We present a domain specific process to enable the verification of observer-based fault detection software. Observer-based fault detection systems, like control systems, yield invariant properties of quadratic types. These quadratic invariants express both safety properties of the software, such as the boundedness of the states, and correctness properties, such as the absence of false alarms from the fault detector. We seek to leverage these quadratic invariants, in an automated way, for the formal verification of the fault detection software. The approach, named the credible autocoding framework, can be characterized as autocoding with proofs. The process starts with the fault detector model, along with its safety and correctness properties, all expressed formally in a synchronous modeling environment such as Simulink. The model is then transformed by a prototype credible autocoder into both code and analyzable annotations for the code. We demonstrate the credible autocoding process on a running example of an output observer fault detector for a 3 degree-of-freedom helicopter control system.
}, keywords = {3-degree-of-freedom helicopter control system, Aerospace Systems, code annotation analysis, control systems, correctness properties, Credible Autocoding, credible autocoding framework, domain specific process, Ellipsoids, Equations, false alarms, Fault detection, fault diagnosis, Formal Methods, formal verification, helicopters, Mathematical model, observer-based fault detection software verification, observers, output observer fault detector model, quadratic-type invariant properties, safety-critical software, Semantics, Simulink, Software, software safety properties, Software Verification, state boundedness, synchronous modeling environment}, doi = {10.1109/ACC.2014.6859131}, author = {Timothy Wang and Ashari, A.E. and Romain Jobredeaux and Eric M. Feron} } @article {cowlagicurvature, title = {Curvature-Bounded Traversability Analysis in Motion Planning for Mobile Robots}, journal = {Transactions on Robotics, IEEE}, volume = {30}, year = {2014}, month = {08/2014}, chapter = {1011}, author = {Raghvendra V. Cowlagi and Panagiotis Tsiotras} } @article {borji2014defending, title = {Defending Yarbus: Eye movements reveal observers{\textquoteright} task}, journal = {Journal of vision}, volume = {14}, year = {2014}, pages = {29}, author = {Ali Borji and Laurent Itti} } @conference {Roberts14iros, title = {Direct Superpixel Labeling for Mobile Robot Navigation Using Learned General Optical Flow Templates}, booktitle = {Intelligent Robots and Systems (IROS), 2014 IEEE/RSJ International Conference on}, year = {2014}, url = {http://frank.dellaert.com/pub/Roberts14iros.pdf}, author = {Richard Roberts and Frank Dellaert} } @conference {6842254, title = {A factor graph approach to estimation and model predictive control on Unmanned Aerial Vehicles}, booktitle = {Unmanned Aircraft Systems (ICUAS), 2014 International Conference on}, year = {2014}, month = {May}, abstract = {In this paper, we present a factor graph framework to solve both estimation and deterministic optimal control problems, and apply it to an obstacle avoidance task on Unmanned Aerial Vehicles (UAVs). We show that factor graphs allow us to consistently use the same optimization method, system dynamics, uncertainty models and other internal and external parameters, which potentially improves the UAV performance as a whole. To this end, we extended the modeling capabilities of factor graphs to represent nonlinear dynamics using constraint factors. For inference, we reformulate Sequential Quadratic Programming as an optimization algorithm on a factor graph with nonlinear constraints. We demonstrate our framework on a simulated quadrotor in an obstacle avoidance application.
}, keywords = {autonomous aerial vehicles, collision avoidance, constraint factors, Cost function, deterministic optimal control, Estimation, external parameters, factor graph approach, graph theory, helicopters, internal parameters, Manifolds, mobile robots, model predictive control, nonlinear constraints, nonlinear dynamical systems, nonlinear dynamics, obstacle avoidance task, optimal control, optimization method, predictive control, quadratic programming, quadrotor, sequential quadratic programming, system dynamics, Trajectory, UAV, uncertainty models, unmanned aerial vehicles, Vectors}, doi = {10.1109/ICUAS.2014.6842254}, author = {Duy-Nguyen Ta and Kobilarov, M. and Frank Dellaert} } @conference {Borji_2014_CVPR, title = {Human vs. Computer in Scene and Object Recognition}, booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)}, year = {2014}, month = {June}, author = {Ali Borji and Laurent Itti} } @conference {Kundu14eccv, title = {Joint Semantic Segmentation and 3D Reconstruction from Monocular Video}, year = {2014}, url = {http://frank.dellaert.com/pubs/Kundu14eccv.pdf}, author = {Abhijit Kundu and Yin Li and Frank Dellaert and Fuxin Li and Jim M. Rehg} } @conference {Ta14iros, title = {Linear-Time Estimation with Tree Assumed Density Filtering and Low-Rank Approximation}, booktitle = {Intelligent Robots and Systems (IROS), 2014 IEEE/RSJ International Conference on}, year = {2014}, url = {http://frank.dellaert.com/pub/Ta14iros.pdf}, author = {Duy-Nguyen Ta and Frank Dellaert} } @conference {6856584, title = {A margin-based approach to threat assessment for autonomous highway navigation}, booktitle = {Intelligent Vehicles Symposium Proceedings, 2014 IEEE}, year = {2014}, month = {June}, abstract = {In this paper we present a new approach to the threat assessment problem for semi-autonomous and fully autonomous vehicles, based on the estimation of the control freedom afforded to a vehicle. Given sensor information available about the surrounding environment, an algorithm is described for identifying fields of safe travel through which the vehicle can safely navigate. Within each candidate field, we then characterize the level of threat, to influence autonomous navigation or driver support inputs. To characterize threat, the fields of safe travel are associated with sets of feasible trajectories generated from a lattice sampled in the vehicle{\textquoteright}s input space. A planner then computes a metric associated with available control freedom from these sampled trajectories. This method potentially allows a semi-autonomous control system to honor safe driver inputs while ensuring safe and robust navigation properties. It could also serve as an input to an autonomous decision-making layer.
}, keywords = {Acceleration, autonomous decision-making layer, autonomous highway navigation, autonomous navigation, control freedom, driver support inputs, fully autonomous vehicles, Lattices, margin-based approach, Measurement, navigation, path planning, road safety, road traffic control, road vehicles, Roads, robust control, robust navigation properties, safe travel, semiautonomous control system, semiautonomous vehicles, sensor information, Sensors, threat assessment problem, threat level, Trajectory, vehicle input space, vehicle navigation safety, Vehicles}, doi = {10.1109/IVS.2014.6856584}, author = {Constantin, A. and Junghee Park and Karl Iagnemma} } @mastersthesis {136, title = {A Margin-based Approach to Vehicle Threat Assessment in a Homotopy Framework for Semi-Autonomous Highway Navigation}, volume = {Master of Science}, year = {2014}, school = {Massachusetts Institute of Technology}, type = {MS}, abstract = {This thesis describes the design of an unified framework for threat assessment and the shared control between a human driver and the onboard controller, based on the notion of fields of safe travel. It allows to perform corridor navigation, trajectory planning, threat assessment and driving assistance in hazardous situations. A new approach to the threat assessment problem is introduced, based upon the estimation of the control freedom afforded to a vehicle. Given sensor information of the surrounding environment, an algorithm first identifies corridors of travel through which the vehicle can safely navigate. The second stage then consists in assessing the potential threat posed to the vehicle in each identified corridor thanks to a metric associated with available control margin. For this purpose, the fields of safe travel are associated with sets of homotopic trajectories generated either from a lattice sampled in the vehicles input space or from a conformal state lattice. This level of threat is the keystone of the system and serves as input to influence autonomous navigation or driver support inputs. The semi-autonomous control system aims to honor safe driver inputs while ensuring safe and robust navigation properties. It ideally operates only during instances of significant threat: it should give a driver full control of the vehicle in {\textquotedblright}low threat{\textquotedblright} situations but apply appropriate levels of computer-controlled actuator effort during {\textquotedblright}high threat{\textquotedblright} situations. This approach preserves the freedom of control of the human driver when he/she remains within a safe navigable corridor, and adjust the vehicle trajectory when its predicted future state falls out of a safe field, or when the lowest threat exceeds some threshold. In fully autonomous mode, this human-inspired motion planning approach ensures collision free navigation and driving comfort.
}, author = {Constantin, A.} } @article {Sanfelice:2014:MPB:2583140.2583488, title = {On Minimum-time Paths of Bounded Curvature with Position-dependent Constraints}, journal = {Automatica}, volume = {50}, year = {2014}, pages = {537{\textendash}546}, keywords = {Dubins vehicle, Hybrid control, Motion planning, optimal control}, issn = {0005-1098}, doi = {10.1016/j.automatica.2013.11.016}, url = {http://dx.doi.org/10.1016/j.automatica.2013.11.016}, author = {Sanfelice, Ricardo G. and Yong, Sze Zheng and Emilio Frazzoli} } @article {10.3389/fncom.2014.00034, title = {Optimal attentional modulation of neural population}, journal = {Frontiers in Computational Neuroscience}, volume = {8}, year = {2014}, abstract = {Top-down attention has often been separately studied in the contexts of either optimal population coding or biasing of visual search. Yet, both are intimately linked, as they entail optimally modulating sensory variables in neural populations according to top-down goals. Designing experiments to probe top-down attentional modulation is difficult because non-linear population dynamics are hard to predict in the absence of a concise theoretical framework. Here, we describe a unified framework that encompasses both contexts. Our work sheds light onto the ongoing debate on whether attention modulates neural response gain, tuning width, and/or preferred feature. We evaluate the framework by conducting simulations for two tasks: (1) classification (discrimination) of two stimuli sa and sb and (2) searching for a target T among distractors D. Results demonstrate that all of gain, tuning, and preferred feature modulation happen to different extents, depending on stimulus conditions and task demands. The theoretical analysis shows that task difficulty (linked to difference Δ between sa and sb, or T, and D) is a crucial factor in optimal modulation, with different effects in discrimination vs. search. Further, our framework allows us to quantify the relative utility of neural parameters. In easy tasks (when Δ is large compared to the density of the neural population), modulating gains and preferred features is sufficient to yield nearly optimal performance; however, in difficult tasks (smaller Δ), modulating tuning width becomes necessary to improve performance. This suggests that the conflicting reports from different experimental studies may be due to differences in tasks and in their difficulties. We further propose future electrophysiology experiments to observe different types of attentional modulation in a same neuron.
