The Shadow DEX-EE is a three-fingered dexterous hand engineered specifically for the demands of reinforcement learning and imitation learning research, where hardware is routinely subjected to repeated impacts, high contact forces, and unpredictable behaviors from untrained policies. Each finger is a self-contained, user-replaceable module—allowing a damaged finger to be swapped out while repairs are completed separately, minimizing downtime in long-running experiments. At approximately 350 mm in height and 4.1 kg for the three-finger configuration, the DEX-EE is larger than a human hand and built for endurance rather than compactness.

The hand delivers 12 degrees of freedom across its three fingers, with tendon-driven actuation using an N+1 motor configuration per finger for added reliability. Control loops operate at up to 10 kHz for force control within motor units and 1 kHz for higher-level position control on a host computer. Each finger carries 155 individual sensor channels covering position, force, and inertial measurements, plus a camera-based distal tactile sensor outputting video at 1280×480 resolution and 50 FPS over a single USB 3.0 connection. A DEX-EE Chiral variant repositions the third finger to resemble a human thumb offset, supporting teleoperation and bi-manual task workflows.