PID Q-LEARNING CONTROL SYSTEM FOR SPEED OF ELECTRIC VEHICLE

• Main Author: Handy Sugianto, Kriesnawan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/50389
• Institution: Institut Teknologi Bandung

Switched Reluctance Motor are a solutions to problems related to motor driving in electric vehicles that have a fairly complex construction. The SRM motor has a very simple construction consisting only of a rotor and stator without any permanent magnets inside. It is necessary to implement a controller on the SRM motor in order to obtain a system response that is more stable and has high performance against changes in the given input rate. In this study, a combined PID control algorithm with Reinforcemenet Learning (QPID / PID Q-Learning) was developed which compensates for errors in system responses due to nonlinear conditions and changes in reference values and can also stabilize the system. The compensation generated by the Reinforcemenet Learning feature is through making error decisions from the system response. In addition, the Ziegler-Nichols PID controller is also applied to the system as a system for QPID performance in reducing the area of the electric vehicle system error. The main objective of the QPID controller is to work in conditions where conventional controllers such as the Ziegler-Nichols PID controller fails, namely in conditions of high uncertainty and nonlinearity. The QPID controller provides performance that is able to make the system stable and has the smallest RMSE value compared to the PID, which is 0.3156 and 0.5080 for the Ziegler-Nichols PID at 10 km/h on horizontal roads. The RMSE value is also the smallest for QPID when there is a change in the reference value from 30 km/h to 40 km/h, namely 0.9036.