Header image  
SGER: Feasible control tasks on battery powered embedded

The PI proposes a research program to establish theoretical foundation for analysis and design of control tasks on embedded computers powered by battery. In modern day applications, feedback control laws and control algorithms are typically implemented as control tasks that are scheduled and executed in an embedded computer under constraints imposed by limited computing resources and limited energy budget. A set of control tasks are feasible if they are schedulable and guarantee performance for physical plants. Since the majority of control tasks are close to periodic, the periods of all control tasks can be modeled
as an (random) operation point in a constrained multi-dimensional nonlinear space. The location of the operation point influences control performance. In addition to control laws, the PI proposes to use computing delay and sampling period as feedback control mechanism. This timing control is implemented as online feedback scheduling algorithms that adjust the location of operation points according to measured computing time and control law performance. A major theoretical contribution of this proposal is to model the operation points as virtual particles with kinetic properties determined by transient performance. Thus the
problem of designing feedback scheduling algorithms can be solved by methods inspired by robot navigation, path following, and obstacle avoidance.