Benchmarking SAC & TD3 with FeedForward, LSTM, BiLSTM, and Transformer encoders on BipedalWalkerHardcore-v3. Best score: 303.6. Includes 8 pretrained models, training logs, evaluation videos, and full reproduction scripts.
reinforcement-learning deep-learning robotics pytorch gymnasium continuous-control actor-critic bipedal-walker
-
Updated
Apr 2, 2026 - TeX