}, issn = {1662-5188}, doi = {10.3389/fncom.2014.00034}, url = {http://www.frontiersin.org/computational_neuroscience/10.3389/fncom.2014.00034/abstract}, author = {Ali Borji and Laurent Itti} } @mastersthesis {bialkowski2014optimizations, title = {Optimizations for sampling-based motion planning algorithms}, year = {2014}, school = {Massachusetts Institute of Technology}, type = {phd}, author = {Joshua Bialkowski} } @inbook {128, title = {Real-Time Near-Optimal Feedback Control of Aggressive Vehicle Maneuvers}, booktitle = {Optimization and Optimal Control in Automotive Systems}, series = {Lecture Notes in Control and Information Sciences}, volume = {455}, year = {2014}, pages = {109-129}, publisher = {Springer International Publishing}, organization = {Springer International Publishing}, isbn = {978-3-319-05370-7}, doi = {10.1007/978-3-319-05371-4_7}, url = {http://dx.doi.org/10.1007/978-3-319-05371-4_7}, author = {Panagiotis Tsiotras and Diaz, RicardoSanz} } @mastersthesis {142, title = {Sampling-based Algorithms for Stochastic Optimal Control}, volume = {PhD}, year = {2014}, school = {Massachusetts Institute of Technology}, type = {PhD}, abstract = {Controlling dynamical systems in uncertain environments is fundamental and essential in several fields, ranging from robotics, healthcare to economics and finance. In these applications, the required tasks can be modeled as continuous-time, continuous-space stochastic optimal control problems. Moreover, risk management is an important requirement of such problems to guarantee safety during the execution of control policies. However, even in the simplest version, finding closed-form or exact algorith- mic solutions for stochastic optimal control problems is computationally challenging.
The main contribution of this thesis is the development of theoretical foundations, and provably-correct and efficient sampling-based algorithms to solve stochastic op- timal control problems in the presence of complex risk constraints.
In the first part of the thesis, we consider the mentioned problems without risk constraints. We propose a novel algorithm called the incremental Markov Decision Process (iMDP) to compute incrementally any-time control policies that approximate arbitrarily well an optimal policy in terms of the expected cost. The main idea is to generate a sequence of finite discretizations of the original problem through random sampling of the state space. At each iteration, the discretized problem is a Markov Decision Process that serves as an incrementally refined model of the original problem. We show that the iMDP algorithm guarantees asymptotic optimality while maintaining low computational and space complexity.
In the second part of the thesis, we consider risk constraints that are expressed as either bounded trajectory performance or bounded probabilities of failure. For the former, we present the first extended iMDP algorithm to approximate arbitrarily well an optimal feedback policy of the constrained problem. For the latter, we present a martingale approach that diffuses a risk constraint into a martingale to construct time-consistent control policies. The martingale stands for the level of risk tolerance that is contingent on available information over time. By augmenting the system dynamics with the martingale, the original risk-constrained problem is transformed into a stochastic target problem. We present the second extended iMDP algorithm to approximate arbitrarily well an optimal feedback policy of the original problem by sampling in the augmented state space and computing proper boundary values for the reformulated problem. In both cases, sequences of policies returned from the extended algorithms are both probabilistically sound and asymptotically optimal.
The effectiveness of these algorithms is demonstrated on robot motion planning and control problems in cluttered environments in the presence of process noise.
}, author = {Vu Anh Huynh} } @book {li2014secrets, title = {The secrets of salient object segmentation}, year = {2014}, publisher = {CVPR}, organization = {CVPR}, author = {Yin Li and Hou, Xiaodi and Koch, Christof and Jim M. Rehg and Yuille, A} } @article {ta2014vistas, title = {Vistas and parallel tracking and mapping with Wall{\textendash}Floor Features: Enabling autonomous flight in man-made environments}, journal = {Robotics and Autonomous Systems}, year = {2014}, author = {Duy-Nguyen Ta and Kyel Ok and Frank Dellaert} } @conference {6751224, title = {Analysis of Scores, Datasets, and Models in Visual Saliency Prediction}, booktitle = {Computer Vision (ICCV), 2013 IEEE International Conference on}, year = {2013}, month = {Dec}, abstract = {Significant recent progress has been made in developing high-quality saliency models. However, less effort has been undertaken on fair assessment of these models, over large standardized datasets and correctly addressing confounding factors. In this study, we pursue a critical and quantitative look at challenges (e.g., center-bias, map smoothing) in saliency modeling and the way they affect model accuracy. We quantitatively compare 32 state-of-the-art models (using the shuffled AUC score to discount center-bias) on 4 benchmark eye movement datasets, for prediction of human fixation locations and scan path sequence. We also account for the role of map smoothing. We find that, although model rankings vary, some (e.g., AWS, LG, AIM, and HouNIPS) consistently outperform other models over all datasets. Some models work well for prediction of both fixation locations and scan path sequence (e.g., Judd, GBVS). Our results show low prediction accuracy for models over emotional stimuli from the NUSEF dataset. Our last benchmark, for the first time, gauges the ability of models to decode the stimulus category from statistics of fixations, saccades, and model saliency values at fixated locations. In this test, ITTI and AIM models win over other models. Our benchmark provides a comprehensive high-level picture of the strengths and weaknesses of many popular models, and suggests future research directions in saliency modeling.
}, keywords = {Accuracy, AIM models, Analytical models, benchmark eye movement datasets, Benchmark testing, computer vision, confounding factors, emotional stimuli, eye movements, Face, fair assessment, fixations, gaze tracking, high level picture, high quality saliency models, human fixation locations, model accuracy, model benchmarking, model rankings, model saliency values, NUSEF dataset, Predictive models, quantitative look, saccades, saliency, saliency modeling, scan path sequence, scores, Smoothing methods, standardized datasets, stimulus category, visual attention, visual saliency prediction, Visualization}, doi = {10.1109/ICCV.2013.118}, author = {Ali Borji and Tavakoli, H.R. and Dicky N. Sihite and Laurent Itti} } @conference {borji2013bayesian, title = {Bayesian optimization explains human active search}, booktitle = {Advances in Neural Information Processing Systems}, year = {2013}, author = {Ali Borji and Laurent Itti} } @conference {Chang_etal13iros, title = {Beobot 2.0: Autonomous Mobile Robot Localization and Navigation in Outdoor Pedestrian Environment}, booktitle = {Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Video Track}, year = {2013}, month = {Nov}, abstract = {We present Beobot 2.0, an autonomous mobile robot designed to operate in unconstrained urban environments. The goal of the project is to create service robots that can be deployed for various tasks that require long range travel. Over the past two years, Beobot has successfully traversed various paths across the USC campus, demonstrating its robustness in recognizing and following different types of roads, avoiding obstacles such as pedestrians and service vehicles, and finding its way to the goal.
}, author = {C.K. Chang and Christian Siagian and Laurent Itti} } @inbook {Itti_Borji13att, title = {Computational models: Bottom-up and top-down aspects}, booktitle = {The Oxford Handbook of Attention}, year = {2013}, pages = {1-20}, author = {Laurent Itti and Ali Borji} } @inbook {149, title = {Deep Learning on Natural Viewing Behaviors to Differentiate Children with Fetal Alcohol Spectrum Disorder}, booktitle = {Intelligent Data Engineering and Automated Learning {\textendash} IDEAL 2013}, series = {Lecture Notes in Computer Science}, volume = {8206}, year = {2013}, pages = {178-185}, publisher = {Springer Berlin Heidelberg}, organization = {Springer Berlin Heidelberg}, keywords = {clinical populations, deep learning, eye movement, natural scenes, sparse representation}, isbn = {978-3-642-41277-6}, doi = {10.1007/978-3-642-41278-3_22}, url = {http://dx.doi.org/10.1007/978-3-642-41278-3_22}, author = {Tseng, Po-He and Paolozza, Angelina and Munoz, DouglasP. and Reynolds, JamesN. and Laurent Itti} } @conference {6696980, title = {Finding planes in LiDAR point clouds for real-time registration}, booktitle = {Intelligent Robots and Systems (IROS), 2013 IEEE/RSJ International Conference on}, year = {2013}, month = {Nov}, abstract = {We present a robust plane finding algorithm that when combined with plane-based frame-to-frame registration gives accurate real-time pose estimation. Our plane extraction is capable of handling large and sparse datasets such as those generated from spinning multi-laser sensors such as the Velodyne HDL-32E LiDAR. We test our algorithm on frame-to-frame registration in a closed-loop indoor path comprising 827 successive 3D laser scans (over 57 million points), using no additional information (e.g., odometry, IMU). Our algorithm outperforms, in both accuracy and time, three state-of-the-art methods, based on iterative closest point (ICP), plane-based randomized Hough transform, and planar region growing.
