RRT*former: Environment-Aware Sampling-Based Motion Planning using Transformer

TL;DR

RRT*former addresses the challenge of sampling-based motion planning in complex and dynamic environments by integrating a Transformer-based sampler into the RRT* framework. The model processes environmental features through a CNN to form $d_{model}$-dimensional embeddings and uses an encoder-only Transformer to generate the next sampling node conditioned on past samples, with a stop criterion based on a distance threshold to the goal. A hybrid sampling strategy balances uniform exploration with transformer-guided sampling via a parameter $\\alpha$, improving sampling efficiency and reducing the number of nodes required to find feasible paths. Experimental results in 2D/3D simulations and Gazebo demonstrate faster convergence, lower initial path costs, and competitive final costs compared to NRRT* and NIRRT*, highlighting practical advantages for real-time, dynamic environments.

Abstract

We investigate the sampling-based optimal path planning problem for robotics in complex and dynamic environments. Most existing sampling-based algorithms neglect environmental information or the information from previous samples. Yet, these pieces of information are highly informative, as leveraging them can provide better heuristics when sampling the next state. In this paper, we propose a novel sampling-based planning algorithm, called \emph{RRT*former}, which integrates the standard RRT* algorithm with a Transformer network in a novel way. Specifically, the Transformer is used to extract features from the environment and leverage information from previous samples to better guide the sampling process. Our extensive experiments demonstrate that, compared to existing sampling-based approaches such as RRT*, Neural RRT*, and their variants, our algorithm achieves considerable improvements in both the optimality of the path and sampling efficiency. The code for our implementation is available on https://github.com/fengmingyang666/RRTformer.

Figures (7)

• Figure 1: Sampler Model. The Sampler consists of three parts. Feature Extractor: extract features from the environment using CNN. Transformer Encoder: generate new sample from previous sampling information and environment features. Condition Validator: determine when to stop sampling by checking whether the new sample is close enough to the goal.
• Figure 2: Base Algorithm
• Figure 3: RRT* Sampler
• Figure 4: Transformer Mixed Sampler
• Figure 5: Demonstration of the random tree generated under different randomization ratios $\alpha$.