DESIGN OF INTELLIGENT CONTROL SYSTEM BASED ON PID CONTROLLER AND REINFORCEMENT LEARNING FOR ELECTRIC VEHICLE SPEED CONTROL
Electric vehicles are increasingly becoming the center of attention, considering that electrification in the world of transportation is currently being intensively campaigned as a solution to dealing with increasing Greenhouse Gases (GHGs) emissions. This makes the development of electric vehicles a...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Subjects: | |
| Online Access: | https://digilib.itb.ac.id/gdl/view/55285 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Electric vehicles are increasingly becoming the center of attention, considering that electrification in the world of transportation is currently being intensively campaigned as a solution to dealing with increasing Greenhouse Gases (GHGs) emissions. This makes the development of electric vehicles a hot topic among academics, including aspects of speed control because vehicle speed is related to energy consumption. Unfortunately, the controller that is widely used, namely PID, is only limited to tracking set-points, not to minimize the impact of disturbance or disturbance rejection. Even though in the operation of electric vehicles various disturbances may occur, such as internal disturbances, namely the slope of the track, and additional loads. Other disturbances can also be in the form of input and output disturbances on the system. For this reason, it is necessary to design a speed control system that can perform the tracking setpoint function as well as disturbance rejection optimally.
In this study a control system was developed which is a combination of PID controllers and Reinforcement Learning (RL). The PID controller is used for the setpoint tracking function by optimizing the accuracy, speed and response stability. Meanwhile, RL in addition to optimizing PID performance, also serves to minimize errors when disturbances occur. In this study, RL is also used to synthesize PID parameters. The RL method used is the Deep Deterministic Policy Gradient (DDPG) because this method has a continuous action and state.
The PID controller synthesized with RL was able to stabilize the system and reach an RMSE value of 9.296, better than the results of the Ziegler-Nichols synthesis of 9.569. In the RL PID controller for all types of disturbances, the RMSE value is smaller than that of the ordinary PID controller. With output noise, the PID controller's RMSE is RL 1.174, while the ordinary PID controller gets 1.274 |
|---|