Showing papers by "Sebastian Thrun published in 2005"

Probabilistic Robotics

Abstract: Planning and navigation algorithms exploit statistics gleaned from uncertain, imperfect real-world environments to guide robots toward their goals and around obstacles.

SCAPE: shape completion and animation of people

Abstract: We introduce the SCAPE method (Shape Completion and Animation for PEople)---a data-driven method for building a human shape model that spans variation in both subject shape and pose. The method is based on a representation that incorporates both articulated and non-rigid deformations. We learn a pose deformation model that derives the non-rigid surface deformation as a function of the pose of the articulated skeleton. We also learn a separate model of variation based on body shape. Our two models can be combined to produce 3D surface models with realistic muscle deformation for different people in different poses, when neither appear in the training set. We show how the model can be used for shape completion --- generating a complete surface mesh given a limited set of markers specifying the target shape. We present applications of shape completion to partial view completion and motion capture animation. In particular, our method is capable of constructing a high-quality animated surface model of a moving person, with realistic muscle deformation, using just a single static scan and a marker motion capture sequence of the person.

An Application of Markov Random Fields to Range Sensing

Abstract: This paper describes a highly successful application of MRFs to the problem of generating high-resolution range images. A new generation of range sensors combines the capture of low-resolution range images with the acquisition of registered high-resolution camera images. The MRF in this paper exploits the fact that discontinuities in range and coloring tend to co-align. This enables it to generate high-resolution, low-noise range images by integrating regular camera images into the range data. We show that by using such an MRF, we can substantially improve over existing range imaging technology.

Anytime dynamic A*: an anytime, replanning algorithm

Abstract: We present a graph-based planning and replanning algorithm able to produce bounded suboptimal solutions in an anytime fashion. Our algorithm tunes the quality of its solution based on available search time, at every step reusing previous search efforts. When updated information regarding the underlying graph is received, the algorithm incrementally repairs its previous solution. The result is an approach that combines the benefits of anytime and incremental planners to provide efficient solutions to complex, dynamic search problems. We present theoretical analysis of the algorithm, experimental results on a simulated robot kinematic arm, and two current applications in dynamic path planning for outdoor mobile robots.

Learning Motion Patterns of People for Compliant Robot Motion

Abstract: Whenever people move through their environments they do not move randomly. Instead, they usually follow specific trajectories or motion patterns corresponding to their intentions. Knowledge about such patterns enables a mobile robot to robustly keep track of persons in its environment and to improve its behavior. In this paper we propose a technique for learning collections of trajectories that characterize typical motion patterns of persons. Data recorded with laser-range finders are clustered using the expectation maximization algorithm. Based on the result of the clustering process, we derive a hidden Markov model that is applied to estimate the current and future positions of persons based on sensory input. We also describe how to incorporate the probabilistic belief about the potential trajectories of persons into the path planning process of a mobile robot. We present several experiments carried out in different environments with a mobile robot equipped with a laser-range scanner and a camera system. The results demonstrate that our approach can reliably learn motion patterns of persons, can robustly estimate and predict positions of persons, and can be used to improve the navigation behavior of a mobile robot.

Shape from symmetry

Abstract: We describe a technique for reconstructing probable occluded surfaces from 3D range images. The technique exploits the fact that many objects possess shape symmetries that can be recognized even from partial 3D views. Our approach identifies probable symmetries and uses them to attend the partial 3D shape model into the occluded space. To accommodate objects consisting of multiple parts, we describe a technique for segmenting objects into parts characterized by different symmetries. Results are provided for a real-world database of 3D range images of common objects, acquired through an active stereo rig

Multi-robot SLAM with Sparse Extended Information Filers

Abstract: We present an algorithm for the multi-robot simultaneous localization and mapping (SLAM) problem. Our algorithm enables teams of robots to build joint maps, even if their relative starting locations are unknown and landmarks are ambiguous—which is presently an open problem in robotics. It achieves this capability through a sparse information filter technique, which represents maps and robot poses by Gaussian Markov random fields. The alignment of local maps into a single global maps is achieved by a tree-based algorithm for searching similar-looking local landmark configurations, paired with a hill climbing algorithm that maximizes the overall likelihood by search in the space of correspondences. We report favorable results obtained with a real-world benchmark data set.

