Posted Content•
STEP: Stochastic Traversability Evaluation and Planning for Risk-Aware Off-road Navigation
TL;DR: In this article, the authors proposed an approach for assessing traversability and planning a safe, feasible, and fast trajectory in real-time, which relies on rapid uncertainty-aware mapping and traversability evaluation, tail risk assessment using the Conditional Value-at-Risk (CVaR), and efficient risk and constraint-aware kinodynamic motion planning using sequential quadratic programming-based predictive control.
Abstract: Although ground robotic autonomy has gained widespread usage in structured and controlled environments, autonomy in unknown and off-road terrain remains a difficult problem. Extreme, off-road, and unstructured environments such as undeveloped wilderness, caves, and rubble pose unique and challenging problems for autonomous navigation. To tackle these problems we propose an approach for assessing traversability and planning a safe, feasible, and fast trajectory in real-time. Our approach, which we name STEP (Stochastic Traversability Evaluation and Planning), relies on: 1) rapid uncertainty-aware mapping and traversability evaluation, 2) tail risk assessment using the Conditional Value-at-Risk (CVaR), and 3) efficient risk and constraint-aware kinodynamic motion planning using sequential quadratic programming-based (SQP) model predictive control (MPC). We analyze our method in simulation and validate its efficacy on wheeled and legged robotic platforms exploring extreme terrains including an abandoned subway and an underground lava tube.
Citations
More filters
Posted Content•
TL;DR: In this article, a scalable value learning framework, PLGRIM (Probabilistic Local and Global Reasoning on Information roadmaps), is proposed to bridge the gap between (i) local, risk-aware resiliency and (ii) global, reward-seeking mission objectives.
Abstract: In order for a robot to explore an unknown environment autonomously, it must account for uncertainty in sensor measurements, hazard assessment, localization, and motion execution Making decisions for maximal reward in a stochastic setting requires learning values and constructing policies over a belief space, ie, probability distribution of the robot-world state Value learning over belief spaces suffer from computational challenges in high-dimensional spaces, such as large spatial environments and long temporal horizons for exploration At the same time, it should be adaptive and resilient to disturbances at run time in order to ensure the robot's safety, as required in many real-world applications This work proposes a scalable value learning framework, PLGRIM (Probabilistic Local and Global Reasoning on Information roadMaps), that bridges the gap between (i) local, risk-aware resiliency and (ii) global, reward-seeking mission objectives By leveraging hierarchical belief space planners with information-rich graph structures, PLGRIM can address large-scale exploration problems while providing locally near-optimal coverage plans PLGRIM is a step toward enabling belief space planners on physical robots operating in unknown and complex environments We validate our proposed framework with a high-fidelity dynamic simulation in diverse environments and with physical hardware, Boston Dynamics' Spot robot, in a lava tube
18 citations
Posted Content•
TL;DR: NeBula as mentioned in this paper is an uncertainty-aware framework that aims at enabling resilient and modular autonomy solutions by performing reasoning and decision making in the belief space (space of probability distributions over the robot and world states).
Abstract: This paper presents and discusses algorithms, hardware, and software architecture developed by the TEAM CoSTAR (Collaborative SubTerranean Autonomous Robots), competing in the DARPA Subterranean Challenge. Specifically, it presents the techniques utilized within the Tunnel (2019) and Urban (2020) competitions, where CoSTAR achieved 2nd and 1st place, respectively. We also discuss CoSTAR's demonstrations in Martian-analog surface and subsurface (lava tubes) exploration. The paper introduces our autonomy solution, referred to as NeBula (Networked Belief-aware Perceptual Autonomy). NeBula is an uncertainty-aware framework that aims at enabling resilient and modular autonomy solutions by performing reasoning and decision making in the belief space (space of probability distributions over the robot and world states). We discuss various components of the NeBula framework, including: (i) geometric and semantic environment mapping; (ii) a multi-modal positioning system; (iii) traversability analysis and local planning; (iv) global motion planning and exploration behavior; (i) risk-aware mission planning; (vi) networking and decentralized reasoning; and (vii) learning-enabled adaptation. We discuss the performance of NeBula on several robot types (e.g. wheeled, legged, flying), in various environments. We discuss the specific results and lessons learned from fielding this solution in the challenging courses of the DARPA Subterranean Challenge competition.
