2023 Fiscal Year Annual Research Report
Sampling-guided symbolic control framework under changing environments
| Project/Area Number |
21K14191
|
|---|
| Research Institution | National Institute of Advanced Industrial Science and Technology |
|---|
Principal Investigator |
プルエクプラサート サシニー 国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 研究員 (50814795)
|
|---|
| Project Period (FY) |
2021-04-01 – 2024-03-31
|
| Keywords | symbolic control / control theory / motion planning / robotics |
|---|
| Outline of Annual Research Achievements |
This research project studies safe control frameworks for semi-controlled environments in which some disturbance may occur. The overall goal of the project is to develop safe and efficient symbolic-structure-based control and verification techniques that are robust enough to handle systems under some nondeterminism and uncertainties. Throughout the research period, we developed efficient symbolic control algorithms for safety verification and control under complex specifications such as temporal logic, and demonstrated their performances by simulations on nonholonomic robots. Moreover, we studied a safe learning approach using discrete symbolic structures.
During the final fiscal year of the project, we also developed a moment approximation method of a stochastic polynomial system, which can be used for system safety verification. Our technique is to estimate the moments of a stochastic polynomial system with finitely many states using Carleman linearization with truncation to obtain a finite-dimensional linear system. We provide efficient online computation methods for this scheme with several error bounds for the approximation. We then demonstrated the effectiveness of our proposed method by simulations on a logistic map with stochastic dynamics and vehicle dynamics subject to stochastic disturbance.
|
