Learning Equality Constraints for Motion Planning on Manifolds
Tags: Deep reinforcement learning, Motion planning

Constrained robot motion planning is a widely used technique to solve complex robot tasks. We consider the problem of learning representations of constraints from demonstrations with a deep neural network, which we call Equality Constraint Manifold Neural Network (ECoMaNN). The idea behind it is to learn a level-set function of the constraint by aligning subspaces in the network with sub-spaces of the data such that it can be integrated into a constrained sampling-based motion planner. We combine both learned constraints and analytically described constraints into the planner and use a projection-based strategy to find valid points. We evaluate ECoMaNN on its representation capabilities of constraint manifolds, the impact of its individual loss terms, and the motions produced with it.

Related Publications

Sutanto, Giovanni; Rayas Fernández, Isabel M.; Englert, Peter; Ramachandran, Ragesh K.; Sukhatme, Gaurav S.: Learning Equality Constraints for Motion Planning on Manifolds. CoRR, vol. abs/2009.11852, 2020
[PDF] [Video]