Embodied Intelligence, Robot Control, and Optimization: From Perception to Physical Action

Recent breakthroughs in embodied AI have shifted the paradigm from purely data-driven learning to intelligence that emerges from physical interaction with the environment. Unlike traditional systems operating in closed, structured spaces, embodied agents—humanoid robots, quadrupeds, dexterous manipulators—must perceive, reason, and act under real-world uncertainty. However, key challenges remain: bridging the sim-to-real gap, ensuring safety during learning, and achieving real-time optimization of complex, high-dimensional control loops. This column invites cutting-edge research at the intersection of embodied AI, control theory, and numerical optimization. 

·Call For Paper

- End-to-end learning for sensorimotor control, tactile feedback integration, and active perception

- Robot Control: Model predictive control (MPC) for locomotion and manipulation, adaptive and robust control under disturbances

- Optimization for Robotics: Real-time trajectory optimization, differentiable simulation for policy learning, and contact-implicit optimization

- Learning and Control Synergies: Reinforcement learning shaped by control priors, Lyapunov-based safe learning, and hybrid model-based/model-free approaches

· Chair