Attention-Based Map Encoding (AME)

"Attention-Based Map Encoding for Learning Generalized Legged Locomotion": an end-to-end RL controller with a terrain-attention map encoder that unifies agile locomotion across quadruped (ANYmal) and humanoid robots over challenging, generalized terrain. From Marco Hutter's Robotic Systems Lab, continuing the ANYmal RL-locomotion lineage; Science Robotics 2025.