Evaluating model-based planning and planner amortization for continuous control
DeepMind（深度思维）；Facebook Reality Labs（脸书现实实验室）
Arunkumar Byravan, Leonard Hasenclever, Piotr Trochim, Mehdi Mirza, Alessandro Davide Ialongo, Yuval Tassa, Jost Tobias Springenberg, Abbas Abdolmaleki, Nicolas… 展开作者
Arunkumar Byravan, Leonard Hasenclever, Piotr Trochim, Mehdi Mirza, Alessandro Davide Ialongo, Yuval Tassa, Jost Tobias Springenberg, Abbas Abdolmaleki, Nicolas Heess, Josh Merel, Martin Riedmiller
备注：
9 pages main text, 30 pages with references and appendix including several ablations and additional experiments. Submitted to ICLR 2022
There is a widespread intuition that model-based control methods should be able to surpass the data efficiency of model-free approaches. In this paper we attempt to evaluate this intuition on various challenging locomotion tasks. We take a hybrid approach, combining model predictive control (MPC) with a learned model and model-free policy learning; the learned policy serves as a proposal for MPC. We find that well-tuned model-free agents are strong baselines even for high DoF control problems but MPC with learned proposals and models (trained on the fly or transferred from related tasks) can significantly improve performance and data efficiency in hard multi-task/multi-goal settings. Finally, we show that it is possible to distil a model-based planner into a policy that amortizes the planning computation without any loss of performance. Videos of agents performing different tasks can be seen at https://sites.google.com/view/mbrl-amortization/home.
Institute of Industrial and Control Engineering, Universitat Politècnica de Catalunya（加泰罗尼亚理工大学工业与控制工程研究所）
Motion planning has evolved from coping with simply geometric problems to physics-based ones that incorporate the kinodynamic and the physical constraints imposed by the robot and the physical world. Therefore, the criteria for evaluating physics-based motion planners goes beyond the computational complexity (e.g. in terms of planning time) usually used as a measure for evaluating geometrical planners, in order to consider also the quality of the solution in terms of dynamical parameters. This study proposes an evaluation criteria and analyzes the performance of several kinodynamic planners, w…
National Institute of Advanced Industrial Science and Technology (AIST)（日本国立产业技术综合研究所（AIST））；Osaka University（大阪大学）
This paper develops a planner to find an optimal assembly sequence to assemble several objects. The input to the planner is the mesh models of the objects, the relative poses between the objects in the assembly, and the final pose of the assembly. The output is an optimal assembly sequence, namely (1) in which order should one assemble the objects, (2) from which directions should the objects be dropped, and (3) candidate grasps of each object. The proposed planner finds the optimal solution by automatically permuting, evaluating, and searching the possible assembly sequences considering stabi…
Rice University（莱斯大学）；TU Berlin（柏林工业大学）；University of Wisconsin-Madison（威斯康星大学麦迪逊分校）
Recently, there has been a wealth of development in motion planning for robotic manipulation new motion planners are continuously proposed, each with their own unique strengths and weaknesses. However, evaluating new planners is challenging and researchers often create their own ad-hoc problems for benchmarking, which is time-consuming, prone to bias, and does not directly compare against other state-of-the-art planners. We present MotionBenchMaker, an open-source tool to generate benchmarking datasets for realistic robot manipulation problems. MotionBenchMaker is designed to be an extensible,…
Stony Brook University（石溪大学）；University of Pennsylvania（宾夕法尼亚大学）
High-performance autonomy often must operate at the boundaries of safety. When external agents are present in a system, the process of ensuring safety without sacrificing performance becomes extremely difficult. In this paper, we present an approach to stress test such systems based on the rapidly exploring random tree (RRT) algorithm. We propose to find faults in such systems through adversarial agent perturbations, where the behaviors of other agents in an otherwise fixed scenario are modified. This creates a large search space of possibilities, which we explore both randomly and with a focu…
University of Toronto（多伦多大学）；Université de Montréal（蒙特利尔大学）；Microsoft（微软）；Facebook AI Research（脸书人工智能研究院）；Vector Institute（向量研究所）
3D scene graphs (3DSGs) are an emerging description; unifying symbolic, topological, and metric scene representations. However, typical 3DSGs contain hundreds of objects and symbols even for small environments; rendering task planning on the full graph impractical. We construct TASKOGRAPHY, the first large-scale robotic task planning benchmark over 3DSGs. While most benchmarking efforts in this area focus on vision-based planning, we systematically study symbolic planning, to decouple planning performance from visual representation learning. We observe that, among existing methods, neither cla…
Computer Science and Artificial Intelligence Laboratory (CSAIL), MIT（麻省理工学院计算机科学与人工智能实验室（CSAIL））；Tagup, Inc.（Tagup公司）；GRASP Laboratory, University of Pennsylvania（宾夕法尼亚大学GRASP实验室）
We describe a task and motion planning architecture for highly dynamic systems that combines a domain-independent sampling-based deliberative planning algorithm with a global reactive planner. We leverage the recent development of a reactive, vector field planner that provides guarantees of reachability to large regions of the environment even in the face of unknown or unforeseen obstacles. The reachability guarantees can be formalized using contracts that allow a deliberative planner to reason purely in terms of those contracts and synthesize a plan by choosing a sequence of reactive behavior…
