From Target Selection to Digging: A Learning-Based Framework for Continuous Autonomous Excavation

2026-09-24

AI Summary

This study addresses the challenges of target adaptation and motion coordination arising from dynamic pile geometry changes during continuous excavation. We propose an autonomous excavation framework integrating terrain perception with reinforcement learning (RL) and imitation learning (IL). By coordinating RL and IL policies through a shared interface and leveraging LiDAR-derived elevation maps for dynamic target selection, the framework decouples task-oriented motion from local digging, enabling closed-loop autonomous operation of hydraulic excavators. Experimental results demonstrate that the proposed system achieves an average payload of 6.52 kg per cycle, substantially outperforming the 2.68 kg baseline, while significantly reducing local motion duration. Furthermore, it successfully executes five consecutive stable excavation cycles, validating its efficiency and adaptability in complex operational scenarios.

Abstract

Repeated excavation continuously reshapes pile geometry, requiring an autonomous excavator to adapt its digging targets and coordinate motion across successive excavation cycles. We present a learning-based framework for continuous autonomous excavation that integrates terrain-aware target selection with reinforcement- and imitation-learning controllers. The framework separates target-conditioned motion from local digging: a shared task-conditioned RL policy controls waypoint-guided approach and loaded transport, while an IL policy learns vision-based digging and lifting from expert demonstrations. Digging targets are selected from LiDAR elevation maps and converted into bucket-tip waypoints for motion control. The control architecture coordinates the learned policies and deterministic unloading through a shared motion interface. The complete system is deployed on a scaled hydraulic excavator with multimodal sensing and closed-loop actuator control. Offline replay and physical experiments demonstrate more consistent target selection, shorter local motion time, and increased payload compared with the respective baselines. The learned digging policy achieves a mean payload of 6.52 kg per completed cycle, compared with 2.68 kg for Fixed Dig. Three five-scoop runs further demonstrate consecutive autonomous excavation under continuously changing pile geometry.

Problem

• autonomous excavation
• target selection
• pile geometry
• continuous digging
• motion coordination

Innovation

• Autonomous Excavation
• Reinforcement Learning
• Imitation Learning
• Target Selection
• Policy Decoupling