}, keywords = {3D laser scans, closed loop systems, closed-loop indoor path, distance measurement, Equations, Hough transform, Hough transforms, ICP, image registration, iterative closest point, Iterative closest point algorithm, iterative methods, Laser radar, Laser theory, LiDAR point clouds, Measurement by laser beam, odometry, optical radar, plane-based frame-to-frame registration, pose estimation, real-time pose estimation, real-time registration, real-time systems, robust plane finding, Sensors, sparse datasets, spinning multilaser sensors, Vectors, Velodyne HDL-32E LiDAR}, doi = {10.1109/IROS.2013.6696980}, author = {Grant, W.S. and Voorhies, R.C. and Laurent Itti} } @conference {Bialkowski.Otte.ea:IROS13, title = {Free-configuration Biased Sampling for Motion Planning}, booktitle = {IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS)}, year = {2013}, keywords = {Motion planning}, url = {http://ares.lids.mit.edu/papers/Bialkowski.Otte.ea.IROS13.pdf}, author = {Joshua Bialkowski and M. W. Otte and Emilio Frazzoli} } @article {6507273, title = {The intelligent copilot: A constraint-based approach to shared-adaptive control of ground vehicles}, journal = {Intelligent Transportation Systems Magazine, IEEE}, volume = {5}, year = {2013}, month = {Summer}, pages = {45-54}, abstract = {This work presents a new approach to semi-autonomous vehicle hazard avoidance and stability control, based on the design and selective enforcement of constraints. This differs from traditional approaches that rely on the planning and tracking of paths and facilitates "minimally-invasive" control for human-machine systems. Instead of forcing a human operator to follow an automation-determined path, the constraint-based approach identifies safe homotopies, and allows the operator to navigate freely within them, introducing control action only as necessary to ensure that the vehicle does not violate safety constraints. This method evaluates candidate homotopies based on "restrictiveness," rather than traditional measures of path goodness, and designs and enforces requisite constraints on the human{\textquoteright}s control commands to ensure that the vehicle never leaves the controllable subset of a desired homotopy. This paper demonstrates the approach in simulation and characterizes its effect on human teleoperation of unmanned ground vehicles via a 20-user, 600-trial study on an outdoor obstacle course. Aggregated across all drivers and experiments, the constraintbased control system required an average of 43\% of the available control authority to reduce collision frequency by 78\% relative to traditional teleoperation, increase average speed by 26\%, and moderate operator steering commands by 34\%.
}, keywords = {adaptive control, collision avoidance, collision frequency, constraint design, constraint-based approach, constraint-based control system, control action, control system synthesis, control systems, Hazards, human teleoperation, human-machine system, intelligent control, intelligent copilot, Man machine systems, man-machine systems, minimally-invasive control, mobile robots, navigation, operator steering command, path goodness, path planning, path tracking, remotely operated vehicles, restrictiveness homotopy, road safety, road traffic control, safety constraint, selective constraint enforcement, semi-autonomous vehicle hazard avoidance, shared-adaptive control, stability, Stability analysis, stability control, unmanned ground vehicle, Vehicle driving, Vehicle safety}, issn = {1939-1390}, doi = {10.1109/MITS.2013.2247796}, author = {Anderson, S. and Sisir Karumanchi and Karl Iagnemma and Walker, J.M.} } @mastersthesis {fathi2013learning, title = {Learning Descriptive Models of Objects and Activities from Egocentric Video}, year = {2013}, author = {Alireza Fathi} } @conference {Li_2013_ICCV, title = {Learning to Predict Gaze in Egocentric Video}, booktitle = {The IEEE International Conference on Computer Vision (ICCV)}, year = {2013}, month = {December}, author = {Yin Li and Alireza Fathi and Jim M. Rehg} } @conference {Siagian_etal13icra, title = {Mobile Robot Navigation System in Outdoor Pedestrian Environment Using Vision-Based Road Recognition}, booktitle = {Proc. IEEE International Conference on Robotics and Automation (ICRA)}, year = {2013}, month = {May}, abstract = {We present a mobile robot navigation system guided by a novel vision-based road recognition approach. The system represents the road as a set of lines extrapolated from the detected image contour segments. These lines enable the robot to maintain its heading by centering the vanishing point in its field of view, and to correct the long term drift from its original lateral position. We integrate odometry and our visual road recognition system into a grid-based local map that estimates the robot pose as well as its surroundings to generate a movement path. Our road recognition system is able to estimate the road center on a standard dataset with 25,076 images to within 11.42\ cm (with respect to roads at least 3\ m wide). It outperforms three other state-of-the-art systems. In addition, we extensively test our navigation system in four busy college campus environments using a wheeled robot. Our tests cover more than 5\ km of autonomous driving without failure. This demonstrates robustness of the proposed approach against challenges that include occlusion by pedestrians, non-standard complex road markings and shapes, shadows, and miscellaneous obstacle objects.
}, author = {Christian Siagian and C.K. Chang and Laurent Itti} } @conference {6630606, title = {Mobility Erosion: High speed motion safety for mobile robots operating in off-road terrain}, booktitle = {Robotics and Automation (ICRA), 2013 IEEE International Conference on}, year = {2013}, month = {May}, abstract = {This paper addresses the problem of ensuring mobile robot motion safety when reacting to soft and hard hazards in a static environment. The work is aimed at off-road navigation for mobile ground robots where soft hazards are posed by varying terrain conditions (e.g. deformable soil, slopes, vegetation). Soft hazards pose operating constraints (i.e. speed limits) to the mobile robot that need to be satisfied to ensure motion safety. This paper presents a new morphological erosion operator that generalizes binary obstacle growing to mobility space (the space of speed limits) to deal with both hard and soft hazards seamlessly. This ensures that topological constraints due to vehicle size as well as momentum are taken into account, and leads to a straight-forward approach to generalize the concept of {\textquoteleft}regions of inevitable collision{\textquoteright} for soft hazards.
}, keywords = {binary obstacle, Gray-scale, hard hazards, Hazards, high speed motion safety, mobile ground robots, mobile robot motion safety, mobile robots, mobility erosion, mobility space, morphological erosion operator, off-road navigation, off-road terrain, off-road vehicles, operating constraints, region of inevitable collision concept, soft hazards, Space vehicles, speed limits, static environment, topological constraints, Topology, varying terrain conditions, vehicle dynamics, vehicle size}, doi = {10.1109/ICRA.2013.6630606}, author = {Sisir Karumanchi and Karl Iagnemma and Scheding, S.} } @conference {335, title = {Monocular Parallel Tracking and Mapping with Odometry Fusion for MAV Navigation in Feature-lacking Environments}, booktitle = {IEEE/RSJ IROS{\textquoteright}13 International Workshop on Vision-based Closed-Loop Control and Navigation of Micro Helicopters in GPS-denied Environments}, year = {2013}, month = {11/2013}, address = {Tokyo, Japan}, url = {http://rpg.ifi.uzh.ch/docs/IROS13workshop/Nguyen.pdf}, author = {Duy-Nguyen Ta and Kyel Ok and Frank Dellaert} } @article {161, title = {Objects do not predict fixations better than early saliency: a re-analysis of Einhauser et al.{\textquoteright}s data.}, journal = {J Vis}, volume = {13}, year = {2013}, month = {2013}, pages = {18}, abstract = {Einh{\"a}user, Spain, and Perona (2008) explored an alternative hypothesis to saliency maps (i.e., spatial image outliers) and claimed that "objects predict fixations better than early saliency." To test their hypothesis, they measured eye movements of human observers while they inspected 93 photographs of common natural scenes (Uncommon Places dataset by Shore, Tillman, \& Schmidt-Wulen 2004; Supplement Figure S4). Subjects were asked to observe an image and, immediately afterwards, to name objects they saw (remembered). Einh{\"a}user et al. showed that a map made of manually drawn object regions, each object weighted by its recall frequency, predicts fixations in individual images better than early saliency. Due to important implications of this hypothesis, we investigate it further. The core of our analysis is explained here. Please refer to the Supplement for details.
}, keywords = {Female, Fixation, Ocular, Humans, Male, Models, Psychological, Pattern Recognition, Visual}, issn = {1534-7362}, doi = {10.1167/13.10.18}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @article {jung2013line, title = {On-line path generation for unmanned aerial vehicles using b-spline path templates}, journal = {Journal of Guidance, Control, and Dynamics}, volume = {36}, year = {2013}, pages = {1642{\textendash}1653}, author = {Jung, Dongwon and Panagiotis Tsiotras} } @mastersthesis {165, title = {Optical Flow Templates for Mobile Robot Environment Understanding}, volume = {PhD}, year = {2013}, month = {05/2013}, school = {Georgia Institute of Technology}, type = {Doctoral}, abstract = {In this work we develop optical flow templates. In doing so, we introduce a practical tool for inferring robot egomotion and semantic superpixel labeling using optical flow in imaging systems with arbitrary optics. In order to do this we develop valuable understand- ing of geometric relationships and mathematical methods that are useful in interpreting optical flow to the robotics and computer vision communities.
