IMPLEMENTATION OF ROBOT MOVEMENT IN SHORTEST PATH PLANNING TESTING PLATFORM WITH HARDWARE

• Main Author: Yugansyah, Cahya
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/66445
• Institution: Institut Teknologi Bandung

Machine-learning (ML) is a data analysis method that automates the construction of analytical models. There are several common ML methods to use. Reinforcement Learning (RL) is one of the most used types of ML and finds the action that produces the greatest reward through trial and error. Computing requirements for heavy RL require a powerful computing device, especially if there are problems with tight time constraints and/or with large data processing, especially in the problem of finding the shortest path. Optimization of the hardware can reduce the large power consumption. This optimization requires a lot of research and design to find the most efficient computing hardware. One of the methods that is used is by implementing hardware accelerator using Field Programmable Gate Array (FPGA). The implementation results from the hardware implemented needs to be tested for certain applications to measure the performance of the device. One of the good applications for implementing RL is smart navigation system. In this book, author will explain about the design of several subsystems for shortest path planning with Q-Learning algorithm testing platform with hardware. The subsystem in question is the robot and its movement on the testing platform with the result that the robot can move in 1,265 seconds to change state, 0.771 seconds to change orientation, and 0.721 seconds to read RFID.