cpRRTC: GPU-Parallel RRT-Connect for Constrained Motion Planning

TL;DR

The paper presents cpRRTC, a GPU-accelerated RRT-Connect framework for constrained motion planning that extends prior GPU methods with constrained projection and intra-block communication. It introduces a parallel projection operator and NVRTC-based runtime code generation to enable early termination and improved parallelism within each thread block. Experimental results show cpRRTC outperforms pRRTC and cuRobo, particularly as obstacle density and constraint complexity increase, indicating strong scalability and practical impact for high-dimensional robotic planning. The approach enhances real-time feasibility of constrained trajectories in cluttered environments, leveraging GPU parallelism and shared memory to accelerate both unconstrained and constrained planning tasks.

Abstract

Motion planning is a fundamental problem in robotics that involves generating feasible trajectories for a robot to follow. Recent advances in parallel computing, particularly through CPU and GPU architectures, have significantly reduced planning times to the order of milliseconds. However, constrained motion planning especially using sampling based methods on GPUs remains underexplored. Prior work such as pRRTC leverages a tracking compiler with a CUDA backend to accelerate forward kinematics and collision checking. While effective in simple settings, their approach struggles with increased complexity in robot models or environments. In this paper, we propose a novel GPU based framework utilizing NVRTC for runtime compilation, enabling efficient handling of high complexity scenarios and supporting constrained motion planning. Experimental results demonstrate that our method achieves superior performance compared to existing approaches.

Figures (7)

• Figure 1: Different ways to project motion segment The curve line is the manifold, while the goal is projecting motion segment to manifold from one blue configurations to another. Upper: The original way to project motion segment in sequence. Lower: The parallel way to project motion segment.
• Figure 2: Illustrative example of the obstacle–segmentation procedure. Left: original Bookshelf Tall scene (1x). Right: 10x version obtained by recursively subdividing each box. Each segmented box is represented by a different color.
• Figure 3: Cumulative distribution of solution times for the Fetch arm on MotionBenchMaker.
• Figure 4: Constrained motion planning task with obstacle(red cuboids). Green plane and line are constrained regions for end-effector.
• Figure 5: Cumulative distribution of solution times for the Fetch arm under line and plane constraint primitives.