This work is motivated by what we perceive as directions for advancing the current state of the art in obstacle detection and scene understanding for mobile robots. Specif- ically, many existing methods build 3D point clouds, which are not directly useful for autonomous navigation and require further processing. Both the step of building the point clouds and the later processing steps are challenging and computationally intensive. Ad- ditionally, many current methods require a calibrated camera, which introduces calibration challenges and places limitations on the types of camera optics that may be used. Wide- angle lenses, systems with mirrors, and multiple cameras all require different calibration models and can be difficult or impossible to calibrate at all. Finally, current pixel and su- perpixel obstacle labeling algorithms typically rely on image appearance. While image appearance is informative, image motion is a direct effect of the scene structure that deter- mines whether a region of the environment is an obstacle.
The egomotion estimation and obstacle labeling methods we develop here based on optical flow templates require very little computation per frame and do not require building point clouds. Additionally, they do not require any specific type of camera optics, nor a calibrated camera. Finally, they label obstacles using optical flow alone without image appearance.
In this thesis we start with optical flow subspaces for egomotion estimation and de- tection of {\textquotedblleft}motion anomalies{\textquotedblright}. We then extend this to multiple subspaces and develop mathematical reasoning to select between them, comprising optical flow templates. Using these we classify environment shapes and label superpixels. Finally, we show how per- forming all learning and inference directly from image spatio-temporal gradients greatly improves computation time and accuracy.
}, author = {Richard Roberts} } @conference {roberts2013optical, title = {Optical Flow Templates for Superpixel Labeling in Autonomous Robot Navigation}, booktitle = {5th Workshop on Planning, Perception and Navigation for Intelligent Vehicles (PPNIV13)}, year = {2013}, author = {Richard Roberts and Frank Dellaert} } @conference {6579835, title = {Optimal motion planning with the half-car dynamical model for autonomous high-speed driving}, booktitle = {American Control Conference (ACC), 2013}, year = {2013}, month = {June}, abstract = {We discuss an implementation of the RRT* optimal motion planning algorithm for the half-car dynamical model to enable autonomous high-speed driving. To develop fast solutions of the associated local steering problem, we observe that the motion of a special point (namely, the front center of oscillation) can be modeled as a double integrator augmented with fictitious inputs. We first map the constraints on tire friction forces to constraints on these augmented inputs, which provides instantaneous, state-dependent bounds on the curvature of geometric paths feasibly traversable by the front center of oscillation. Next, we map the vehicle{\textquoteright}s actual inputs to the augmented inputs. The local steering problem for the half-car dynamical model can then be transformed to a simpler steering problem for the front center of oscillation, which we solve efficiently by first constructing a curvature-bounded geometric path and then imposing a suitable speed profile on this geometric path. Finally, we demonstrate the efficacy of the proposed motion planner via numerical simulation results.
}, keywords = {Acceleration, augmented input, autonomous high-speed driving, bounded geometric path curvature, constraint mapping, double integrator augmentation, Friction, half car dynamical model, Heuristic algorithms, mobile robots, numerical simulation, optimal control, oscillation front center, oscillations, path planning, Planning, robot dynamics, RRT* optimal motion planning algorithm, state dependent bound, steering problem, steering systems, tire friction force, Tires, Trajectory, tyres, vehicle actual input mapping, Vehicles}, doi = {10.1109/ACC.2013.6579835}, author = {Jeong Hwan Jeon and Raghvendra V. Cowlagi and Steven C. Peters and Sertac Karaman and Emilio Frazzoli and Panagiotis Tsiotras and Karl Iagnemma} } @article {borji2013quantitative, title = {Quantitative analysis of human-model agreement in visual saliency modeling: A comparative study}, journal = {Image Processing, IEEE Transactions on}, volume = {22}, year = {2013}, pages = {55{\textendash}69}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @article {6548082, title = {Reliably Creating Collision Avoidance Advisories in Piloted Simulations}, journal = {Human-Machine Systems, IEEE Transactions on}, volume = {43}, year = {2013}, pages = {416-420}, abstract = {We demonstrate a novel method to reliably generate collision avoidance advisories, in piloted simulations, by the widely used traffic alert and collision avoidance system (TCAS). The TCAS advisory issued to a pilot is highly sensitive to the trajectory of an intruder aircraft relative to the ownship flown by the pilot. In realistic piloted simulations, a prescripted intruder trajectory will not reliably result in the relative dynamics that lead to a desired TCAS advisory. Further, the complexity of the TCAS logic requires a novel method for mapping trajectories to the range of possible advisories. We propose to use a rapidly exploring random tree algorithm in large-scale fast-time simulations to establish the mapping between the space of relative trajectories and TCAS advisories. These trajectories are then created in piloted simulations through guidance algorithms. Results demonstrate the ease of use and robustness of this method, and its potential for pilot training and for research and development.
}, keywords = {air traffic control, aircraft, collision avoidance, guidance algorithms, human factors, intruder aircraft, intruder trajectory, mapping trajectories, pilot training, piloted simulations, relative dynamics, reliably creating collision avoidance advisories, simulation, TCAS advisory, TCAS logic, traffic alert and collision avoidance system, trajectory control}, issn = {2168-2291}, doi = {10.1109/TSMC.2013.2258668}, author = {Popescu, V. and Cleveland, W.P. and Boidot, E. and Pritchett, A.R. and Eric M. Feron and Zoetrum, J.J.} } @conference {6580549, title = {Sampling-based algorithms for continuous-time POMDPs}, booktitle = {American Control Conference (ACC), 2013}, year = {2013}, month = {June}, abstract = {This paper focuses on a continuous-time, continuous-space formulation of the stochastic optimal control problem with nonlinear dynamics and observation noise. We lay the mathematical foundations to construct, via incremental sampling, an approximating sequence of discrete-time finite-state partially observable Markov decision processes (POMDPs), such that the behavior of successive approximations converges to the behavior of the original continuous system in an appropriate sense. We also show that the optimal cost function and control policies for these POMDP approximations converge almost surely to their counterparts for the underlying continuous system in the limit. We demonstrate this approach on two popular continuous-time problems, viz., the Linear-Quadratic-Gaussian (LQG) control problem and the light-dark domain problem.
}, keywords = {Approximation methods, approximation theory, continuous time systems, continuous-space formulation, continuous-time formulation, continuous-time POMDP, Cost function, cost optimal control, discrete time systems, discrete-time finite-state partially observable Markov decision processes, incremental sampling-based algorithms, light-dark domain problem, linear quadratic Gaussian control, linear-quadratic-Gaussian control problem, LQG, Manganese, Markov processes, mathematical foundations, nonlinear control systems, nonlinear dynamics, observation noise, optimal cost control policies, optimal cost function policies, sampling methods, sequence approximation, stochastic optimal control problem, stochastic systems, Tin, Trajectory}, doi = {10.1109/ACC.2013.6580549}, author = {Chaudhari, P. and Sertac Karaman and Hsu, D. and Emilio Frazzoli} } @conference {DBLP:conf/icra/KaramanF13, title = {Sampling-based Optimal Motion Planning for Non-holonomic Dynamical Systems}, booktitle = {ICRA}, year = {2013}, author = {Sertac Karaman and Emilio Frazzoli} } @conference {Noori_Itti13schema, title = {Schema-driven, space-supported random accessible memory systems for manipulation of symbolic working memory}, booktitle = {Proc. 35th annual conference of the cognitive science society}, year = {2013}, month = {Aug}, abstract = {We present an execution model for manipulation of working memory content during intellectual symbolic working memory tasks, which allows random access of WM content through a schema-operated sensory-motor spatial working memory. The core concept of this framework is binding symbolic items to spatial locations which are accessible {\em via} selective mechanisms of attention in space. An operational schema implements basic WM management operations such as insertion, deletion and fetching through sequences of shifts in spatial attention towards registry locations. We apply the model to a serial recall task (both forward and backward orders). We show that the model provides a better fit to human data in backward recall compared to forward recall, which conforms with the evidence for leveraging spatial strategies for backward recall and phonological strategies for forward recall in normal subjects. We discuss additional possible implications of our model and its assumption of spatial organization of WM content and access through shifts of attention.
}, author = {N. Noori and Laurent Itti} } @conference {6697178, title = {Situation awareness via sensor-equipped eyeglasses}, booktitle = {Intelligent Robots and Systems (IROS), 2013 IEEE/RSJ International Conference on}, year = {2013}, month = {Nov}, abstract = {New smartphone technologies are emerging which combine head-mounted displays (HMD) with standard functions such as receiving phone calls, emails, and helping with navigation. This opens new opportunities to explore cyber robotics algorithms (robotics sensors and human motor plant). To make these devices more adaptive to the environmental conditions, user behavior, and user preferences, it is important to allow the sensor-equipped devices to efficiently adapt and respond to user activities (e.g., disable incoming phone calls in an elevator, activate video recording while car driving). This paper hence presents a situation awareness system (SAS) for head-mounted smartphones. After collecting data from inertial sensors (accelerometers, gyroscopes), and video data (camera), SAS performs activity classification in three steps. Step 1 transforms inertial sensor data into a head orientation-independent and stable normalized coordinate system. Step 2 extracts critical features (statistical, physical, GIST). Step 3 classifies activities (Naive Bayes classifier), distinguishes between environments (Support Vector Machine), and finally combines both results (Hidden Markov Model) for further improvement. SAS has been implemented on a sensor-equipped eyeglasses prototype and achieved high accuracy (81.5\%) when distinguishing between 20 real-world activities.
