will propose a doctoral thesis entitled,
Reconfigurable Reliable Robotics
On
[date & time] Monday, December 2 at 3:00 p.m.
[building & room] Montgomery Knight Building 317
Abstract
Due to emerging on-demand mobility, many recent topics of researches are focused on various types of autonomous electric vehicle systems such as a self-driving car, a delivery drone service, and urban air mobility. Two crucial considerations of these autonomous systems are reliability and safety. This research addresses a formal framework for provably guaranteeing reliability and safety in the context of safety-critical cyber-physical systems, which apply to previously mentioned real-world systems. This framework consists of three main contributions. The first one develops a design automation technique for assuring absolute reliability. This design method adopts an Aerospace Recommend Practice documents as standards for formulating a redundancy optimization problem based on a geometric program. The solution to this problem indicates the minimum number of redundancy components, e.g., sensors, computing units, and actuators, needed for achieving the desired reliability.
The second contribution focuses on an approach that mitigates the consequential effects of failures that could occur on those redundancy components. This approach provides two reliability optimization-based methodology for hardware reconfiguring or software reallocating. Depending on where a fault has occurred on the system, the reconfiguration/reallocation optimization problem can be cast as either a mixed-integer linear programming problem or a rank optimization problem.
The third contribution is shifted toward a safety aspect of the cyber-physical system concerning component failures. Notably, the aim is to ensure that the system's state-space always stays in a known safe-space during its operation despite the loss or a partial loss of its controllability or observability. This idea utilizes concepts of a control allocation problem and a set invariance principle, and provides a quadratic programming optimization problem for calculating an optimal control policy.
In addition to numerical examples provided at the end of each subsection, this proposed framework will be realizable through experimental implementations that demonstrate the reconfigurable reliable robotic system.
Committee