A Cross-Environment and Cross-Embodiment Path Planning Framework via a Conditional Diffusion Model

TL;DR

GADGET presents a diffusion-based path planning framework that generalizes to unseen environments and robotic embodiments without retraining by conditioning trajectories on voxelized scene geometry and start/goal states, and by integrating safety through a CBF-inspired guidance during diffusion. The method combines classifier-free scene conditioning with CBF-based safety shaping to produce collision-free, near-optimal joint-space trajectories across multiple robot arms, demonstrated in spherical, bin-picking, and shelf tasks, with real-world Kinova Gen3 deployment. Key contributions include a geometry-centric voxel encoding, end-to-end conditional training, inference-time safety guidance, and a robot-adaptable formulation enabling zero-shot transfer across serial-chain manipulators. The approach achieves high success rates with low collision intensity and competitive path lengths, and it demonstrates practical potential for deployment-ready motion planning in dynamic manufacturing contexts. Overall, GADGET unifies environment-aware generation and safety-aware inference to provide scalable, cross-domain motion planning for diverse hardware without retraining.

Abstract

Path planning for a robotic system in high-dimensional cluttered environments needs to be efficient, safe, and adaptable for different environments and hardware. Conventional methods face high computation time and require extensive parameter tuning, while prior learning-based methods still fail to generalize effectively. The primary goal of this research is to develop a path planning framework capable of generalizing to unseen environments and new robotic manipulators without the need for retraining. We present GADGET (Generalizable and Adaptive Diffusion-Guided Environment-aware Trajectory generation), a diffusion-based planning model that generates joint-space trajectories conditioned on voxelized scene representations as well as start and goal configurations. A key innovation is GADGET's hybrid dual-conditioning mechanism that combines classifier-free guidance via learned scene encoding with classifier-guided Control Barrier Function (CBF) safety shaping, integrating environment awareness with real-time collision avoidance directly in the denoising process. This design supports zero-shot transfer to new environments and robotic embodiments without retraining. Experimental results show that GADGET achieves high success rates with low collision intensity in spherical-obstacle, bin-picking, and shelf environments, with CBF guidance further improving safety. Moreover, comparative evaluations indicate strong performance relative to both sampling-based and learning-based baselines. Furthermore, GADGET provides transferability across Franka Panda, Kinova Gen3 (6/7-DoF), and UR5 robots, and physical execution on a Kinova Gen3 demonstrates its ability to generate safe, collision-free trajectories in real-world settings.

Figures (9)

• Figure 1: Forward and reverse (denoising) process of Diffusion model.
• Figure 2: Generalizable and Adaptive Diffusion-Guided Environment-aware Trajectory Generation (GADGET) framework.
• Figure 3: Simulation environments. (a) Spherical Obstacles: Used for training; during training and testing, obstacle positions are randomized. (b) Bin Picking: A test-only environment not encountered during training. (c) Shelf Manipulation: Another test-only environment unseen during training, representing structured manipulation tasks. In all scenes, the red shadow of the robot denotes the target (goal) configuration.
• Figure 4: Success rate (left) and collision intensity (right) across spherical, bin, and shelf environments. GADGET outperforms learning-based baselines (MPNet, MPD, DP3), achieving higher success with lower collision rates, while CBF guidance further enhances safety.
• Figure 5: t-SNE visualization of scene embeddings for spherical, bin, and shelf environments. The embeddings form distinct clusters, confirming that the latent distributions differ across environment types while still supporting generalization in planning.