Improving Trajectory Stitching with Flow Models

TL;DR

This paper introduces a novel flow model-based approach to improve trajectory stitching in generative models, enabling better planning for complex robotic tasks including obstacle avoidance, both in simulation and real hardware.

Contribution

It identifies limitations of existing generative models in trajectory stitching and proposes architectural, dataset, and training enhancements to overcome these issues.

Findings

Significantly improved obstacle avoidance capabilities.

Effective planning with out-of-distribution boundary conditions.

Superior performance over baselines in simulation and real-world tests.

Abstract

Generative models have shown great promise as trajectory planners, given their affinity to modeling complex distributions and guidable inference process. Previous works have successfully applied these in the context of robotic manipulation but perform poorly when the required solution does not exist as a complete trajectory within the training set. We identify that this is a result of being unable to plan via stitching, and subsequently address the architectural and dataset choices needed to remedy this. On top of this, we propose a novel addition to the training and inference procedures to both stabilize and enhance these capabilities. We demonstrate the efficacy of our approach by generating plans with out of distribution boundary conditions and performing obstacle avoidance on the Franka Panda in simulation and on real hardware. In both of these tasks our method performs significantly better than the baselines and is able to avoid obstacles up to four times as large.