}, keywords = {accelerometers, activity classification, Bayes methods, camera, Cameras, critical feature extraction, cyber robotics algorithms, Databases, environmental conditions, feature extraction, gyroscopes, Head, head orientation-independent system, head-mounted displays, head-mounted smartphones, helmet mounted displays, hidden Markov model, hidden Markov models, HMD, human motor plant, inertial sensor data, inertial sensors, naive Bayes classifier, normalized coordinate system, pattern classification, robotics sensors, robots, SAS, Sensor systems, sensor-equipped eyeglasses prototype, situation awareness system, smart phones, smartphone technologies, support vector machine, support vector machines, ubiquitous computing, user behavior, user preferences, video data}, doi = {10.1109/IROS.2013.6697178}, author = {Windau, J. and Laurent Itti} } @article {berg2013spatiotemporal, title = {Spatiotemporal processing of saliency signals in the primate: a behavioral and neurophysiological investigation}, year = {2013}, author = {Berg, David J} } @conference {6580156, title = {Time-optimal vehicle posture control to mitigate unavoidable collisions using conventional control inputs}, booktitle = {American Control Conference (ACC), 2013}, year = {2013}, month = {June}, abstract = {This paper analyzes the mitigation of an unavoidable T-bone collision, where an {\textquotedblleft}intelligent{\textquotedblright} vehicle executes an aggressive time-optimal rotation to achieve a favorable relative orientation with another vehicle prior to impact. The current paper extends the previous work by the authors on this problem, by modeling additional vehicle dynamics (neglected in the prior work) and by utilizing conventionally available control commands (that is, steering, braking, handbrake) for the maneuvering vehicle. The commands can either be applied directly by a trained driver, or (as in the majority of cases) can be executed with the help of a combination of an Active Front Steering (AFS) and an Electronic Stability Control (ESC) system onboard the vehicle. The optimal yaw rotation maneuver is analyzed for different initial speeds on both dry and wet asphalt. The results confirm the existence of an {\textquotedblleft}option zone{\textquotedblright} for some cases, within which such an aggressive maneuver may be possible and perhaps even preferable to straight line braking.
}, keywords = {active front steering, AFS, asphalt, automobiles, braking, conventional control inputs, dry asphalt, electronic stability control system, ESC, Friction, intelligent vehicle, optimal yaw rotation maneuver, position control, Safety, steering systems, straight line braking, time optimal control, time-optimal rotation, time-optimal vehicle posture control, Tires, Torque, unavoidable T-bone collision mitigation, vehicle dynamics, Vehicles, wet asphalt, Wheels}, doi = {10.1109/ACC.2013.6580156}, author = {Chakraborty, I. and Panagiotis Tsiotras and Sanz Diaz, R.} } @conference {Noori_Itti13traces, title = {Traces of intellectual working memory tasks on visual-spatial short-term memory}, booktitle = {Proc. 35th annual conference of the cognitive science society}, year = {2013}, month = {Aug}, abstract = {We measured the sensitivity in detecting a change in the location of one of two visual targets over a short period of time to investigate the impact of a secondary intellectual symbolic WM task on the visuospatial short-term memory, in a dual-task paradigm. We observed that engaging in a WM task that involves manipulation of symbolic information impacts the ability to detect a location change, and this impact does not change when more time is allocated to the WM task. Furthermore, we observed that the impact of a mental sorting task on the ability to detect location changes is spatially selective to the horizontal orientation. Our results suggest a possible role for sensory-motor working memory, which supports perception-action schemas in manipulation of information during the intellectual symbolic working memory tasks.
}, author = {N. Noori and Laurent Itti} } @conference {DBLP:conf/icra/ArslanT13, title = {Use of Relaxation Methods in Sampling-Based Algorithms for Optimal Motion Planning}, booktitle = {ICRA 2013}, year = {2013}, author = {Oktay Arslan and Panagiotis Tsiotras} } @conference {Li2013SegTrack, title = {Video Segmentation by Tracking Many Figure-Ground Segments}, booktitle = {IEEE International Conference on Computer Vision}, year = {2013}, author = {Fuxin Li and Taeyoung Kim and Ahmad Humayun and David Tsai and Jim M. Rehg} } @conference {Li_2013_ICCV, title = {Video Segmentation by Tracking Many Figure-Ground Segments}, booktitle = {The IEEE International Conference on Computer Vision (ICCV)}, year = {2013}, month = {December}, author = {Fuxin Li and Taeyoung Kim and Ahmad Humayun and David Tsai and Jim M. Rehg} } @article {borji2013stands, title = {What stands out in a scene? A study of human explicit saliency judgment}, journal = {Vision research}, volume = {91}, year = {2013}, pages = {62{\textendash}77}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @article {ieee.tran.systems.man.cybernetics/Borji13, title = {Where to look next? Modeling Top-down Visual Attention in Complex Interactive Environments}, journal = {IEEE Transactions on Systems, Man, and Cybernetics, Part A}, volume = {44}, year = {2013}, month = {05/2014}, pages = {15}, chapter = {523}, abstract = {Several visual attention models have been proposed for describing eye movements over simple stimuli and tasks such as free viewing or visual search. Yet, to date, there exists no computational framework that can reliably mimic human gaze behavior in more complex environments and tasks such as urban driving. In addition, benchmark datasets, scoring techniques, and top-down model architectures are not yet well understood. In this paper, we describe new task-dependent approaches for modeling top-down overt visual attention based on graphical models for probabilistic inference and reasoning. We describe a dynamic Bayesian network that infers probability distributions over attended objects and spatial locations directly from observed data. Probabilistic inference in our model is performed over object-related functions that are fed from manual annotations of objects in video scenes or by state-of-the-art object detection/recognition algorithms. Evaluating over approximately 3 h (approximately 315 000 eye fixations and 12 000 saccades) of observers playing three video games (time-scheduling, driving, and flight combat), we show that our approach is significantly more predictive of eye fixations compared to: 1) simpler classifier-based models also developed here that map a signature of a scene (multimodal information from gist, bottom-up saliency, physical actions, and events) to eye positions; 2) 14 state-of-the-art bottom-up saliency models; and 3) brute-force algorithms such as mean eye position. Our results show that the proposed model is more effective in employing and reasoning over spatio-temporal visual data compared with the state-of-the-art.
}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @conference {Noori_Itti13you, title = {Where what you count is what really counts}, booktitle = {Proc. 35th annual conference of the cognitive science society}, year = {2013}, month = {Aug}, abstract = {We noticed that human subjects were notably faster and more accurate in concurrent counting of three location-based events while they ignored the identity of targets, compared to concurrent counting of three identity-based events while they ignored the locations. In a control experiment, subjects performed a location-based triple counting task, while now also paying attention to the target identity. This did not incur any additional cost, compared to the cost of the location-based counting. Performing each of these tasks relies on maintaining three running numerical counters, and on switching between them to increase each one. Our results suggest that switching between these counters has lower cost when they are associated to spatial locations, compared to when they are associated to identities. This difference is not affected when additionally processing the identity of items. We argue that this might be related to the advantage of the space in switching attention between internal representations.
}, author = {N. Noori and Laurent Itti} } @conference {Beall12fusion, title = {Attitude Heading Reference System with Rotation-Aiding Visual Landmarks}, year = {2012}, url = {http://frank.dellaert.com/pubs/Beall12fusion.pdf}, author = {Chris Beall and Duy-Nguyen Ta and Kyel Ok and Frank Dellaert} } @conference {DBLP:conf/eccv/RobertsSD12, title = {Classifying Environment Shape from Image Motion for Mobile Robot Path Following}, booktitle = {ECCV}, year = {2012}, author = {Richard Roberts and Natesh Srinivasan and Frank Dellaert} } @conference {DBLP:conf/cvpr/RobertsSD12, title = {Classifying Large-Scale Environment Shapes with Linear Optical Flow Templates}, booktitle = {CVPR}, year = {2012}, author = {Richard Roberts and Natesh Srinivasan and Frank Dellaert} } @conference {conf/ivs/AndersonKI12, title = {Constraint-based planning and control for safe, semi-autonomous operation of vehicles.}, booktitle = {Intelligent Vehicles Symposium}, year = {2012}, publisher = {IEEE}, organization = {IEEE}, keywords = {dblp}, isbn = {978-1-4673-2119-8}, url = {http://dblp.uni-trier.de/db/conf/ivs/ivs2012.html$\#$AndersonKI12}, author = {Anderson, S. and Sisir Karumanchi and Karl Iagnemma} } @conference {6232153, title = {Constraint-based planning and control for safe, semi-autonomous operation of vehicles}, booktitle = {Intelligent Vehicles Symposium (IV), 2012 IEEE}, year = {2012}, month = {June}, abstract = {This paper presents a new approach to semi-autonomous vehicle hazard avoidance and stability control, based on the design and selective enforcement of constraints. This differs from traditional approaches that rely on the planning and tracking of paths. This emphasis on constraints facilitates {\textquotedblleft}minimally-invasive{\textquotedblright} control for human-machine systems; instead of forcing a human operator to follow an automation-determined path, the constraint-based approach identifies safe homotopies, and allows the operator to navigate freely within them, introducing control action only as necessary to ensure that the vehicle does not violate safety constraints. The method evaluates candidate homotopies based on {\textquotedblleft}restrictiveness{\textquotedblright}, rather than traditional measures of path goodness, and designs and enforces requisite constraints on the human{\textquoteright}s control commands to ensure that the vehicle never leaves the controllable subset of a desired homotopy. Identification of these homotopic classes in off-road environments is performed using geometric constructs. The goodness of competing homotopies and their associated constraints is then characterized using geometric heuristics. Finally, input limits satisfying homotopy and vehicle dynamic constraints are enforced using threat-based feedback mechanisms to ensure that the vehicle avoids collisions and instability while preserving the human operator{\textquoteright}s situational awareness and mental models. The methods developed in this work are shown in simulation and experimentally demonstrated in safe, high-speed teleoperation of an unmanned ground vehicle.
