Perturbed Optimal Control for Connected and Automated Vehicles

Name: Shobhit Gupta

Abstract:
Global regulatory targets for reducing CO2 emissions along with the customer demand is driving the automotive sector towards energy efficient transportation. Powertrain electrification offers great potential to improve the fuel economy due to the extra control flexibility compared to vehicles with a single power source. The benefits of the electrification can be significantly reduced when auxiliaries such as the vehicle climate control system directly competes with the powertrain for battery energy, reducing the range and energy efficiency. Connected and Automated Vehicles (CAVs) can increase the energy savings by allowing to switch from instantaneous optimization to predictive optimization by leveraging information from advanced navigation systems, Vehicle-to-Infrastructure (V2I) and Vehicle-to-Vehicle (V2V) communication. In this work, two energy optimization problems for CAVs are studied. First is to jointly optimize the vehicle and powertrain dynamics and the second is to optimize the vehicle climate control system. The focus of this work is to combine the Dynamic Programming (DP), Approximate Dynamic Programming (ADP) and perturbation theory based approaches to solve the energy optimization problems with variations in external inputs and parameters. The uncertainty in or variability of some key parameters can have a significant impact on the plant model, objective function or constraints, ultimately affecting the fuel economy and travel time. To this end, two computationally efficient sub-optimal DP-based methodologies are proposed that compensates for these variations by formulating the problem as a perturbed optimal control problem. The first approach develops a systematic framework to evaluate the sensitivity of the value function to parameters based on the perturbation theory. The sensitivity expression is then used to devise a methodology to account for the mismatches in the observed parameters by correcting the terminal cost of a Model Predictive Control (MPC). Two case-studies are considered with variations in vehicle payload and auxiliary power load. The second approach develops an algorithm for solving perturbed dynamic programs, where first-order corrections are applied to a pre-computed optimal strategy and its corresponding value function. The technique is developed to handle perturbations of external inputs and parameters affecting system dynamics, objective and constraint functions, allowing the application to a wide variety of perturbed problems. The method is applied to the energy optimization of a vehicle climate control system, formulated as a constrained dynamic program where the external conditions and target set-point are perturbed from their nominal values. Finally, the developed algorithm is used to solve the vehicle climate control optimization problem as a MPC. Simulations result show 10-15% energy savings from a baseline strategy, and 80% reduction in computation time from a DP solution, with only minimal effect on the overall controller performance.

Zoom Link (or alternative) - if available
https://osu.zoom.us/j/3329468109?pwd=RlZQNEtxZDgvRWFHajlQYWxuY2VZQT09 (Meeting ID: 332 946 8109, Password: 019377)

Committee Members
Dr. Marcello Canova
Dr. Stephanie Stockar
Dr. Abhishek Gupta