Control system theory plays an instrumental role in enabling reliable autonomous operations, including SLAM (simultaneous localization and mapping), path planning, trajectory tracking, etc. In particular, physics-based optimization policies such as LQR (linear quadratic regulator), MPC (model predictive control), and MHE (moving horizon estimation), etc., have been the powerhouse behind many advanced robot applications. On the other hand, deep learning methods can often achieve state-of-the-art performance and generalizability through training on a large amount of data in a model-free fashion.
There is an increasing interest in combining them to achieve the best of both worlds: efficient training, fast adaptation to environments, safety and robust guarantees, semantic understanding, agile and dexterous performance, etc. For example, incorporating system models can help significantly reduce training samples and time. Especially, recent works have developed efficient analytical gradients for differentiation through MPC and MHE policies, enabling seamless embedding of optimal policies into deep neural networks, which achieved superior control performance.
However, fundamental challenges still exist in further pushing the performance boundary in both areas. This invited session aims to provide a platform for a diverse group of researchers and practitioners from different areas to report their recent research and application progress on learning and control for autonomous robots.
Topics of interest include, but are not limited to:
- Learning-based control
- Safe reinforcement learning
- Joint optimization of MPC, MHE with neural networks
- Hyperparameter optimization and bilevel optimization
- Hybrid model-free and model-based optimal control
- Distributed learning and optimization for autonomous robots
- Meta-learning, continual learning for autonomous robots
- Neural ODE for adaptive optimal control
Accepted Paper list and Schedule
| Time | Title | Presenter(s) |
|---|---|---|
| 16:00 - 16:20 | Learning for Online Mixed-Integer Model Predictive Control with Parametric Optimality Certificates | Nair, Siddharth, University of California, Berkeley Russo, Luigi, Università del Sannio Glielmo, Luigi, University of Napoli Federico II Borrelli, Francesco, University of California at Berkeley |
| 16:20 - 16:40 | Nonlinear MPC for Quadrotors in Close-Proximity Flight with Neural Network Downwash Prediction | Li, Jinjie, Beihang University Han, Liang, Beihang University Yu, Haoyang, Beihang University Lin, Yuheng, Beihang University Li, Qingdong, Beihang University Ren, Zhang, Beijing University of Aeronautics and Astronautics |
| 16:40 - 17:00 | Optimal Scheduling for Remote Estimation with an Auxiliary Transmission Scheme | Li, Zitian, Guangdong University of Technology Yang, Lixin, Queensland University of Technology Jia, Yijin, Guangdong University of Technology Huang, Zenghong, Guangdong University of Technology Lv, Weijun, Guangdong University of Technology Xu, Yong, Guangdong University of Technology |
| 17:00 - 17:20 | Deriving Rewards for Reinforcement Learning from Symbolic Behaviour Descriptions of Bipedal Walking | Harnack, Daniel, German Research Center for Artificial Intelligence, DFKI Lüth, Christoph, Deutsches Forschungszentrum für Künstliche Intelligenz Gross, Lukas, DFKI Kumar, Shivesh, German Research Center for Artificial Intelligence, DFKI GmbH Kirchner, Frank, Robotics Innovation Center, DFKI and Department of Mathematics and Informatics, University of Bremen |
| 17:20 - 17:40 | Optimizing Field-of-View for Multi-Agent Path Finding via Reinforcement Learning: A Performance and Communication Overhead Study | Cheng, Hoi Chuen, The Hong Kong University of Science and Technology Shi, Ling, The Hong Kong University of Science and Technology Yue, Chik Patrick, The Hong Kong University of Science and Technology |
| 17:40 - 18:00 | Learning Koopman Operators with Control Using Bi-level Optimization | Huang, Daning, Pennsylvania State University Prasetyo, Muhammad Bayu, Pennsylvania State University Yu, Yin, Penn State University Geng, Junyi, Pennsylvania State University |
Organizers
National University of Singapore |
University at Buffalo |
Carnegie Mellon University |
Carnegie Mellon University |
Purdue University |