Figure - uploaded by Jonathan Passmore
Content may be subject to copyright.
Source publication
This mixed‐methods study reviewed the role of coaching in the driver development environment. The study sought to explore the impact of coaching as a learning methodology and to compare this with an instruction‐based approach. The study involved a mixed‐methods sequential design. The first part of the study was a randomized controlled trial (RCT) a...
Context in source publication
Citations
Every year many people lose their lives due to driving accidents. Many factors affect driving accidents, including the quality of driving training. To improve the quality of driving training, a novel approach was introduced by Hemmati and Nahvi in 2019 based on a master-slave teleoperation method. However, that method did not guarantee stability.
In this thesis, the stability of the driver training teleoperation system was analyzed under friction disturbance and delay. Stability enhancement was achieved by designing a model predictive controller instead of PID controller. The driving model contains two parts: the vehicle model and the driver model. First, a new driver model describing the driver’s behavior was developed. The new model consists of a neuro-muscular subsystem, perception subsystem, and vestibular subsystem. The driver’s reaction time and prediction time are the main variables of the driver model. Two models were designed for the vehicle model. The first one models a 4-wheel sedan and is implemented in CarSim, and the second one contains an observer’s model that uses a bicycle model. Next, the brushless DC motors of the master-slave system were modeled. Then, a model predictive controller was designed to improve the stability of the system in the presence of friction and sensor and communication delays. The system response resulting from the two controllers of the PID and the model predictive control were compared. The effects of delay in the PID controller were studied based on the Lyapunov-Krasovskii method using linear inequality matrices. Also, the stability of the model predictive control system was studied by evaluating a cost function. To enhance the performance of the model predictive controller, equations for the actuator dynamics were developed. The model was further augmented by taking into account the friction model. The model predictive control has a high computational cost that may decrease the stability margin or even make the system unstable at high-frequency inputs. To overcome this problem, offline optimal and sub-optimal optimization methods were used, including explicit MPC, simplified explicit MPC, and sub-optimal explicit MPC. The results were verified with Matlab and CarSim for the double lane-change scenario. The developed model achieved better performance compared with the PID method for the two inputs of impulse and high-frequency harmonic with constant and time-variable delays. The advantages of MPC over PID were observed in three parts: less overshoot, less settling time, and more stability margin. The PID controller was unstable at high-frequency inputs with variable time delay, unlike MPC. The stability of the PID controller was examined by the Lyapunov-Krasovskii method, while the cost function was used for the MPC. It was shown that the stability of the system decreases as soon as the delay rises, but the MPC had a more robust response to the rise of delay, so it was decided to use MPC instead of PID for the double lane-change scenario.
The experimental results showed response improvements for the proposed controller. At the stall of the DC motor, the current rises due to the static friction force. After overcoming the static friction force, the current suddenly drops due to a lower Coulomb friction force. The current drop is accurately observed in the developed model. The developed models resulted in 1.9 and 6.7 times more accurate responses at counterclockwise and clockwise directions, respectively. It can be seen that the simplified explicit MPC has 26.5% lower computational time over explicit MPC and has 58% higher robustness against uncertainty with respect to sub-optimal explicit MPC. Unlike PID and the primal-dual method, which were unstable at high-frequency inputs, the simplified explicit MPC is stable. Better performance of the simplified explicit MPC was obtained when comparing the control method’s robustness and computational cost.
