STAMP: Differentiable Task and Motion Planning via Stein Variational Gradient Descent

TL;DR

STAMP introduces a differentiable, gradient-based approach to task and motion planning that efficiently finds multiple diverse solutions by relaxing the problem into a continuous domain and leveraging parallelized physics simulation.

Contribution

This work presents a novel TAMP method using Stein Variational Gradient Descent to optimize discrete and continuous plans simultaneously within a differentiable physics framework.

Findings

Finds multiple diverse plans in a single run

Significantly faster than existing TAMP methods

Effective on various robotics planning problems

Abstract

Planning for sequential robotics tasks often requires integrated symbolic and geometric reasoning. TAMP algorithms typically solve these problems by performing a tree search over high-level task sequences while checking for kinematic and dynamic feasibility. This can be inefficient because, typically, candidate task plans resulting from the tree search ignore geometric information. This often leads to motion planning failures that require expensive backtracking steps to find alternative task plans. We propose a novel approach to TAMP called Stein Task and Motion Planning (STAMP) that relaxes the hybrid optimization problem into a continuous domain. This allows us to leverage gradients from differentiable physics simulation to fully optimize discrete and continuous plan parameters for TAMP. In particular, we solve the optimization problem using a gradient-based variational inference algorithm called Stein Variational Gradient Descent. This allows us to find a distribution of solutions within a single optimization run. Furthermore, we use an off-the-shelf differentiable physics simulator that is parallelized on the GPU to run parallelized inference over diverse plan parameters. We demonstrate our method on a variety of problems and show that it can find multiple diverse plans in a single optimization run while also being significantly faster than existing approaches.