Start Date
10-5-2019 10:15 AM
End Date
10-5-2019 11:30 AM
Document Type
Event
Keywords
Line following robot, Pathfinding, PID control, reinforcement learning
Description
The development of a line-following robot is, for many, a building block into further applications of robotics. One of the main focuses is pathfinding without prior knowledge of the route or environment. In the case of a simple line- following robots, the task is limited, allowing for much room in experimenting with different approaches. In this paper, we present our dual-control line following robot - ATR LineBot. The first control mechanism uses two series of line sensors in combination with a custom PID controller for speed optimization. The additional sensors allows for sophisticated planning and correction to be done. Our second control mechanism utilizes a double layered neural network for remote learning of optimal speed. Utilizing the output of line detection sensors and motor encoder as feature vectors, the neural network can optimize control. ATR LineBot, the dual-control line following robot, uses both PID controlling and reinforcement learning for successful navigation and serves as a platform for improving the traditional building block of robotics. Thus, we present a smart line following robot comprising a robust design, which is compatible with any circumstance as well as efficient in achieving its ideal speed to perceive its goal.
DOI
https://doi.org/10.5038/KRZG8576
Design Evaluation of Self-Learning Line Following Robot
The development of a line-following robot is, for many, a building block into further applications of robotics. One of the main focuses is pathfinding without prior knowledge of the route or environment. In the case of a simple line- following robots, the task is limited, allowing for much room in experimenting with different approaches. In this paper, we present our dual-control line following robot - ATR LineBot. The first control mechanism uses two series of line sensors in combination with a custom PID controller for speed optimization. The additional sensors allows for sophisticated planning and correction to be done. Our second control mechanism utilizes a double layered neural network for remote learning of optimal speed. Utilizing the output of line detection sensors and motor encoder as feature vectors, the neural network can optimize control. ATR LineBot, the dual-control line following robot, uses both PID controlling and reinforcement learning for successful navigation and serves as a platform for improving the traditional building block of robotics. Thus, we present a smart line following robot comprising a robust design, which is compatible with any circumstance as well as efficient in achieving its ideal speed to perceive its goal.