ELMP: Efficient Learning for Motion Planning via Analytical Policy Gradients
2026-06-30 • Robotics
Robotics
AI summaryⓘ
The authors present ELMP, a new way to quickly adapt robot motion planners to new environments without needing large expert datasets. Instead of generating expensive expert data, ELMP fine-tunes itself using direct feedback from the robot's movements and collisions. They also improve generalization by including the shape of the robot's tools as point clouds. Their method works much faster and more reliably than traditional approaches and was tested on a real robot arm.
Neural Motion Plannersself-supervised learningfine-tuningkinematic chainpoint cloudscollision detectionrobot motion planningFranka Emika Pandainference latency
Authors
Yixiao Li, Tifanny Portela, Jordis Herrmann, René Zurbrügg, Marco Hutter
Abstract
Neural Motion Planners (NMPs) enable fast reactive motion generation, but adapting them to new environments typically requires recollecting large expert datasets, which is computationally prohibitive. We propose ELMP, a framework for data-efficient adaptation via self-supervised fine-tuning. Rather than generating additional expert trajectories with expensive global planners, ELMP directly optimizes the policy through a differentiable kinematic layer using dense collision, target-reaching, and smoothness objectives. This replaces expert data generation with rapid problem sampling, reducing per-sample adaptation cost by roughly two orders of magnitude. To further support robust generalization across changing kinematic chains, we introduce a mechanism to explicitly encode tool geometry via point clouds. Benchmarked against classical and neural baselines, ELMP achieves an 84.8% average success rate with orders-of-magnitude lower cold-start latency than classical methods. In unseen environments, self-supervised fine-tuning improves success rate from 57.3% (zero-shot) to 89.8%, removing the data collection bottleneck. Our approach maintains millisecond-level inference latency and is validated on a physical Franka Emika Panda robot.