UC Merced

An autonomous vehicle must navigate dynamic and complex environments, requiring advanced perception and control systems to ensure safety and efficiency. In general, the technologies used in self-driving cars are divided into perception and control. Control is the "brain" of the vehicle, planning its route and making decisions about how to move. Perception acts as the "eyes," interpreting the environment and feeding crucial information to the control system. The control algorithm used in this study is Model Predictive Control (MPC) which is known for its real time optimization-based approach and capability to have multiple inputs and outputs. MPC is used to plan feasible routes while respecting the vehicle’s physical constraints and dynamically changing environments, such as varying road conditions, road curvature, obstacles, and traffic. To optimize the MPC capabilities for different cases, multiple driving scenarios are considered. Initially, we designed an MPC that ensures the feasibility and stability of the vehicle that needs to come to a stop at a traffic light over a short period of time, and at different road frictions. Then we developed an MPC for a multiple-turn trajectory, where the other controllers fail to stably track the trajectory. Finally, we advance the MPC controller for a multiple vehicle platooning, ensuring safe and efficient coordination among vehicles by maintaining appropriate inter-vehicle distances and avoiding collisions. In summary, by addressing key challenges such as multi-vehicle coordination, traffic signal management, and obstacle avoidance, this thesis demonstrates the versatility and robustness of MPC in managing diverse autonomous driving scenarios.