Finding approximate POMDP solutions through belief compression

Abstract: Standard value function approaches to finding policies for Partially Observable Markov Decision Processes (POMDPs) are generally considered to be intractable for large models. The intractability of these algorithms is to a large extent a consequence of computing an exact, optimal policy over the entire belief space. However, in real-world POMDP problems, computing the optimal policy for the full belief space is often unnecessary for good control even for problems with complicated policy classes. The beliefs experienced by the controller often lie near a structured, low-dimensional subspace embedded in the high-dimensional belief space. Finding a good approximation to the optimal value function for only this subspace can be much easier than computing the full value function. We introduce a new method for solving large-scale POMDPs by reducing the dimensionality of the belief space. We use Exponential family Principal Components Analysis (Collins, Dasgupta, & Schapire, 2002) to represent sparse, high-dimensional belief spaces using small sets of learned features of the belief state. We then plan only in terms of the low-dimensional belief features. By planning in this low-dimensional space, we can find policies for POMDP models that are orders of magnitude larger than models that can be handled by conventional techniques. We demonstrate the use of this algorithm on a synthetic problem and on mobile robot navigation tasks.

Discriminative Training of Kalman Filters.

Abstract: Kalman filters are a workhorse of robotics and are routinely used in state-estimation problems. However, their performance critically depends on a large number of modeling parameters which can be very difficult to obtain, and are often set via significant manual tweaking and at a great cost of engineering time. In this paper, we propose a method for automatically learning the noise parameters of a Kalman filter. We also demonstrate on a commercial wheeled rover that our Kalman filter’s learned noise covariance parameters—obtained quickly and fully automatically—significantly outperform an earlier, carefully and laboriously hand-designed one.

Towards Lazy Data Association in SLAM

Abstract: We present a lazy data association algorithm for the simultaneous localization and mapping (SLAM) problem. Our approach uses a tree-structured Bayesian representation of map posteriors that makes it possible to revise data association decisions arbitrarily far into the past. We describe a criterion for detecting and repairing poor data association decisions. This technique makes it possible to acquire maps of large-scale environments with many loops, with a minimum of computational overhead for the management of multiple data association hypotheses. A empirical comparison with the popular FastSLAM algorithm shows the advantage of lazy over proactive data association.

Affine Structure From Sound

Abstract: We consider the problem of localizing a set of microphones together with a set of external acoustic events (e.g., hand claps), emitted at unknown times and unknown locations. We propose a solution that approximates this problem under a far field approximation defined in the calculus of affine geometry, and that relies on singular value decomposition (SVD) to recover the affine structure of the problem. We then define low-dimensional optimization techniques for embedding the solution into Euclidean geometry, and further techniques for recovering the locations and emission times of the acoustic events. The approach is useful for the calibration of ad-hoc microphone arrays and sensor networks.

Adaptive Road Following using Self-Supervised Learning and Reverse Optical Flow.

Abstract: The majority of current image-based road following algorithms operate, at least in part, by assuming the presence of structural or visual cues unique to the roadway. As a result, these algorithms are poorly suited to the task of tracking unstructured roads typical in desert environments. In this paper, we propose a road following algorithm that operates in a selfsupervised learning regime, allowing it to adapt to changing road conditions while making no assumptions about the general structure or appearance of the road surface. An application of optical flow techniques, paired with one-dimensional template matching, allows identification of regions in the current camera image that closely resemble the learned appearance of the road in the recent past. The algorithm assumes the vehicle lies on the road in order to form templates of the road’s appearance. A dynamic programming variant is then applied to optimize the 1-D template match results while enforcing a constraint on the maximum road curvature expected. Algorithm output images, as well as quantitative results, are presented for three distinct road types encountered in actual driving video acquired in the California Mojave Desert.

Game Theoretic Control for Robot Teams

Abstract: In the real world, noisy sensors and limited communication make it difficult for robot teams to coordinate in tightly coupled tasks. Team members cannot simply apply single-robot solution techniques for partially observable problems in parallel because they do not take into account the recursive effect that reasoning about the beliefs of others has on policy generation. Instead, we must turn to a game theoretic approach to model the problem correctly. Partially observable stochastic games (POSGs) provide a solution model for decentralized robot teams, however, this model quickly becomes intractable. In previous work we presented an algorithm for lookahead search in POSGs. Here we present an extension which reduces computation during lookahead by clustering similar observation histories together. We show that by clustering histories which have similar profiles of predicted reward, we can greatly reduce the computation time required to solve a POSG while maintaining a good approximation to the optimal policy. We demonstrate the power of the clustering algorithm in a real-time robot controller as well as for a simple benchmark problem.

A probabilistic language based upon sampling functions

Abstract: As probabilistic computations play an increasing role in solving various problems, researchers have designed probabilistic languages that treat probability distributions as primitive datatypes. Most probabilistic languages, however, focus only on discrete distributions and have limited expressive power. In this paper, we present a probabilistic language, called λο, which uniformly supports all kinds of probability distributions -- discrete distributions, continuous distributions, and even those belonging to neither group. Its mathematical basis is sampling functions, i.e., mappings from the unit interval (0.0,1.0] to probability domains.We also briefly describe the implementation of λο as an extension of Objective CAML and demonstrate its practicality with three applications in robotics: robot localization, people tracking, and robotic mapping. All experiments have been carried out with real robots.

