moleworksResearch in robotics and machine autonomy.
Paper & video
Perception, planning, and learned control coordinate bulk material transfer on a 40-ton hydraulic machine.
This research combines a learned grasp-location planner, obstacle-aware motion planning, and learned hydraulic controllers to pick up, transport, and unload material with a freely swinging gripper.
Field video
A hydraulic excavator engages the ground and lifts material in a field research demonstration.
Excavation brings the machine into sustained contact with the ground. This field clip shows bucket engagement through lift, one of the material-interaction tasks that motivates our work on reusable manipulation skills.
Paper & video
A policy trained in simulation uses sparse LiDAR and machine feedback to lift boulders with a standard bucket.
Tested on a 12-ton excavator, the learned policy coordinates the arm and bucket while adapting its digging motion to different soil resistance. It uses sparse LiDAR and machine feedback; an operator initializes the target segmentation.