Jianqiang Ding

To free humans from hazardous tasks, robotics must be flexible enough to operate safely in critical environments.
Ideally, mobile robots like drones and autonomous vehicles must be able to avoid collisions with potential obstacles, and power systems must maintain stability despite fluctuations in energy supply. With the ever-growing complexity of these systems and the concurrent surge of data available for control design, deriving provable data-driven control strategies have become crucial for achieving reliable control of complex robotic systems in safe-critical environments.

This research aims to develop a theoretical and computational framework that enables the deployment of data-driven nonlinear control algorithms in real world scenarios with provable guarantees. The focus of this study is not only on the realization of nonlinear systems with rigorous error estimation and the synthesis of uncertainty-aware formal certificates, but also on the improvements of interpretability when integrating learning techniques.  By investigating prediction error bounds of linear approximation approaches, such as Koopman theory and feedback linearization methodology, the research seeks to establish a formal framework for accurate state prediction in response to control inputs. Furthermore, this study also explores to deliver rigorous state estimation for real-world nonlinear systems when considering noise in data collection and controller deployment. Furthermore, this research investigates the construction of uncertainty-aware formal certificates to serve as metrics to quantify deviations of system behaviors from requirements, providing provable guarantees for the safe implementation of data-driven control policies.

To advance the reliable control of increasingly complex robotics, the ultimate goal of this research is to develop a theoretical framework for provably trustworthy decision making in robotic systems, ensuring operational safety and achieving design objectives in safety-critical and uncertain environments.