McMaster University
Intelligent Energy Management Strategy for Eco-driving in Connected and Autonomous Hybrid Electric Vehicles
Abstract
dc:description.abstractThis thesis focuses on developing an intelligent energy management strategy for eco-driving in Connected and Autonomous Hybrid Electric Vehicles (CA-HEV's), which can be implemented in real-time. The strategy is divided into two layers, i.e. the upper level controller and the lower level controller. The upper level controller can be executed on the remote server. It is responsible for extracting the information from the driver about the trip and the vehicle information using the communication capabilities of the CA-HEV. The gathered information is then utilized by dynamic programming (DP), which is implemented in a bi-layer fashion to reduce the computation burden on the server. The outer layer of the DP algorithm and the optimal velocity trajectory and the inner layer optimizes the power distribution in the powertrain to minimize fuel consumption alongside maintaining charge balance conditions. These global optimal results are evaluated for an ideal environment without any traffic information. The lower level controller is responsible for real-time implementation on vehicles in the real world environment and is based on a well-accredited reinforcement learning (RL) strategy, i.e., Q-learning. The RL-based controller optimally distributes the power in a CA-HEV and maintains charge balance conditions. Furthermore, the RL-based controller is also trained on the remote server based on global optimal results obtained from the DP algorithm. The optimal parameter information is then resent to the vehicle's embedded controller for real-time implementation. Simulations are performed for Toyata Prius (2010) on MATLAB and Simulink, and road information is gathered from SUMO. Simulation results provide a comparative study between the global optimal and the RL-based controller. To validate the adaptiveness of the RL-based controller, it is also tested on two approximate real-world drivecycles and its performance is compared against global optimal results evaluated using DP.
Degree
thesis:*- Department dc:contributor.department
- Mechanical Engineering
- Year dc:date.issued
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Rathore, Aashit
- Advisor dc:contributor.advisor
-
- Emadi, Ali
Subjects
dc:subject × 1Rights
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Identifier URI
- https://doi.org/10.71548/18484