Deep-RL policy-transfer paper reports hardware results on ATRIAS

Evidence notes

• The paper studies whether deep reinforcement-learning policies can transfer from simulation to hardware when embedded inside a structured walking controller.
• Experiments on ATRIAS examine action spaces, cost functions, and the rate of transfer from high-fidelity simulation to the physical biped.
• The result is a historically important ATRIAS entry for combining learned policies with interpretable controller structure.