}, keywords = {automation-determined path, collision avoidance, constraint-based approach, constraint-based control, constraint-based planning, geometric constructs, geometric heuristics, hazard avoidance, Hazards, high-speed teleoperation, homotopic classes, human-machine interaction, human-machine systems, Humans, man-machine systems, mental models, minimally-invasive control, navigation, obstacle avoidance, off-road environments, path planning, path tracking, Planning, remotely operated vehicles, road safety, safe vehicle operation, Semi-Autonomous control, semi-autonomous vehicle operation, shared adaptive control, situational awareness, stability, stability control, teleoperation, threat-based feedback mechanisms, Torque, unmanned ground vehicle, unmanned ground vehicles, vehicle dynamic constraints, vehicle dynamics, Vehicles, Wheels}, doi = {10.1109/IVS.2012.6232153}, author = {Anderson, S.} } @article {doi:10.1117/12.919265, title = {Constraint-based semi-autonomy for unmanned ground vehicles using local sensing}, year = {2012}, pages = {83870K-83870K-8}, doi = {10.1117/12.919265}, url = {http://dx.doi.org/10.1117/12.919265}, author = {Anderson, S. and Sisir Karumanchi and Bryan Johnson and Victor Perlin and Mitchell Rohde and Karl Iagnemma} } @conference {DBLP:conf/cvpr/BorjiI12, title = {Exploiting local and global patch rarities for saliency detection}, booktitle = {CVPR}, year = {2012}, author = {Ali Borji and Laurent Itti} } @article {journals/trob/CowlagiT12, title = {Hierarchical Motion Planning With Dynamical Feasibility Guarantees for Mobile Robotic Vehicles.}, journal = {IEEE Transactions on Robotics}, volume = {28}, year = {2012}, pages = {379-395}, keywords = {dblp}, url = {http://dblp.uni-trier.de/db/journals/trob/trob28.html$\#$CowlagiT12}, author = {Raghvendra V. Cowlagi and Panagiotis Tsiotras} } @conference {DBLP:conf/icra/CowlagiT12, title = {Hierarchical Motion Planning with Kinodynamic Feasibility Guarantees: Local trajectory Planning via Model Predictive Control}, booktitle = {ICRA}, year = {2012}, author = {Raghvendra V. Cowlagi and Panagiotis Tsiotras} } @mastersthesis {mit/chaudhari.MSThesis, title = {Incremental Sampling based Algorithms for State Estimation}, volume = {Master of Science in Aeronautics and Astronautics}, year = {2012}, school = {Massachusetts Institute of Technology}, type = {Master}, author = {Chaudhari, P.} } @article {Marat_Itti12vc, title = {Influence of the Amount of Context Learned for Improving Object Classification when Simultaneously Learning Object and Contextual Cues}, journal = {Visual Cognition}, volume = {20}, year = {2012}, month = {Apr}, pages = {580-602}, type = {mod;cv;sc}, abstract = {Humans use visual context to improve object recognition. Yet, many machine vision algorithms still focus on local object features, discarding surrounding features as unwanted clutter. Here we study the impact of learning contextual cues while training an object classifier. In a new image database with 10 object categories and 28,800 images, objects were presented in contextual or uniform backgrounds. Both the fraction of contextual backgrounds during training and the spatial extent of context were analysed. Local object features and broader context features were extracted by two biologically inspired algorithms, previously used for object and scene classification, respectively: HMAX, applied to a tight window around every object, and a {\textquoteright}Gist{\textquoteright} algorithm, applied to a larger yet still localized window. The descriptors from both algorithms were combined and processed by a Support Vector Machine. The recognition rate increased from 29 percent, without contextual cues, to 43 percent for objects presented in their context.
}, author = {Sophie Marat and Laurent Itti} } @conference {DBLP:conf/cvpr/WangLJSWR12, title = {Learning sparse covariance patterns for natural scenes}, booktitle = {CVPR}, year = {2012}, author = {Liwei Wang and Yin Li and Jiaya Jia and Jian Sun and David Wipf and Jim M. Rehg} } @conference {Borji_etal12icra, title = {Modeling the Influence of Action on Spatial Attention in Visual Interactive Environments}, booktitle = {Proc. IEEE International Conference on Robotics and Automation (ICRA)}, year = {2012}, month = {May}, abstract = {A large number of studies have been reported on top-down influences of visual attention. However, less progress have been made in understanding and modeling its mechanisms in real-world tasks. In this paper, we propose an approach for learning spatial attention taking into account influences of physical actions on top-down attention. For this purpose, we focus on interactive visual environments (video games) which are modest real-world simulations, where a player has to attend to certain aspects of visual stimuli and perform actions to achieve a goal. The basic idea is to learn a mapping from current mental state of the game player, represented by past actions and observations, to its gaze fixation. A data-driven approach is followed where we train a model from the data of some players and test it over a new subject. In particular, two contributions this paper makes are: 1) employing multimodal information including mean eye position, gist of a scene, physical actions, bottom-up saliency, and tagged events for state representation and 2) analysis of different methods of combining bottom-up and top-down influences. Comparing with other top-down task-driven and bottom-up spatio-temporal models, our approach shows higher NSS scores in predicting eye positions.
}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @conference {DBLP:conf/icra/BorjiSI12, title = {Modeling the influence of action on spatial attention in visual interactive environments}, booktitle = {ICRA}, year = {2012}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @article {DBLP:journals/tsmc/CowlagiT12, title = {Multiresolution Motion Planning for Autonomous Agents via Wavelet-Based Cell Decompositions}, journal = {IEEE Transactions on Systems, Man, and Cybernetics, Part B}, volume = {42}, year = {2012}, pages = {1455-1469}, author = {Raghvendra V. Cowlagi and Panagiotis Tsiotras} } @conference {Borji_etal12aaai, title = {An Object-based Bayesian Framework for Top-down Visual Attention}, booktitle = {Proc. Twenty-Sixth AAAI Conference on Artificial Intelligence (AAAI-12), Toronto, Canada}, year = {2012}, month = {Aug}, abstract = {We introduce a new task-independent framework to model top-down overt visual attention based on graphical models for probabilistic inference and reasoning. We describe a Dynamic Bayesian Network (DBN) that infers probability distributions over attended objects and spatial locations directly from observed data. Probabilistic inference in our model is performed over object-related functions which are fed from manual an- notations of objects in video scenes or by state-of-the-art object detection models. Evaluating over appx. 3 hours (appx. 315,000 eye fixations and 12,600 saccades) of observers playing 3 video games (time-scheduling, driving, and flight combat), we show that our approach is significantly more predictive of eye fixations com- pared to: 1) simpler classifier-based models also developed here that map a signature of a scene (multi-modal information from gist, bottom-up saliency, physical actions, and events) to eye positions, 2) 14 state-of-the-art bottom-up saliency models, and 3) brute-force algo- rithms such as mean eye position. Our results show that the proposed model is more effective in employing and reasoning over spatio-temporal visual data.
}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @mastersthesis {thesis/MIT/Peters12, title = {Optimal Planning and Control for Hazard Avoidance of Front-Wheel Steered Ground Vehicles}, volume = {Doctor of Philosophy}, year = {2012}, pages = {137}, school = {Massachusetts Institute of Technology}, type = {PhD}, abstract = {Hazard avoidance is an important capability for safe operation of robotic vehicles at high speed. It is also an important consideration for passenger vehicle safety, as thousands are killed each year in passenger vehicle accidents caused by driver error. Even when hazard locations are known, high-speed hazard avoidance presents challenges in real-time motion planning and control of nonlinear and potentially unstable vehicle dynamics.
This thesis presents methods for planning and control of optimal hazard avoidance maneuvers for a bicycle model with front-wheel steering and wheel slip. The planning problem is posed as an optimization problem in which constrained dynamic quantities, such as friction circle utilization, are minimized, while ensuring a minimum clearance from hazards. These optimal trajectories can be computed numerically, though real-time computation requires simple models and constraints.
