Design and Implementation of Communication System of Mobile Robot and FPGA for Smart Navigation Problem

Imam Firdaus · · 2 minute read
ML RL Learning FPGA SoC TCP/IP WiFi

ABSTRACT

🔗

Design and Implementation of Communication System of Mobile Robot and FPGA for Smart Navigation Problem

🔗

By

🔗

Mohamad Imam Firdaus

🔗

NIM: 13218025

🔗

(Undergraduate Program in Electrical Engineering)

🔗

Q-Learning is one of the commonly used Reinforcement Learning methods. To accelerate the performance of Q-Learning, a Q-Learning acceleration hardware based on System on Chip was created. The system is implemented on the FPGA development board PYNQ-Z1. To carry out the testing and implementation process in real cases, a Mobile Robot is made that can be integrated with the FPGA development board. The Mobile Robot will run on Grid World which is compatible with the Q-Learning Accelerator work process to be tested. In the test system, it is important for the robot to get input that matches the hardware accelerator needs. The robot must also get the output issued from the Hardware Accelerator. In order to transmit environmental conditions and receive action from the Hardware Accelerator, the Robot must have a communication interface compatible with the FPGA Development Board and the robot components themselves. On the FPGA Development Board there are several communication protocols that need to be met. These protocols include network protocols such as TCP / IP and USB serial. TCP/IP is needed for controlling FPGA performance by testers. This TCP/IP will support applications such as Jupyter Notebook and SSH. USB serial is useful for communication between the robot and the FPGA.

To meet the needs mentioned above, a docking system is designed on the robot. The docking system will support mentioned connectivity and add power delivery and mechanical mounting. To meet communication needs, there are 2 components. The first component is the Ralink RT5370 WiFi dongle. This WiFi dongle was chosen because it uses a USB port for connection and is also supported by the Linux Kernel without the need to install additional drivers. Next is the USB Hub to increase the number of USB connections that can be made. This is because the robot requires 2 USB connections, and the FPGA development board only has one port. To meet the needs of power delivery, the docking system will contain a USB Powerbank that provides 5V power to the FPGA. For self-mounting, the robot chassis will be specially designed to accommodate the docking system at the top of the robot.

Keywords: Q-Learning, FPGA, SoC, TCP/IP, WiFi