Potential Based Diffusion Motion Planning
This addresses motion planning for robotics, offering a composable solution that generalizes to various constraints, though it appears incremental as it builds on potential-based methods.
The paper tackles the problem of motion planning in high-dimensional spaces by training a neural network to learn easily optimizable potentials over trajectories, significantly outperforming classical and recent learned methods and avoiding local minima.
Effective motion planning in high dimensional spaces is a long-standing open problem in robotics. One class of traditional motion planning algorithms corresponds to potential-based motion planning. An advantage of potential based motion planning is composability -- different motion constraints can be easily combined by adding corresponding potentials. However, constructing motion paths from potentials requires solving a global optimization across configuration space potential landscape, which is often prone to local minima. We propose a new approach towards learning potential based motion planning, where we train a neural network to capture and learn an easily optimizable potentials over motion planning trajectories. We illustrate the effectiveness of such approach, significantly outperforming both classical and recent learned motion planning approaches and avoiding issues with local minima. We further illustrate its inherent composability, enabling us to generalize to a multitude of different motion constraints.