Differentiable Spatial Planning using Transformers
This addresses the problem of efficient and generalizable path planning for robotics and AI systems, offering a novel method that improves over existing data-driven approaches.
The paper tackles spatial path planning by learning a differentiable planner from data, proposing Spatial Planning Transformers (SPT) that plan over long-range dependencies, and it outperforms prior state-of-the-art methods with absolute improvements of 7-19% in manipulation and navigation tasks.
We consider the problem of spatial path planning. In contrast to the classical solutions which optimize a new plan from scratch and assume access to the full map with ground truth obstacle locations, we learn a planner from the data in a differentiable manner that allows us to leverage statistical regularities from past data. We propose Spatial Planning Transformers (SPT), which given an obstacle map learns to generate actions by planning over long-range spatial dependencies, unlike prior data-driven planners that propagate information locally via convolutional structure in an iterative manner. In the setting where the ground truth map is not known to the agent, we leverage pre-trained SPTs in an end-to-end framework that has the structure of mapper and planner built into it which allows seamless generalization to out-of-distribution maps and goals. SPTs outperform prior state-of-the-art differentiable planners across all the setups for both manipulation and navigation tasks, leading to an absolute improvement of 7-19%.