THREAD: Trajectory Planning for Hybrid Rigid-Soft Manipulators with Environment-Aware Diffusion

2026-06-19

AI Summary

This work addresses the challenges of trajectory planning for hybrid rigid-soft robotic arms operating in confined spaces, where environmental contacts often render planned trajectories infeasible and the kinematic coupling between rigid and soft segments is commonly neglected. To overcome these limitations, we propose the first diffusion model–based trajectory planner that learns a prior over physically feasible backbone trajectories under local geometric constraints. Our approach jointly models curvature, smoothness, and collision avoidance to enable end-to-end, environment-aware trajectory generation while explicitly capturing the motion coupling between rigid and soft segments. Leveraging a physics-informed loss function and a cross-embodiment transfer strategy—combining simulation pretraining with minimal online fine-tuning—our method achieves a 92.4% task success rate in simulation (reducing collisions by a factor of five) and successfully threads the arm through a real-world slit only 1.3 times the diameter of its soft segment.

Abstract

Manipulation in confined environments, such as threading a manipulator through narrow apertures, remains a fundamental challenge, especially for conventional rigid robots. Hybrid rigid-soft manipulators offer promise but face two compounding planning challenges: backbone shapes feasible in free space become infeasible under environmental contact, and planning rigid and soft segments independently ignores their kinematic coupling. We present THREAD, the first diffusion-based trajectory planner for hybrid manipulation, learning a generative prior over physically realizable backbone trajectories conditioned on local environment geometry, with physics-inspired losses encoding curvature, smoothness, and collision constraints jointly across both segments. Trained in simulation, THREAD achieves 92.4% task success with 5x fewer collisions than the strongest baseline. We show cross-embodiment real-world transfer with minimal online updates, successfully threading through apertures as small as 1.3x the soft segment diameter.

Problem

• trajectory planning
• hybrid rigid-soft manipulators
• confined environments
• kinematic coupling
• environmental contact

Innovation

• diffusion-based planning
• hybrid rigid-soft manipulators
• environment-aware trajectory generation
• kinematic coupling
• physics-informed constraints