Motion planning and controller design are essential in many autonomous applications, such as self-driving cars, surveillance with drones, and assistive robotics for home or medical applications. In these areas, the dynamical systems can be characterized as cyber-physical systems, in which the physical parts can be modeled by differential equations in continuous time, while digital computers are used to perform discrete state sampling and control updates. However, many existing frameworks were developed solely for the cyber part of the system. Ignoring continuous phenomena during discrete-time implementation could render undesired and even unsafe behavior. In this dissertation, I developed a motion planning and control framework for safety-critical systems to satisfy spatial and temporal constraints under a zeroth-order hold control implementation.
In the first part of the dissertation, I introduce a self-triggered controller, with control barrier functions constraints for safety and control Lyapunov functions constraints for stability. The controls are generated by solving quadratic programs sequentially over time. Instead of using a periodic control mechanism, where the controls are updated at a constant rate, I proactively calculate the next update time instant, given the constraints, to ensure continuous-time safety and stability.
In the second part of the dissertation, I introduce a correct-by-construction control synthesis framework where the system is required to satisfy Signal Temporal Logic formulas. I utilize the lower bounds of the control barrier functions and the predicate functions over a time interval to ensure continuous-time satisfaction of a formula.
The focus of the third part of the dissertation is a sampling-based motion planner. The motion planner is based on Rapidly-exploring Random Trees. I introduce a function that generates collision-free trajectories and controls on-the-fly to avoid separate collision checks during edge extensions. The developed motion planner works for nonlinear systems and can be efficiently solved in real-time.
| Identifer | oai:union.ndltd.org:bu.edu/oai:open.bu.edu:2144/41481 |
| Date | 29 September 2020 |
| Creators | Yang, Guang |
| Contributors | Belta, Calin A., Tron, Roberto |
| Source Sets | Boston University |
| Language | en_US |
| Detected Language | English |
| Type | Thesis/Dissertation |
| Rights | Attribution 4.0 International, http://creativecommons.org/licenses/by/4.0/ |
Page generated in 0.0024 seconds