Many adaptive and learning control problems can be formulated as optimization problems aiming at minimizing an objective function related to system performance: in this context, several research directions arise such as optimal control formulations for uncertain systems via the Bellman’s optimality principle, or solving the exploitation-exploration dilemma of controlling the system as well as possible based on current system knowledge (exploitation) or experimenting with the system so as to learn about its behavior and control it better in the future (exploration).
Representative Themes
- Adaptive dynamic programming / Approximate dynamic programming
- Distributed learning and distributed optimization
- Cognitive adaptive control, adaptive optimal control
- Adaptive dual control
Representative Applications
- Self-optimizing energy efficient systems (e.g. of energy systems)
- Self-optimizing formations (e.g. of unmanned vehicles)
- Learning-based robots
Representative Results
- Yue D., Baldi S., Cao J., and De Schutter B., “Distributed adaptive optimization with weight-balancing”, IEEE Transactions on Automatic Control, scheduled for publication, 2021. doi:10.1109/TAC.2021.3071651
- Dai P., Yu W., Wang H., and Baldi S., “Distributed actor-critic algorithms for multi-agent reinforcement learning over directed graphs”, IEEE Transactions on Neural Networks and Learning Systems, scheduled for publication, 2022. doi:10.1109/TNNLS.2021.3139138
- Michailidis I., Baldi S., Kosmatopoulos E. B., and Ioannou P. A., “Adaptive optimal control for large-scale nonlinear systems”. IEEE Transactions on Automatic Control, Vol. 62(11), pp. 5567-5577, 2017. doi:10.1109/TAC.2017.2684458