10 citations
Posted Content•
26 Mar 2021
TL;DR: In this article, the authors consider the stochastic shortest path planning problem in MDPs, i.e., the problem of designing policies that ensure reaching a goal state from a given initial state with minimum accrued cost.
Abstract: We consider the stochastic shortest path planning problem in MDPs, i.e., the problem of designing policies that ensure reaching a goal state from a given initial state with minimum accrued cost. In order to account for rare but important realizations of the system, we consider a nested dynamic coherent risk total cost functional rather than the conventional risk-neutral total expected cost. Under some assumptions, we show that optimal, stationary, Markovian policies exist and can be found via a special Bellman's equation. We propose a computational technique based on difference convex programs (DCPs) to find the associated value functions and therefore the risk-averse policies. A rover navigation MDP is used to illustrate the proposed methodology with conditional-value-at-risk (CVaR) and entropic-value-at-risk (EVaR) coherent risk measures.
3 citations
Posted Content•
TL;DR: In this article, the authors proposed a non-sampling based method to assess the risk for trajectories of autonomous vehicles when probabilistic predictions of other agents' futures are generated by deep neural networks (DNNs).
Abstract: This paper presents fast non-sampling based methods to assess the risk for trajectories of autonomous vehicles when probabilistic predictions of other agents' futures are generated by deep neural networks (DNNs). The presented methods address a wide range of representations for uncertain predictions including both Gaussian and non-Gaussian mixture models to predict both agent positions and control inputs conditioned on the scene contexts. We show that the problem of risk assessment when Gaussian mixture models (GMMs) of agent positions are learned can be solved rapidly to arbitrary levels of accuracy with existing numerical methods. To address the problem of risk assessment for non-Gaussian mixture models of agent position, we propose finding upper bounds on risk using nonlinear Chebyshev's Inequality and sums-of-squares (SOS) programming; they are both of interest as the former is much faster while the latter can be arbitrarily tight. These approaches only require higher order statistical moments of agent positions to determine upper bounds on risk. To perform risk assessment when models are learned for agent control inputs as opposed to positions, we propagate the moments of uncertain control inputs through the nonlinear motion dynamics to obtain the exact moments of uncertain position over the planning horizon. To this end, we construct deterministic linear dynamical systems that govern the exact time evolution of the moments of uncertain position in the presence of uncertain control inputs. The presented methods are demonstrated on realistic predictions from DNNs trained on the Argoverse and CARLA datasets and are shown to be effective for rapidly assessing the probability of low probability events.
Posted Content•
TL;DR: In this paper, the learning and non-learning-based components are designed in a way such that they can be effectively combined in a self-supervised manner in an open-world off-road navigation task.
Abstract: Geometric methods for solving open-world off-road navigation tasks, by learning occupancy and metric maps, provide good generalization but can be brittle in outdoor environments that violate their assumptions (e.g., tall grass). Learning-based methods can directly learn collision-free behavior from raw observations, but are difficult to integrate with standard geometry-based pipelines. This creates an unfortunate conflict -- either use learning and lose out on well-understood geometric navigational components, or do not use it, in favor of extensively hand-tuned geometry-based cost maps. In this work, we reject this dichotomy by designing the learning and non-learning-based components in a way such that they can be effectively combined in a self-supervised manner. Both components contribute to a planning criterion: the learned component contributes predicted traversability as rewards, while the geometric component contributes obstacle cost information. We instantiate and comparatively evaluate our system in both in-distribution and out-of-distribution environments, showing that this approach inherits complementary gains from the learned and geometric components and significantly outperforms either of them. Videos of our results are hosted at https://sites.google.com/view/hybrid-imitative-planning
References
More filters
TL;DR: In this paper, the authors present and justify a set of four desirable properties for measures of risk, and call the measures satisfying these properties "coherent", and demonstrate the universality of scenario-based methods for providing coherent measures.
Abstract: In this paper we study both market risks and nonmarket risks, without complete markets assumption, and discuss methods of measurement of these risks. We present and justify a set of four desirable properties for measures of risk, and call the measures satisfying these properties “coherent.” We examine the measures of risk provided and the related actions required by SPAN, by the SEC=NASD rules, and by quantile-based methods. We demonstrate the universality of scenario-based methods for providing coherent measures. We offer suggestions concerning the SEC method. We also suggest a method to repair the failure of subadditivity of quantile-based methods.
