Posterior Sampling for Anytime Motion Planning on Graphs with Expensive-to-Evaluate Edges

Hou, Brian; Choudhury, Sanjiban; Lee, Gilwoo; Mandalika, Aditya; Srinivasa, Siddhartha S.

doi:10.1109/icra40945.2020.9197014

Cited by 9 publications

(10 citation statements)

References 34 publications

(39 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…If the goal is not reached the robot receives some large negative reward.2. Observations ψ are shared and the obstacles remain fixed between iterations.This Repeated BTP is analogous to the Experienced Lazy Path Search problem, for which Thompson sampling (within an algorithm called PSMP) has bounded regret compared to the optimal policy always taking the shortest path (Hou et al 2020). Consider the strategy that attempts the path from Thompson sampling, and if a collision occurs backtracks to the start and executes the shortest path found so far.…”

Section: A1 Heuristic Estimates Of Q-valuesmentioning

confidence: 99%

The blindfolded traveler’s problem: A search framework for motion planning with contact estimates

Saund¹,

Choudhury

Srinivasa

et al. 2023

The International Journal of Robotics Research

Self Cite

View full text Add to dashboard Cite

We address the problem of robot motion planning under uncertainty where the only observations are through contact with the environment. Such problems are typically solved by planning optimistically assuming unknown space is free, moving along the planned path and re-planning if the robot collides. However this approach can be very inefficient, leading to many unnecessary collisions and unproductive motion. We propose a new formulation, the Blindfolded Traveler’s Problem (BTP), for planning on a graph containing edges with unknown validity, with true validity observed only through attempted traversal by the robot. The solution to a BTP is a policy indicating the next edge to attempt given previous observations and an initial belief. We prove that BTP is NP-complete and show that exact modeling of the belief is intractable, therefore we present several approximation-based policies and beliefs. For the policy we propose graph search with edge weights augmented by the probability of collision. For the belief representation we propose a weighted Mixture of Experts of Collision Hypothesis Sets and a Manifold Particle Filter. Empirical evaluation in simulation and on a real robot arm shows that our proposed approach vastly outperforms several baselines as well as a previous approach that does not employ the BTP framework.

show abstract

Section: A1 Heuristic Estimates Of Q-valuesmentioning

confidence: 99%

The blindfolded traveler’s problem: A search framework for motion planning with contact estimates

Saund¹,

Choudhury

Srinivasa

et al. 2023

The International Journal of Robotics Research

Self Cite

View full text Add to dashboard Cite

show abstract

“…While (Cohen, Phillips, and Likhachev 2015;Haghtalab et al 2018;Mandalika et al 2019) are designed to search directly for the optimal path, our Algorithm 1 is an Anytime algorithm. Note that PSMP (Hou et al 2020) is also based on LAZYSP and is an anytime algorithm. However, it is not related to our problem.…”

Section: Related Workmentioning

confidence: 99%

MA3: Model-Accuracy Aware Anytime Planning with Simulation Verification for Navigating Complex Terrains

Das

Conover

Eum

et al. 2022

SOCS

View full text Add to dashboard Cite

Off-road and unstructured environments often contain complex patches of various types of terrain, rough elevation changes, deformable objects, etc. An autonomous ground vehicle traversing such environments experiences physical interactions that are extremely hard to model at scale and thus very hard to predict. Nevertheless, planning a safely traversable path through such an environment requires the ability to predict the outcomes of these interactions instead of avoiding them. One approach to doing this is to learn the interaction model offline based on collected data. Unfortunately, though, this requires large amounts of data and can often be brittle. Alternatively, models using physics-based simulators can generate large data and provide a reliable prediction. However, they are very slow to query online within the planning loop. This work proposes an algorithmic framework that utilizes the combination of a learned model and a physics-based simulation model for fast planning. Specifically, it uses the learned model as much as possible to accelerate planning while sparsely using the physics-based simulator to verify the feasibility of the planned path. We provide a theoretical analysis of the algorithm and its empirical evaluation showing a significant reduction in planning times.

show abstract

“…• Identifying samples that are guaranteed to be valid (Bialkowski et al, 2016;Bialkowski et al, 2013) • Using a learned model in place of a collision detector (Burns and Brock, 2005;Huh and Lee, 2016;Das and Yip, 2020;Kew et al, 2020;Yu and Gao, 2021) • Determining the order in which to collision check nodes or edges (Pan et al, 2013;Bhardwaj et al, 2019;Choudhury et al, 2017;Choudhury et al, 2018;Hou et al, 2020)…”

Section: Categories Of Work On Collision Checkingmentioning

confidence: 99%

“…Posterior Sampling for Motion Planning (PSMP) (Hou et al, 2020) formulates anytime search on graphs as an instance of Bayesian Reinforcement Learning (Bayesian RL). Unlike prior work, PSMP aims for anytime performance by leveraging learned posteriors on edge collisions to quickly discover an initial feasible path and progressively yield shorter paths.…”

Section: Determining the Order In Which To Collision Check Nodes/edgesmentioning

confidence: 99%

A Survey on the Integration of Machine Learning with Sampling-based Motion Planning

et al. 2022

View full text Add to dashboard Cite

Sampling-based methods are widely adopted solutions for robot motion planning. The methods are straightforward to implement, effective in practice for many robotic systems. It is often possible to prove that they have desirable properties, such as probabilistic completeness and asymptotic optimality. Nevertheless, they still face challenges as the complexity of the underlying planning problem increases, especially under tight computation time constraints, which impact the quality of returned solutions or given inaccurate models. This has motivated machine learning to improve the computational efficiency and applicability of Sampling-Based Motion Planners (SBMPs). This survey reviews such integrative efforts and aims to provide a classification of the alternative directions that have been explored in the literature. It first discusses how learning has been used to enhance key components of SBMPs, such as node sampling, collision detection, distance or nearest neighbor computation, local

show abstract

Posterior Sampling for Anytime Motion Planning on Graphs with Expensive-to-Evaluate Edges

Cited by 9 publications

References 34 publications

The blindfolded traveler’s problem: A search framework for motion planning with contact estimates

The blindfolded traveler’s problem: A search framework for motion planning with contact estimates

MA3: Model-Accuracy Aware Anytime Planning with Simulation Verification for Navigating Complex Terrains

A Survey on the Integration of Machine Learning with Sampling-based Motion Planning

Contact Info

Product

Resources

About