|Institution:||University of Michigan|
|Keywords:||Stochastic fuel-efficient cruise control; Constrained control; Aerospace Engineering; Engineering|
|Full text PDF:||http://hdl.handle.net/2027.42/111399|
This dissertation presents contributions to fuel-efficient control of vehicle speed and constrained control with applications to aircraft. ???In the first part of this dissertation a stochastic approach to fuel-efficient vehicle speed control is developed. This approach encompasses stochastic modeling of road grade and traffic speed and uses the application of stochastic dynamic programming to generate vehicle speed control policies that are optimized for the trade-off between fuel consumption and travel time. It is shown that the policies lead to the emergence of time-varying vehicle speed patterns, often referred to as pulse and glide (PnG). Through simulations and experiments it is confirmed that these time-varying vehicle speed profiles are more fuel-efficient than driving at a comparable constant speed. A practical implementation strategy of these patterns is then developed and demonstrated. Also, several additional contributions are made to approaches for stochastic modeling of road grade and vehicle speed that include the use of Kullback-Liebler divergence and divergence rate and a stochastic jump-like model for the behavior of the road grade. In the second part of the dissertation, contributions to constrained control with applications to aircraft are described. Recoverable sets and integral safe sets of initial states of constrained closed-loop systems are introduced first and computational procedures of such sets based on linear discrete-time models are given. An approach to constrained flight planning based on chaining recoverable sets or integral safe sets is described and illustrated with a simulation example. Finally, two control schemes that exploit integral safe sets are proposed. The first scheme, referred to as the controller state governor (CSG), resets the controller state (typically an integrator) to enforce the constraints and enlarge the set of plant states that can be recovered without constraint violation. The second scheme, referred to as the controller state and reference governor (CSRG), combines the controller state governor with the reference governor control architecture and provides the capability of simultaneously modifying the reference command and the controller state to enforce the constraints. Theoretical results that characterize the response properties of both schemes are presented. Examples are reported that illustrate the operation of these schemes.