8,651 citations
TL;DR: In this paper, a new approach to optimize or hedging a portfolio of financial instruments to reduce risk is presented and tested on applications, which focuses on minimizing Conditional Value-at-Risk (CVaR) rather than minimizing Value at Risk (VaR), but portfolios with low CVaR necessarily have low VaR as well.
Abstract: A new approach to optimizing or hedging a portfolio of nancial instruments to reduce risk is presented and tested on applications. It focuses on minimizing Conditional Value-at-Risk (CVaR) rather than minimizing Value-at-Risk (VaR), but portfolios with low CVaR necessarily have low VaR as well. CVaR, also called Mean Excess Loss, Mean Shortfall, or Tail VaR, is anyway considered to be a more consistent measure of risk than VaR. Central to the new approach is a technique for portfolio optimization which calculates VaR and optimizes CVaR simultaneously. This technique is suitable for use by investment companies, brokerage rms, mutual funds, and any business that evaluates risks. It can be combined with analytical or scenario-based methods to optimize portfolios with large numbers of instruments, in which case the calculations often come down to linear programming or nonsmooth programming. The methodology can be applied also to the optimization of percentiles in contexts outside of nance.
5,622 citations
Book•
26 Jan 2012
TL;DR: In this article, the authors present a model predictive controller for a water heating system, which is based on the T Polynomial Process (TOP) model of the MPC.
Abstract: 1 Introduction to Model Predictive Control.- 1.1 MPC Strategy.- 1.2 Historical Perspective.- 1.3 Industrial Technology.- 1.4 Outline of the Chapters.- 2 Model Predictive Controllers.- 2.1 MPC Elements.- 2.1.1 Prediction Model.- 2.1.2 Objective Function.- 2.1.3 Obtaining the Control Law.- 2.2 Review of Some MPC Algorithms.- 2.3 State Space Formulation.- 3 Commercial Model Predictive Control Schemes.- 3.1 Dynamic Matrix Control.- 3.1.1 Prediction.- 3.1.2 Measurable Disturbances.- 3.1.3 Control Algorithm.- 3.2 Model Algorithmic Control.- 3.2.1 Process Model and Prediction.- 3.2.2 Control Law.- 3.3 Predictive Functional Control.- 3.3.1 Formulation.- 3.4 Case Study: A Water Heater.- 3.5 Exercises.- 4 Generalized Predictive Control.- 4.1 Introduction.- 4.2 Formulation of Generalized Predictive Control.- 4.3 The Coloured Noise Case.- 4.4 An Example.- 4.5 Closed-Loop Relationships.- 4.6 The Role of the T Polynomial.- 4.6.1 Selection of the T Polynomial.- 4.6.2 Relationships with Other Formulations.- 4.7 The P Polynomial.- 4.8 Consideration of Measurable Disturbances.- 4.9 Use of a Different Predictor in GPC.- 4.9.1 Equivalent Structure.- 4.9.2 A Comparative Example.- 4.10 Constrained Receding Horizon Predictive Control.- 4.10.1 Computation of the Control Law.- 4.10.2 Properties.- 4.11 Stable GPC.- 4.11.1 Formulation of the Control Law.- 4.12 Exercises.- 5 Simple Implementation of GPC for Industrial Processes.- 5.1 Plant Model.- 5.1.1 Plant Identification: The Reaction Curve Method.- 5.2 The Dead Time Multiple of the Sampling Time Case.- 5.2.1 Discrete Plant Model.- 5.2.2 Problem Formulation.- 5.2.3 Computation of the Controller Parameters.- 5.2.4 Role of the Control-weighting Factor.- 5.2.5 Implementation Algorithm.- 5.2.6 An Implementation Example.- 5.3 The Dead Time Nonmultiple of the Sampling Time Case.- 5.3.1 Discrete Model of the Plant.- 5.3.2 Controller Parameters.- 5.3.3 Example.- 5.4 Integrating Processes.- 5.4.1 Derivation of the Control Law.- 5.4.2 Controller Parameters.- 5.4.3 Example.- 5.5 Consideration of Ramp Setpoints.- 5.5.1 Example.- 5.6 Comparison with Standard GPC.