Learning and Autonomous Control

The Learning and Autonomous Control (LAC) team focuses on robot control methods spanning the whole range from high-level cognitive approaches including adaptation and learning to low-level motion control. We develop novel algorithms, methods, and solutions in the following areas of robotics: 

• control of mechanical systems - model-based and data-driven control of complex Lagrangian dynamics, control-based exploitation of physical intelligence, computational design, reduced-order modeling of high-dimensional systems;
• robot autonomy - knowledge processing systems, motion planning, real-time inference for decision-making, model predictive control, and social interaction with other agents;
• robot learning - reinforcement learning, iterative learning control, and deep learning for teaching robots interactively, from human demonstrations and also for continual improvement of the robot’s performance;
• multi-robot control - real-time coordination, dynamic vehicle routing and task assignment, multi-robot learning, and security and privacy aspects of multi-robot systems; 

We apply our novel methods to industrial robotics, (mobile) manipulation, locomotion, soft robots, exoskeletons, agri-food and retail robotics, healthcare robotics, autonomous driving, aerial vehicles, and intelligent transportation.