Robot hand study reports higher rotation performance with symmetric fingers

A robotics preprint reports that a symmetric finger arrangement on the low-cost LEAP Hand achieved a 32.0% higher mean object rotation count and a 25.8% higher aggregate rotation rate than an anthropomorphic arrangement. Both hands used vision-based tactile sensors and a shared reinforcement learning framework. Hardware experiments tested tactile policies on ten objects under multiple hand orientations. Additional hardware force recordings showed a 19.9% lower mean total fingertip load for the symmetric arrangement. The study also tested 12 unseen objects in simulation, where richer tactile feedback improved rotation progress for both designs. The experiments concern continuous rotation within a robot hand, limiting how broadly the performance gains apply.