Score-Guided Motion Planning: Learning the Gradient Field of Promising Regions¶
Authors: Shaochen Wang, Qilin Wu, Qing Huang, Zhuo Cheng
Published: 2026 (Conference Paper)
Source: ICASSP 2026 - 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Algorithm: Score-Guided Motion Planning
DOI: 10.1109/ICASSP55912.2026.11460477
Summary¶
Abstract¶
In this paper, a score-guided sampling framework, ScorePlanner, is proposed to address the critical challenge of sampling inefficiency in sampling-based motion planning. ScorePlanner fundamentally reframes the sampling problem: instead of relying on blind, uniform exploration, it introduces a guided generation process. This process leverages a learned score function, the gradient of the log-probability of promising configurations, to actively steer samples towards the connective backbone of the promising space. ScorePlanner can be seamlessly integrated as a modular sampler into foundational planners like RRT and RRT* to dramatically improve sampling efficiency and accelerate convergence to high-quality solutions. Extensive experiments in a variety of challenging environments are conducted to demonstrate the performance gains of the proposed framework.