- 5.7 Stability Robustness Analysis.- 5.7.1 Structured Uncertainties.- 5.7.2 Unstructured Uncertainties.- 5.7.3 General Comments.- 5.8 Composition Control in an Evaporator.- 5.8.1 Description of the Process.- 5.8.2 Obtaining the Linear Model.- 5.8.3 Controller Design.- 5.8.4 Results.- 5.9 Exercises.- 6 Multivariable Model Predictive Control.- 6.1 Derivation of Multivariable GPC.- 6.1.1 White Noise Case.- 6.1.2 Coloured Noise Case.- 6.1.3 Measurable Disturbances.- 6.2 Obtaining a Matrix Fraction Description.- 6.2.1 Transfer Matrix Representation.- 6.2.2 Parametric Identification.- 6.3 State Space Formulation.- 6.3.1 Matrix Fraction and State Space Equivalences.- 6.4 Case Study: Flight Control.- 6.5 Convolution Models Formulation.- 6.6 Case Study: Chemical Reactor.- 6.6.1 Plant Description.- 6.6.2 Obtaining the Plant Model.- 6.6.3 Control Law.- 6.6.4 Simulation Results.- 6.7 Dead Time Problems.- 6.8 Case Study: Distillation Column.- 6.9 Multivariable MPC and Transmission Zeros.- 6.9.1 Simulation Example.- 6.9.2 Tuning MPC for Processes with OUD Zeros.- 6.10 Exercises.- 7 Constrained Model Predictive Control.- 7.1 Constraints and MPC.- 7.1.1 Constraint General Form.- 7.1.2 Illustrative Examples.- 7.2 Constraints and Optimization.- 7.3 Revision of Main Quadratic Programming Algorithms.- 7.3.1 The Active Set Methods.- 7.3.2 Feasible Direction Methods.- 7.3.3 Initial Feasible Point.- 7.3.4 Pivoting Methods.- 7.4 Constraints Handling.- 7.4.1 Slew Rate Constraints.- 7.4.2 Amplitude Constraints.- 7.4.3 Output Constraints.- 7.4.4 Constraint Reduction.- 7.5 1-norm.- 7.6 Case Study: A Compressor.- 7.7 Constraint Management.- 7.7.1 Feasibility.- 7.7.2 Techniques for Improving Feasibility.- 7.8 Constrained MPC and Stability.- 7.9 Multiobjective MPC.- 7.9.1 Priorization of Objectives.- 7.10 Exercises.- 8 Robust Model Predictive Control.- 8.1 Process Models and Uncertainties.- 8.1.1 Truncated Impulse Response Uncertainties.- 8.1.2 Matrix Fraction Description Uncertainties.- 8.1.3 Global Uncertainties.- 8.2 Objective Functions.- 8.2.1 Quadratic Cost Function.- 8.2.2 ?-? norm.- 8.2.3 1-norm.- 8.3 Robustness by Imposing Constraints.- 8.4 Constraint Handling.- 8.5 Illustrative Examples.- 8.5.1 Bounds on the Output.- 8.5.2 Uncertainties in the Gain.- 8.6 Robust MPC and Linear Matrix Inequalities.- 8.7 Closed-Loop Predictions.- 8.7.1 An Illustrative Example.- 8.7.2 Increasing the Number of Decision Variables.- 8.7.3 Dynamic Programming Approach.- 8.7.4 Linear Feedback.- 8.7.5 An Illustrative Example.- 8.8 Exercises.- 9 Nonlinear Model Predictive Control.- 9.1 Nonlinear MPC Versus Linear MPC.- 9.2 Nonlinear Models.- 9.2.1 Empirical Models.- 9.2.2 Fundamental Models.- 9.2.3 Grey-box Models.- 9.2.4 Modelling Example.- 9.3 Solution of the NMPC Problem.- 9.3.1 Problem Formulation.- 9.3.2 Solution.- 9.4 Techniques for Nonlinear Predictive Control.- 9.4.1 Extended Linear MPC.- 9.4.2 Local Models.- 9.4.3 Suboptimal NPMC.- 9.4.4 Use of Short Horizons.- 9.4.5 Decomposition of the Control Sequence.- 9.4.6 Feedback Linearization.- 9.4.7 MPC Based on Volterra Models.- 9.4.8 Neural Networks.- 9.4.9 Commercial Products.- 9.5 Stability and Nonlinear MPC.- 9.6 Case Study: pH Neutralization Process.- 9.6.1 Process Model.- 9.6.2 Results.- 9.7 Exercises.- 10 Model Predictive Control and Hybrid Systems.- 10.1 Hybrid System Modelling.- 10.2 Example: A Jacket Cooled Batch Reactor.- 10.2.1 Mixed Logical Dynamical Systems.- 10.2.2 Example.- 10.3 Model Predictive Control of MLD Systems.- 10.3.1 Branch and Bound Mixed Integer Programming.- 10.3.2 An Illustrative Example.- 10.4 Piecewise Affine Systems.- 10.4.1 Example: Tankwith Different Area Sections.