Fast Swept Volume Estimation with Deep Learning

Chiang, Hao-Tien Lewis; Faust, Aleksandra; Sugaya, Satomi; Tapia, Lydia

doi:10.1007/978-3-030-44051-0_4

Cited by 10 publications

(7 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Second, we briefly demonstrate that swept volume geometries ( S V ( c 1 , c 2 ) ) can be accurately and efficiently estimated by DNNs. This paper extends our previous work (Chiang et al 2018) through the following additional contributions. First, we comprehensively evaluate the accuracy of DNN distance measure learning on seven robots, encompassing rigid-body, prismatic, and revolute joints and closed-loop kinematic chain motion.…”

Section: Introductionsupporting

confidence: 72%

Fast deep swept volume estimator

Chiang

Baxter

Sugaya

et al. 2020

The International Journal of Robotics Research

Self Cite

View full text Add to dashboard Cite

Despite decades of research on efficient swept volume computation for robotics, computing the exact swept volume is intractable and approximate swept volume algorithms have been computationally prohibitive for applications such as motion and task planning. In this work, we employ deep neural networks (DNNs) for fast swept volume estimation. Since swept volume is a property of robot kinematics, a DNN can be trained off-line once in a supervised manner and deployed in any environment. The trained DNN is fast during on-line swept volume geometry or size inferences. Results show that DNNs can accurately and rapidly estimate swept volumes caused by rotational, translational, and prismatic joint motions. Sampling-based planners using the learned distance are up to five times more efficient and identify paths with smaller swept volumes on simulated and physical robots. Results also show that swept volume geometry estimation with a DNN is over 98.9% accurate and 1,200 times faster than an octree-based swept volume algorithm.

show abstract

Section: Introductionsupporting

confidence: 72%

Fast deep swept volume estimator

Chiang

Baxter

Sugaya

et al. 2020

The International Journal of Robotics Research

Self Cite

View full text Add to dashboard Cite

show abstract

“…One approach seeks to identify and evaluate only important regions of the robot's state space, thus reducing the number of required collision checks. Some methods learn heuristics for promising paths [25], [5], learn a distribution of promising regions [18], or evaluate only the most promising edges [6]. Our approach, which learns to accelerate the actual collision checking procedure, is complementary and can be used in conjunction with these approaches.…”

Section: Related Workmentioning

confidence: 99%

Neural Collision Clearance Estimator for Batched Motion Planning

Kew

Ichter

Bandari

et al. 2021

Springer Proceedings in Advanced Robotics

Self Cite

View full text Add to dashboard Cite

Collision checking is a well known bottleneck in sampling-based motion planning due to its computational expense and the large number of checks required. To alleviate this bottleneck, we present a fast neural network collision checking heuristic, ClearanceNet, and incorporate it within a planning algorithm, ClearanceNet-RRT (CN-RRT). ClearanceNet takes as input a robot pose and the location of all obstacles in the workspace and learns to predict the clearance, i.e., distance to nearest obstacle. CN-RRT then efficiently computes a motion plan by leveraging three key features of ClearanceNet. First, as neural network inference is massively parallel, CN-RRT explores the space via a parallel RRT, which expands nodes in parallel, allowing for thousands of collision checks at once. Second, CN-RRT adaptively relaxes its clearance threshold for more difficult problems. Third, to repair errors, CN-RRT shifts states towards higher clearance through a gradient-based approach that uses the analytic gradient of ClearanceNet. Once a path is found, any errors are repaired via RRT over the misclassified sections, thus maintaining the theoretical guarantees of sampling-based motion planning. We evaluate the collision checking speed, planning speed, and motion plan efficiency in configuration spaces with up to 30 degrees of freedom. The collision checking achieves speedups of more than two orders of magnitude over traditional collision detection methods. Sampling-based planning over multiple robotic arms in new environment configurations achieves speedups of up to 51% over a baseline, with paths up to 25% more efficient. Experiments on a physical Fetch robot reaching into shelves in a cluttered environment confirm the feasibility of this method on real robots.

show abstract

“…In this work, our chance constraints are also influenced by the cross-communication uncertainty of a heterogeneous robot team (see III-B). Additionally, while obstacle constraints can be explicitly used in the optimization, other works show a learning-based approach to avoid collisions by modeling the distribution of promising regions for travel [10] or predicting the separation distance between the robot and its surroundings [11]. Our work is a hybrid approach, where we use the RNN to predict uncertainty of state estimates (which affect the 'size' of polyhedral obstacles), but still use these obstacles as constraints in our SMPC optimization.…”

Section: Relative Distances To Obstacles/global Goalmentioning

confidence: 99%

“…Equation ( 14) returns a positive distance if the robot is 'outside' the obstacle boundary, and a negative distance if the robot is 'inside' the obstacle boundary (a negative distance would cause the line constraint to fail). By definition of constraint (10), only one or more of the line constraints need to be satisfied per obstacle O j , which is ensured by using binary integer variables under constraints ( 15) and ( 16) (e.g., for line constraint i belonging to O j , if I i,j = 1, the robot is outside this obstacle). For simplicity, we assume we have a 'perfect' object detection system.…”

Section: A Stochastic Model Predictive Control Formulationmentioning

confidence: 99%

SABER: Data-Driven Motion Planner for Autonomously Navigating Heterogeneous Robots

Schperberg,

Tsuei,

Soatto

et al. 2021

Preprint

View full text Add to dashboard Cite

Fast Swept Volume Estimation with Deep Learning

Cited by 10 publications

References 19 publications

Fast deep swept volume estimator

Fast deep swept volume estimator

Neural Collision Clearance Estimator for Batched Motion Planning

SABER: Data-Driven Motion Planner for Autonomously Navigating Heterogeneous Robots

Contact Info

Product

Resources

About