Parallel Stochastic Hill- Climbing with Small Teams

Abstract: We address the basic problem of coordinating the actions of multiple robots that are working toward a common goal. This kind of problem is NP-hard, because in order to coordinate a system of n robots, it is in principle necessary to generate and evaluate a number of actions or plans that is exponential in n (assuming P ≠ NP). However, we suggest that many instances of coordination problems, despite the NP-hardness of the overall class of problems, do not in practice require exponential computation in order to arrive at good solutions. In such problems, it is not necessary to consider all possible actions of the n robots; instead an algorithm may restrict its attention to interactions within small teams, and still produce high-quality solutions.

The Information-Form Data Association Filter

Abstract: This paper presents a new filter for online data association problems in high-dimensional spaces. The key innovation is a representation of the data association posterior in information form, in which the "proximity" of objects and tracks are expressed by numerical links. Updating these links requires linear time, compared to exponential time required for computing the exact posterior probabilities. The paper derives the algorithm formally and provides comparative results using data obtained by a real-world camera array and by a large-scale sensor network simulation.

Robotic mapping with polygonal random fields

Abstract: Two types of probabilistic maps are popular in the mobile robotics literature: occupancy grids and geometric maps. Occupancy grids have the advantages of simplicity and speed, but they represent only a restricted class of maps and they make incorrect independence assumptions. On the other hand, current geometric approaches, which characterize the environment by features such as line segments, can represent complex environments compactly. However, they do not reason explicitly about occupancy, a necessity for motion planning; and, they lack a complete probability model over environmental structures. In this paper we present a probabilistic mapping technique based on polygonal random fields (PRF), which combines the advantages of both approaches. Our approach explicitly represents occupancy using a geometric representation, and it is based upon a consistent probability distribution over environments which avoids the incorrect independence assumptions made by occupancy grids. We show how sampling techniques for PRFs can be applied to localized laser and sonar data, and we demonstrate significant improvements in mapping performance over occupancy grids.

Search-based planning for large dynamic environments

Abstract: Agents operating in the real world often have to act under the conditions where time is critical: there is a limit on the time they can afford to spend on deliberating what action to execute next. Planners used by such agents must produce the best plans they can find within the amount of time available. The strategy of always searching for an optimal plan becomes infeasible in these scenarios. Instead, we must use an anytime planner. Anytime planners operate by quickly finding a highly suboptimal plan first, and then improving it until the available time runs out. In addition to the constraints on time, world models used by planners are usually imperfect and environments are often dynamic. The execution of a plan therefore often results in unexpected outcomes. An agent then needs to update the model accordingly and re-execute a planner on the new model. A planner that has a replanning capability (a.k.a. an incremental planner) can substantially speed up each planning episode in such cases, as it tries to make use of the results of previous planning efforts in finding a new plan. Combining anytime with replanning capabilities is thus beneficial. For one, at each planning episode it allows the planner to produce a better plan within the available time: both in finding the first plan as well as in improving it, the planner can use its replanning capability to accelerate the process. In addition, the combination allows one to interleave planning and execution effectively. While the agent executes the current plan, the planner can continue improving it without having to discard all of its efforts every time the model of the world is updated. This thesis concentrates on graph-based searches. It presents an alternative view of A* search, a widely popular heuristic search in AI, and then uses this view to easily derive three versions of A* search: an anytime version, an incremental version and a version that is both anytime and incremental. Each of these algorithms is also able to provide a non-trivial bound on the suboptimality of the solution it generates. We believe that the simplicity, the existence of suboptimality bounds and the generality of the presented methods contribute to the research and development of planners well suited for systems operating in the real world.

Lazy inference on object identities in wireless sensor networks

Abstract: Maintaining the identities of moving objects is an important aspect of most multi-object tracking applications Uncertainty in sensor data, coupled with the intrinsic combinatorial difficulty of the data association problem, suggests probabilistic formulations over the set of possible identities While an explicit representation of a distribution over all associations may require exponential storage and computation, in practice the information provided by this distribution is accessed only in certain stylized ways, as when asking for the identity of a given track, or the track with a given identity Exploiting this observation, we proposed in [1] a practical solution to this problem based on maintaining marginal probabilities and demonstrated its effectiveness in the context of tracking within a wireless sensor network That method, unfortunately, requires extensive communication in the network whenever new identity observations are made, in order for normalization operations to keep the marginals consistent [2] In this paper, we have proposed a very different solution based on accumulated log-likelihoods that can postpone all normalization computations until actual identity queries are made In this manner the continuous communication and computational expense of repeated normalizations is avoided and that effort is expended only when actual queries are made of the network We compare the two methods in terms of their computational complexities, inference accuracies, and distributed implementations Simulation and experimental results from a RFID system are also presented

Autonomous Helicopter Tracking and Localization Using a Self-Surveying Camera Array.

