An official website of the United States government
Recent work has demonstrated that motion planners’ performance can be significantly improved by retrieving past experiences from a database. Typically, the experience database is queried for past similar problems using a similarity function defined over the motion planning problems. However, to date, most works rely on simple hand-crafted similarity functions and fail to generalize outside their corresponding training dataset. To address this limitation, we propose (FIRE), a framework that extracts local representation of planning problems and learns a similarity function over them. To generate the training data we introduce a novel self-supervised method that identifies similar and dissimilar pairs of local primitives from past solution paths. With these pairs, a Siamese network is trained with the contrastive loss and the similarity function is realized in the network’s latent space. We evaluate FIRE on an 8-DOF manipulator in five categories of motion planning problems with sensed environments. Our experiments show that FIRE retrieves relevant experiences which can informatively guide sampling-based planners even in problems outside its training distribution, outperforming other baselines.
more »
« less
Learning Sampling Distributions Using Local 3D Workspace Decompositions for Motion Planning in High Dimensions
Earlier work has shown that reusing experience from prior motion planning problems can improve the efficiency of similar, future motion planning queries. However, for robots with many degrees-of-freedom, these methods exhibit poor generalization across different environments and often require large datasets that are impractical to gather. We present SPARK and FLAME, two experience-based frameworks for sampling-based planning applicable to complex manipulators in 3D environments. Both combine samplers associated with features from a workspace decomposition into a global biased sampling distribution. SPARK decomposes the environment based on exact geometry while FLAME is more general, and uses an octree-based decomposition obtained from sensor data. We demonstrate the effectiveness of SPARK and FLAME on a real and simulated Fetch robot tasked with challenging pick-and-place manipulation problems. Our approaches can be trained incrementally and significantly improve performance with only a handful of examples, generalizing better over diverse tasks and environments as compared to prior approaches.
more »
« less
- Award ID(s):
- 1718478
- PAR ID:
- 10280236
- Date Published:
- Journal Name:
- Proceedings of the International Conference on Robotics and Automation 2021
- ISSN:
- 1050-4729
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Proving motion planning infeasibility is an important part of a complete motion planner. Common approaches for high-dimensional motion planning are only probabilistically complete. Previously, we presented an algorithm to construct infeasibility proofs by applying machine learning to sampled configurations from a bidirectional sampling-based planner. In this work, we prove that the learned manifold converges to an infeasibility proof exponentially. Combining prior approaches for sampling-based planning and our converging infeasibility proofs, we propose the term asymptotic completeness to describe the property of returning a plan or infeasibility proof in the limit. We compare the empirical convergence of different sampling strategies to validate our analysis.more » « less
-
Approaches to autonomous navigation for unmanned ground vehicles rely on motion planning algorithms that optimize maneuvers under kinematic and environmental constraints. Algorithms that combine heuristic search with local optimization are well suited to domains where solution optimality is favored over speed and memory resources are limited as they often improve the optimality of solutions without increasing the sampling density. To address the runtime performance limitations of such algorithms, this paper introduces Predictively Adapted State Lattices, an extension of recombinant motion planning search space construction that adapts the representation by selecting regions to optimize using a learned model trained to predict the expected improvement. The model aids in prioritizing computations that optimize regions where significant improvement is anticipated. We evaluate the performance of the proposed method through statistical and qualitative comparisons to alternative State Lattice approaches for a simulated mobile robot with nonholonomic constraints. Results demonstrate an advance in the ability of recombinant motion planning search spaces to improve relative optimality at reduced runtime in varyingly complex environments.more » « less
-
null (Ed.)Robot motion planning is one of the important elements in robotics. In environments full of obstacles, it is always challenging to find a collision-free and dynamically feasible path between the robot's initial configuration and goal configuration. While many motion planning algorithms have been proposed in the past, each of them has its pros and cons. This work presents a benchmark which implements and compares existing planning algorithms on a variety of problems with extensive simulation. Based on that, we also propose a hybrid planning algorithm, RRT*-CFS, that combines the merits of sampling-based planning methods and optimization-based planning methods. The first layer, RRT*, quickly samples a semi-optimal path. The second layer, CFS, performs sequential convex optimization given the reference path from RRT*. The proposed RRT*-CFS has feasibility and convergence guarantees. Simulation results show that RRT*-CFS benefits from the hybrid structure and performs robustly in various scenarios including the narrow passage problems.more » « less
-
Mobile manipulation tasks such as opening a door, pulling open a drawer, or lifting a toilet lid require constrained motion of the end-effector under environmental and task constraints. This, coupled with partial information in novel environments, makes it challenging to employ classical motion planning approaches at test time. Our key insight is to cast it as a learning problem to leverage past experience of solving similar planning problems to directly predict motion plans for mobile manipulation tasks in novel situations at test time. To enable this, we develop a simulator, ArtObjSim, that simulates articulated objects placed in real scenes. We then introduce SeqIK+θ0, a fast and flexible representation for motion plans. Finally, we learn models that use SeqIK+θ0 to quickly predict motion plans for articulating novel objects at test time. Experimental evaluation shows improved speed and accuracy at generating motion plans than pure search-based methods and pure learning methods.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