- 10.4.2 Reach Set, Controllable Set, and STG Algorithm.- 10.5 Exercises.- 11 Fast Methods for Implementing Model Predictive Control.- 11.1 Piecewise Affinity of MPC.- 11.2 MPC and Multiparametric Programming.- 11.3 Piecewise Implementation of MPC.- 11.3.1 Illustrative Example: The Double Integrator.- 11.3.2 Nonconstant References and Measurable Disturbances.- 11.3.3 Example.- 11.3.4 The 1-norm and ?-norm Cases.- 11.4 Fast Implementation of MPC forUncertain Systems.- 11.4.1 Example.- 11.4.2 The Closed-Loop Min-max MPC.- 11.5 Approximated Implementation for MPC.- 11.6 Fast Implementation of MPC and Dead Time Considerations.- 11.7 Exercises.- 12 Applications.- 12.1 Solar Power Plant.- 12.1.1 Selftuning GPC Control Strategy.- 12.1.2 Gain Scheduling Generalized Predictive Control.- 12.2 Pilot Plant.- 12.2.1 Plant Description.- 12.2.2 Plant Control.- 12.2.3 Flow Control.- 12.2.4 Temperature Control at the Exchanger Output.- 12.2.5 Temperature Control in the Tank.- 12.2.6 Level Control.- 12.2.7 Remarks.- 12.3 Model Predictive Control in a Sugar Refinery.- 12.4 Olive Oil Mill.- 12.4.1 Plant Description.- 12.4.2 Process Modelling and Validation.- 12.4.3 Controller Synthesis.- 12.4.4 Experimental Results.- 12.5 Mobile Robot.- 12.5.1 Problem Definition.- 12.5.2 Prediction Model.- 12.5.3 Parametrization of the Desired Path.- 12.5.4 Potential Function for Considering Fixed Obstacles.- 12.5.5 The Neural Network Approach.- 12.5.6 Training Phase.- 12.5.7 Results.- A Revision of the Simplex Method.- A.1 Equality Constraints.- A.2 Finding an Initial Solution.- A.3 Inequality Constraints.- B Dynamic Programming and Linear Quadratic Optimal Control.- B.1 LinearQuadratic Problem.- B.2 InfiniteHorizon.- References.
3,913 citations
TL;DR: Since its popularization in the late 1970s, Sequential Quadratic Programming (SQP) has arguably become the most successful method for solving nonlinearly constrained optimization problems.
Abstract: Since its popularization in the late 1970s, Sequential Quadratic Programming (SQP) has arguably become the most successful method for solving nonlinearly constrained optimization problems. As with most optimization methods, SQP is not a single algorithm, but rather a conceptual method from which numerous specific algorithms have evolved. Backed by a solid theoretical and computational foundation, both commercial and public-domain SQP algorithms have been developed and used to solve a remarkably large set of important practical problems. Recently large-scale versions have been devised and tested with promising results.
1,765 citations
09 May 2011
TL;DR: It is experimentally show that the stochastic nature of STOMP allows it to overcome local minima that gradient-based methods like CHOMP can get stuck in.
Abstract: We present a new approach to motion planning using a stochastic trajectory optimization framework. The approach relies on generating noisy trajectories to explore the space around an initial (possibly infeasible) trajectory, which are then combined to produced an updated trajectory with lower cost. A cost function based on a combination of obstacle and smoothness cost is optimized in each iteration. No gradient information is required for the particular optimization algorithm that we use and so general costs for which derivatives may not be available (e.g. costs corresponding to constraints and motor torques) can be included in the cost function. We demonstrate the approach both in simulation and on a mobile manipulation system for unconstrained and constrained tasks. We experimentally show that the stochastic nature of STOMP allows it to overcome local minima that gradient-based methods like CHOMP can get stuck in.
817 citations