Abstract: A Self-surveying Camera Array (SSCA) is a vision-based local-area positioning system consisting of multiple ground-deployed cameras that are capable of self-surveying their extrinsic parameters while tracking and localizing a moving target. This paper presents the self-surveying algorithm being used to track a target helicopter in each camera frame and to localize the helicopter in an array-fixed frame. Three cameras are deployed independently in an arbitrary arrangement that allows each camera to view the helicopter's flight volume. The helicopter then flies an unplanned path that allows the cameras to calibrate the relative locations and orientations by utilizing a self-surveying algorithm that is extended from the well-known structure from motion algorithm and the bundle adjustment technique. This yields the cameras'extrinsic parameters enabling real-time helicopter positioning via triangulation. This paper also presents results from field trials, which verify the feasibility of the SSCA as a readily-deployable system applicable to helicopter tracking and localization. The results demonstrate that, compared to the differential GPS solution as true reference, the SSCA alone is capable of positioning the helicopter with meter-level accuracy. The SSCA has been integrated with onboard inertial sensors providing a reliable positioning system to enable successful autonomous hovering.

Learning Activity-Based Ground Models from a Moving Helicopter Platform

Abstract: We present a method for learning activity-based ground models based on a multiple particle filter approach to motion tracking in video acquired from a moving aerial platform. Such models offer a number of potential benefits. In this paper we demonstrate the ability of activity-based models to improve the performance of an object motion tracker as well as their applicability to global registration of video sequences.

A distributed approach to passive localization for sensor networks

Abstract: Sensors that know their location, from microphones to vibration sensors, can support a wider arena of applications than their location unaware counterparts. We offer a method for sensors to determine their own location relative to one another by using only exogenous sounds and the differences in the arrivals of these sounds at different sensors. We present a distributed and computationally efficient solution that offers accuracy on par with more active and computationally intense methods.

Active Sensing for High-Speed Offroad Driving

Abstract: In this paper we propose an active control strategy for scanning laser sensors on autonomous vehicles traveling offroad at high speeds. As speed increases the amount of sensor information about the terrain decreases. We address the problem of sensor control in the context of this speed-coverage trade off. The algorithm and testing methodologies are described with results comparing our active sensing method to a passive sensing method.

Linear Systems and Control

Abstract: This chapter contains sections titled: State Space Representation, Stability, LTI Control Systems, Observing LTI Systems, Discrete Time Systems

Sampling-Based Algorithms

Abstract: This chapter contains sections titled: Probabilistic Roadmaps, Single-Query Sampling-Based Planners, Integration of Planners: Sampling-Based Roadmap of Trees, Analysis of PRM, Beyond Basic Path Planning, Problems

A programming language for probabilistic computation

Abstract: As probabilistic computations play an increasing role in solving various problems, researchers have designed probabilistic languages to facilitate their modeling Most of the existing probabilistic languages, however, focus only on discrete distributions, and there has been little effort to develop probabilistic languages whose expressive power is beyond discrete distributions This dissertation presents a probabilistic language, called PTP (ProbabilisTic Programming), which supports all kinds of probability distributions The key idea behind PTP is to use sampling functions, ie, mappings from the unit interval (00,10] to probability domains, to specify probability distributions By using sampling functions as its mathematical basis, PTP provides a unified representation scheme for probability distributions, without drawing a syntactic or semantic distinction between different kinds of probability distributions Independently of PTP, we develop a linguistic framework, called λ c, to account for computational effects in general λ c extends a monadic language by applying the possible world interpretation of modal logic A characteristic feature of λc is the distinction between stateful computational effects, called world effects , and contextual computational effects, called control effects PTP arises as an instance of λc with a language construct for probabilistic choices We use a sound and complete translator of PTP to embed it in Objective CAML The use of PTP is demonstrated with three applications in robotics: robot localization, people tracking, and robotic mapping Thus PTP serves as another example of high-level language applied to a problem domain where imperative languages have been traditionally dominant

Nonholonomic and Underactuated Systems

Abstract: This chapter contains sections titled: Preliminaries, Control Systems, Controllability, Simple Mechanical Control Systems, Motion Planning, Problems

175 miles through the desert

Abstract: The DARPA Grand Challenge is one of the biggest open challenges for the robotics community to date. It requires a robotic vehicle to follow a given route of up to 175 miles across punishing desert terrain, without any human supervision. The Challenge was first held in 2004, in which the best performing team failed after 7.3 miles of autonomous driving. The speaker heads one out of 195 teams worldwide competing for the 2 Million Dollar price. Thrun will present the work of the Stanford Racing Team, which is developing an automated car capable of desert driving at up to 50km/h. He will report on research in areas as diverse as computer vision, control, fault-tolerant systems, machine learning, motion planning, data fusion, and 3-D environment modeling.