Robowflex is a software library for robot motion planning in industrial and research applications, leveraging the popular MoveIt library and Robot Operating System (ROS) middleware. Robowflex provides an augmented API for crafting and manipulating motion planning queries within a single program, making motion planning with MoveIt easy. Robowflex's high-level API simplifies many common use-cases while still providing low-level access to the MoveIt library when needed. Robowflex is particularly useful for 1) developing new motion planners, 2) evaluating motion planners, and 3) complex problems t…
A Survey on Hybrid Motion Planning Methods for Automated Driving Systems
WMG, University of Warwick（华威大学WMG学院）；School of Electronics, Electrical Engineering and Computer Science, Queen’s University of Belfast（贝尔法斯特女王大学电子、电气工程与计算机科学学院）
Look as You Leap: Planning Simultaneous Motion and Perception for High-DOF Robots
Department of Computer Science, Rice University（计算机科学系，里士大学）；Ken Kennedy Institute, Rice University（肯尼迪研究所，里士大学）；Houston Methodist Academic Institute（休斯顿 Methodist 学术研究院）
We present RAPPIDS: a novel collision checking and planning algorithm for multicopters that is capable of quickly finding local collision-free trajectories given a single depth image from an onboard camera. The primary contribution of this work is a new pyramid-based spatial partitioning method that enables rapid collision detection between candidate trajectories and the environment. By leveraging the efficiency of our collision checking method, we shown how a local planning algorithm can be run at high rates on computationally constrained hardware, evaluating thousands of candidate trajectori…
Gradient-based planners are widely used for quadrotor local planning, in which a Euclidean Signed Distance Field (ESDF) is crucial for evaluating gradient magnitude and direction. Nevertheless, computing such a field has much redundancy since the trajectory optimization procedure only covers a very limited subspace of the ESDF updating range. In this paper, an ESDF-free gradient-based planning framework is proposed, which significantly reduces computation time. The main improvement is that the collision term in the penalty function is formulated by comparing the colliding trajectory with a col…
This paper introduces a method to compute a sparse lattice planner control set that is suited to a particular task by learning from a representative dataset of vehicle paths. To do this, we use a scoring measure similar to the Fréchet distance and propose an algorithm for evaluating a given control set according to the scoring measure. Control actions are then selected from a dense control set according to an objective function that rewards improvements in matching the dataset while also encouraging sparsity. This method is evaluated across several experiments involving real and synthetic data…
University of North Carolina at Chapel Hill（北卡罗来纳大学教堂山分校）；University of North Carolina School of Medicine（北卡罗来纳大学医学院）
The prospect of using autonomous robots to enhance the capabilities of physicians and enable novel procedures has led to considerable efforts in developing medical robots and incorporating autonomous capabilities. Motion planning is a core component for any such system working in an environment that demands near perfect levels of safety, reliability, and precision. Despite the extensive and promising work that has gone into developing motion planners for medical robots, a standardized and clinically-meaningful way to compare existing algorithms and evaluate novel planners and robots is not wel…
Stanford University（斯坦福大学）；NVIDIA（英伟达）；University of Toronto（多伦多大学）；Vector Institute（向量研究所）
Evaluating and improving planning for autonomous vehicles requires scalable generation of long-tail traffic scenarios. To be useful, these scenarios must be realistic and challenging, but not impossible to drive through safely. In this work, we introduce STRIVE, a method to automatically generate challenging scenarios that cause a given planner to produce undesirable behavior, like collisions. To maintain scenario plausibility, the key idea is to leverage a learned model of traffic motion in the form of a graph-based conditional VAE. Scenario generation is formulated as an optimization in the …
Asymptotically Optimal Lazy Lifelong Sampling-based Algorithm for Efficient Motion Planning in Dynamic Environments
Department of Mechanical Engineering, City University of Hongkong（机械工程系，香港城市大学）；Department of Electronic and Electrical Engineering, Southern University of Science and Technology（电子与电气工程系，南方科技大学）；Department of Electronic and Computer Engineering, Hong Kong University of Science and Technology（电子与计算机工程系，香港科技大学）
Decision-making under uncertainty is a crucial ability for autonomous systems. In its most general form, this problem can be formulated as a Partially Observable Markov Decision Process (POMDP). The solution policy of a POMDP can be implicitly encoded as a value function. In partially observable settings, the value function is typically learned via forward simulation of the system evolution. Focusing on accurate and long-range risk assessment, we propose a novel method, where the value function is learned in different phases via a bi-directional search in belief space. A backward value learnin…
FRENETIX: A High-Performance and Modular Motion Planning Framework for Autonomous Driving
Institute of Automotive Technology, Technical University of Munich（慕尼黑工业大学汽车技术研究所）；Munich Institute of Robotics and Machine Intelligence（慕尼黑机器人与机器智能研究所）；TUM School of Engineering and Design, Technical University of Munich（慕尼黑工业大学工程与设计学院）；Professorship of Cyber-physical Systems, Technical University of Munich（慕尼黑工业大学信息物理系统教授席位）
GeCCo -- a Generalist Contact-Conditioned Policy for Loco-Manipulation Skills on Legged Robots
Dynamic Robot Systems Group, Oxford Robotics Institute, University of Oxford（牛津大学机器人研究所动态机器人系统组）；Sony AI（索尼人工智能）；Dyson Institute of Engineering and Technology（戴森工程与技术研究所）