To simplify the computation of optimal avoidance trajectories, analytical solutions to the optimal planning problem are presented for a point mass subject to an acceleration magnitude constraint, which is analogous to a tire friction circle constraint. The optimal point mass solutions are extended to a nonlinear bicycle model by defining a flatness- based trajectory tracking controller using tire force control. This controller decouples the bicycle dynamics into a point mass at the front center of oscillation with an additional degree of freedom related to the vehicle yaw dynamics. Structure is identified in the yaw dynamics and is exploited to characterize stability limits. Simulation results verify the stability properties of the yaw dynamics.
These results were applied to a semi-autonomous driver assistance system and demonstrated experimentally on a full-sized passenger vehicle. Efficient computation of point mass avoidance maneuvers was used as a cost-to-go for real-time numerical optimization of trajectories for a bicycle model. The experimental system switches control authority between the driver and an automatic avoidance controller so that the driver retains control authority in benign situations, and the automatic controller avoids hazards automatically in hazardous situations.
}, author = {Steven C. Peters} } @conference {6523899, title = {Planning random path distributions for ambush games in unstructured environments}, booktitle = {Safety, Security, and Rescue Robotics (SSRR), 2012 IEEE International Symposium on}, year = {2012}, keywords = {ambush game, ambush games, environment characteristics, game theory, linear optimization, linear program, linear programming, LP solver algorithm, path planning, probability distributions, random path distribution planning, random processes, statistical distributions, unstructured environment, zero sum noncooperative game}, doi = {10.1109/SSRR.2012.6523899}, author = {Boidot, E. and Eric M. Feron} } @conference {DBLP:conf/cvpr/BorjiSI12, title = {Probabilistic learning of task-specific visual attention}, booktitle = {CVPR}, year = {2012}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @conference {DBLP:conf/cdc/HuynhF12, title = {Probabilistically-sound and Asymptotically-optimal Algorithm for Stochastic Control with Trajectory Constraints}, booktitle = {CDC-ECE}, year = {2012}, author = {Vu Anh Huynh and Emilio Frazzoli} } @conference {DBLP:conf/nfm/Herencia-ZapanaJOGFPA12, title = {PVS Linear Algebra Libraries for Verification of Control Software Algorithms in C/ACSL}, booktitle = {NASA Formal Methods}, year = {2012}, author = {Heber Herencia-Zapana and Romain Jobredeaux and Sam Owre and Pierre-Loic Garoche and Eric M. Feron and Gilberto Perez and Pablo Ascariz} } @article {Borji_etal12tip, title = {Quantitative Analysis of Human-Model Agreement in Visual Saliency Modeling: A Comparative Study}, journal = {IEEE Transactions on Image Processing}, year = {2012}, pages = {1-16 (in press)}, type = {mod;bu}, abstract = {Visual attention is a process that enables biological and machine vision systems to select the most relevant regions from a scene. Relevance is determined by two components: 1) top-down factors driven by task and 2) bottom-up factors that highlight image regions that are different from their surroundings. The latter are often referred to as {\textquoteright}visual saliency{\textquoteright}. Modeling bottom-up visual saliency has been the subject of numerous research efforts during the past 20 years, with many successful applications in computer vision and robotics. Available models have been tested with different datasets (e.g., synthetic psychological search arrays, natural images or videos) using different evaluation scores (e.g., search slopes, comparison to human eye tracking) and parameter settings. This has made direct comparison of models difficult. Here we perform an exhaustive comparison of 35 state-of-the-art saliency models over 54 challenging synthetic patterns, 3 natural image datasets, and 2 video datasets, using 3 evaluation scores. We find that although model rankings vary, some models consistently perform better. Analysis of datasets reveals that existing datasets are highly center-biased, which influences some of the evaluation scores. Computational complexity analysis shows that some models are very fast, yet yield competitive eye movement prediction accuracy. Different models often have common easy/difficult stimuli. Furthermore, several concerns in visual saliency modeling, eye movement datasets, and evaluation scores are discussed and insights for future work are provided. Our study allows one to assess the state-of-the-art, helps organizing this rapidly growing field, and sets a unified comparison framework for gauging future efforts, similar to the PASCAL VOC challenge in the object recognition and detection domains.
}, url = {http://ilab.usc.edu/publications/doc/Borji_etal12tip.pdf}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @conference {Roberts12spie, title = {Saliency Detection and Model-based Tracking: a Two Part Vision System for Small Robot Navigation in Forested Environment}, booktitle = {Unmanned Systems Technology XIV - SPIE Defense, Security and Sensing}, year = {2012}, url = {http://frank.dellaert.com/pubs/Roberts12spie.pdf}, author = {Richard Roberts and Duy-Nguyen Ta and Julian Straub and Frank Dellaert} } @conference {DBLP:conf/eccv/BorjiSI12, title = {Salient Object Detection: A Benchmark}, booktitle = {ECCV (2)}, year = {2012}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @conference {DBLP:conf/cdc/ChaudhariKF12, title = {Sampling-based Algorithm for Filtering using Markov chain Approximations}, booktitle = {CDC-ECE}, year = {2012}, author = {Chaudhari, P. and Sertac Karaman and Emilio Frazzoli} } @mastersthesis {100, title = {Sampling-based Algorithms for Optimal Path Planning Problems}, volume = {Doctor of Philosophy}, year = {2012}, pages = {152}, school = {Massachusetts Institute of Technology}, type = {PhD}, abstract = {Sampling-based motion planning received increasing attention during the last decade. In particular, some of the leading paradigms, such the Probabilistic RoadMap (PRM) and the Rapidly-exploring Random Tree (RRT) algorithms, have been demonstrated on several robotic platforms, and found applications well outside the robotics domain. However, a large portion of this research effort has been limited to the classical feasible path planning problem, which asks for finding a path that starts from an initial configuration and reaches a goal configuration while avoiding collision with obstacles.
The main contribution of this dissertation is a novel class of algorithms that extend the application domain of sampling-based methods to two new directions: optimal path planning and path planning with complex task specifications.
Regarding the optimal path planning problem, we first show that the existing algorithms either lack asymptotic optimality, i.e., almost-sure convergence to optimal solutions, or they lack computational efficiency: on one hand, neither the RRT nor the k-nearest PRM (for any fixed k) is asymptotically optimal; on the other hand, the simple PRM algorithm, where the connections are sought within fixed radius balls, is not computationally as efficient as the RRT or the efficient PRM variants. Subsequently, we propose two novel algorithms, called PRM* and RRT*, both of which guarantee asymptotic optimality without sacrificing computational efficiency. In fact, the proposed algorithms and the most efficient existing algorithms, such as the RRT, have the same asymptotic computational complexity.
Regarding the path planning problem with complex task specifications, we propose an incremental sampling-based algorithm that is provably correct and probabilistically complete, i.e., it generates a correct-by-design path that satisfies a given determinis- tic μ-calculus specification, when such a path exists, with probability approaching to one as the number of samples approaches infinity. For this purpose, we develop two key ingredients. First, we propose an incremental sampling-based algorithm, called the RRG, that generates a representative set of paths in the form of a graph, with guaranteed almost-sure convergence towards feasible paths. Second, we propose an in- cremental local model-checking algorithm for the deterministic μ-calculus. Moreover, with the help of these tools and the ideas behind the RRT*, we construct algorithms that also guarantee almost sure convergence to optimal solutions.
}, author = {Sertac Karaman} } @conference {DBLP:conf/cvpr/FathiHR12, title = {Social interactions: A first-person perspective}, booktitle = {CVPR}, year = {2012}, author = {Alireza Fathi and Jessica K. Hodgins and Jim M. Rehg} } @article {Borji_Itti12pami, title = {State-of-the-art in Visual Attention Modeling}, journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence}, year = {2012}, type = {mod;bu;rev}, abstract = {Modeling visual attention - particularly stimulus-driven, saliency-based attention - has been a very active research area over the past 25 years. Many different models of attention are now available, which aside from lending theoretical contributions to other fields, have demonstrated successful applications in computer vision, mobile robotics, and cognitive systems. Here we review, from a computational perspective, the basic concepts of attention implemented in these models. We present a taxonomy of nearly 65 models, which provides a critical comparison of approaches, their capabilities, and shortcomings. In particular, thirteen criteria derived from behavioral and computational studies are formulated for qualitative comparison of attention models. Furthermore, we address several challenging issues with models, including biological plausibility of the computations, correlation with eye movement datasets, bottom-up and top-down dissociation, and constructing meaningful performance measures. Finally, we highlight current research trends in attention modeling and provide insights for future.
}, url = {http://ilab.usc.edu/publications/doc/Borji_Itti12pami.pdf}, author = {Ali Borji and Laurent Itti} } @article {ieee.tr.control.systems.technology-PetersAPTI12, title = {Threat-based hazard avoidance for semi- autonomous vehicles using nonlinear model predictive control}, journal = {IEEE Transactions on Control Systems Technology}, year = {2012}, author = {Steven C. Peters and Anderson, S. and Tom Pilutti and H. Eric Tseng and Karl Iagnemma} } @article {Shen_Itti12vr, title = {Top-down Influences on Visual Attention during Listening are Modulated by Observer Sex}, journal = {Vision Research}, volume = {65}, year = {2012}, month = {Jul}, pages = {62-76}, type = {psy;eye;bu}, abstract = {In conversation, women have a small advantage in decoding non-verbal communication compared to men. In light of these findings, we sought to determine whether sex differences also existed in visual attention during a related listening task, and if so, if the differences existed among attention to high-level aspects of the scene or to conspicuous visual features. Using eye-tracking and computational techniques, we present direct evidence that men and women orient attention differently during conversational listening. We tracked the eyes of 15 men and 19 women who watched and listened to 84 clips featuring 12 different speakers in various outdoor settings. At the fixation following each saccadic eye movement, we analyzed the type of object that was fixated. Men gazed more often at the mouth and women at the eyes of the speaker. Women more often exhibited {\textquoteright}distracted{\textquoteright} saccades directed away from the speaker and towards a background scene element. Examining the multi-scale center-surround variation in low-level visual features (static: color, intensity, orientation, and dynamic: motion energy), we found that men consistently selected regions which expressed more variation in dynamic features, which can be attributed to a male preference for motion and a female preference for areas that may contain nonverbal information about the speaker. In sum, significant differences were observed, which we speculate arise from different integration strategies of visual cues in selecting the final target of attention. Our findings have implications for studies of sex in nonverbal communication, as well as for more predictive models of visual attention.
}, author = {John Shen and Laurent Itti} } @conference {Ok12vicomor, title = {Vistas and Wall-Floor Intersection Features - Enabling Autonomous Flight in Man-made Environments}, booktitle = {Workshop on Visual Control of Mobile Robots}, year = {2012}, url = {http://frank.dellaert.com/pubs/Ok12vicomor.pdf}, author = {Kyel Ok and Duy-Nguyen Ta and Frank Dellaert} } @conference {DBLP:conf/cdc/JeonKF11, title = {Anytime computation of time-optimal off-road vehicle maneuvers using the RRT}, booktitle = {CDC-ECE}, year = {2011}, author = {Jeong Hwan Jeon and Sertac Karaman and Emilio Frazzoli} } @conference {Perez.Karaman.ea:IROS11, title = {Asymptotically-optimal Path Planning for Manipulation using Incremental Sampling-based Algorithms}, booktitle = {Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, year = {2011}, author = {Gilberto Perez and Sertac Karaman and Walter and Shkolnik and Emilio Frazzoli and Teller} } @conference {6096122, title = {Autocoding control software with proofs I: Annotation translation}, booktitle = {Digital Avionics Systems Conference (DASC), 2011 IEEE/AIAA 30th}, year = {2011}, month = {oct.}, abstract = {In an effort to meet the reliability standards that control systems operating in safety-critical roles require, we have started laying the foundation for a tool-set that migrates control theory properties and proofs into the software implementation of those control systems designs. By using this tool the engineer can provide a more rigorous guarantee of the quality of the software and initiate the formal verification process. The tool focuses on control software in order to leverage the domain knowledge from existing mathematical techniques for the analysis and synthesis of control systems. As a first step in the development of the tool-set, we have created a prototype of a Scilab to C translator with proof annotation support. Though limited in its current functionalities, the development of this prototype allowed us to identify the key issues which will be used to further refine the translator. This paper describes the prototype and the further improvements planned for the translator.
}, keywords = {annotation translation, autocoding control software, C language, C translator, control engineering computing, control software, control system analysis, control system design, control system synthesis, formal verification process, mathematical techniques, program compilers, program interpreters, program verification, proof annotation support, reliability standards, safety-critical software, Scilab prototype, software implementation, software prototyping, software reliability, software tools, theorem proving, tool-set}, doi = {10.1109/DASC.2011.6096122}, url = {papers/06096122.pdf}, author = {Romain Jobredeaux and Timothy Wang and Eric M. Feron} } @conference {http://dx.doi.org/10.5244/C.25.78, title = {Combining Self Training and Active Learning for Video Segmentation}, booktitle = {Proceedings of the British Machine Vision Conference}, year = {2011}, publisher = {BMVA Press}, organization = {BMVA Press}, isbn = {1-901725-43-X}, author = {Alireza Fathi and Maria Florina Balcan and Xiaofeng Ren and Jim M. Rehg} } @conference {Borji_etal11bmvc, title = {Computational Modeling of Top-down Visual Attention in Interactive Environments}, booktitle = {Proc. British Machine Vision Conference (BMVC 2011)}, year = {2011}, month = {Sep}, abstract = {Modeling how visual saliency guides the deployment of attention over visual scenes has attracted much interest recently - among both computer vision and experimental/computational researchers - since visual attention is a key function of both machine and biological vision systems. Research efforts in computer vision have mostly been focused on modeling bottom-up saliency. Strong influences on attention and eye movements, however, come from instantaneous task demands. Here, we propose models of top-down visual guidance considering task influences. The new models estimate the state of a human subject performing a task (here, playing video games), and map that state to an eye position. Factors influencing state come from scene gist, physical actions, events, and bottom-up saliency. Proposed models fall into two categories. In the first category, we use classical discriminative classifiers, including Regres- sion, kNN and SVM. In the second category, we use Bayesian Networks to combine all the multi-modal factors in a unified framework. Our approaches significantly outperform 15 competing bottom-up and top-down attention models in predicting future eye fixations.
}, author = {Ali Borji and Dicky N. Sihite and Laurent Itti} } @mastersthesis {gatech/cowlagi.PhDThesis, title = {Hierarchical Motion Planning for Autonomous Aerial and Terrestrial Vehicles}, volume = {Doctor of Philosophy}, year = {2011}, school = {Georgia Institute of Technology}, type = {PhD}, author = {Raghvendra V. Cowlagi} } @conference {DBLP:conf/cvpr/FathiRR11, title = {Learning to recognize objects in egocentric activities}, booktitle = {CVPR}, year = {2011}, author = {Alireza Fathi and Xiaofeng Ren and Jim M. Rehg} } @conference {Bialkowski.Karaman.ea:IROS11, title = {Massively Parallelizing the RRT and the RRT*}, booktitle = {IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS)}, year = {2011}, author = {Joshua Bialkowski and Sertac Karaman and Emilio Frazzoli} } @conference {DBLP:conf/iros/ChangSI10, title = {Mobile robot vision navigation \& localization using Gist and Saliency}, booktitle = {IROS}, year = {2011}, author = {Chin-Kai Chang and Christian Siagian and Laurent Itti} } @conference {conf/iros/LuHAT11, title = {Multi-scale LPA* with low worst-case complexity guarantees.}, booktitle = {IROS}, year = {2011}, publisher = {IEEE}, organization = {IEEE}, keywords = {dblp}, isbn = {978-1-61284-454-1}, url = {http://dblp.uni-trier.de/db/conf/iros/iros2011.html$\#$LuHAT11}, author = {Yibiao Lu and Xiaoming Huo and Oktay Arslan and Panagiotis Tsiotras} } @conference {DBLP:conf/iros/ArslanTH11, title = {Solving Shortest Path Problems with Curvature Constraints using Beamlets}, booktitle = {IROS}, year = {2011}, author = {Oktay Arslan and Panagiotis Tsiotras and Xiaoming Huo} } @article {ieee.control.magazine-ArndtBPID11, title = {Two-Wheel Self-Balancing of a Four-Wheeled Vehicle}, journal = {IEEE Control Systems Magazine}, year = {2011}, month = {April 2011}, author = {David Arndt and James E. Bobrow and Steven C. Peters and Karl Iagnemma and Steven Dubowsky} } @conference {DBLP:conf/iccv/FathiFR11, title = {Understanding egocentric activities}, booktitle = {ICCV}, year = {2011}, author = {Alireza Fathi and Ali Farhadi and Jim M. Rehg} } @conference {DBLP:dblp_conf/cdc/ChakrabortyTL11, title = {Vehicle posture control through aggressive maneuvering for mitigation of T-bone collisions.}, booktitle = {CDC-ECE}, year = {2011}, author = {Chakraborty, I. and Panagiotis Tsiotras and Jianbo Lu and Jianbo Lu} } @conference {DBLP:conf/iros/ChangSI10, title = {Mobile robot vision navigation {\&} localization using Gist and Saliency}, booktitle = {IROS}, year = {2010}, author = {C.K. Chang and Christian Siagian and Laurent Itti} } @conference {cowlagi2010multi, title = {Multi-resolution Path Planning: Theoretical Analysis, Efficient Implementation, and Extensions to Dynamic Environments}, booktitle = {Decision and Control (CDC), 2010 49th IEEE Conference on}, year = {2010}, publisher = {IEEE}, organization = {IEEE}, author = {Raghvendra V. Cowlagi and Panagiotis Tsiotras} } @article {citeulike:9543669, title = {Optimal kinodynamic motion planning using incremental sampling based methods}, year = {2010}, keywords = {rrt}, url = {http://sertac.scripts.mit.edu/web/wp-content/papercite-data/pdf/karaman.frazzoli-cdc10.pdf}, author = {Sertac Karaman and Emilio Frazzoli} } @article {Velenis.Frazzoli.ea:IJVAS09, title = {Steady-State Cornering Equilibria and Stabilization for a Vehicle During Extreme Operating Conditions.}, journal = {Int. Journal of Vehicle Autonomous Systems}, year = {2010}, author = {Efstathios Velenis and Emilio Frazzoli and Panagiotis Tsiotras} }