Mobile robotics maze solving using Wall following Algorithm
The goal of this project was to autonomously guide a robot through a maze course. I used a wall following algorithm to achieve this. I used an off-center kinematic control, which basically minimizes the positional and velocity error between an off center point and the center point of the robot, to compute joint velocities for the robot. I decided to use a lumibot differential drive mobile robot, adding proximity sensors on the left, right sides and front to get distance from the wall relative to the robot. To test the code, I started out with an easy maze. This maze course had gaps in between, and as a result, code generated errors whenever the proximity sensor distance data did not read float or integer values. Also, distance data were used in calculations, therefore nil values could not be used for running calculations. An easy fix was to substitute the max range of the sensor whenever the distance returned a nil value . However, this slightly affected the motion as you can tell from the video. The robot is programmed to keep at a defined distance from the wall throughout the course, therefore, whenever the proximity reads a higher or lower value, the robot adjusts to maintain this clearance (0.5 in this case)
Then I tried a more difficult maze, with 90 degree turns and curves. Also, fixed the jerky turns in the previous maze by breaking down different motion states (curve, turn, wall following, stop) and coded the kinematic equations in functions that return the joint velocities at the different states. This allowed for smoother state switching.
Below is a video with theoretical calculations of wall following and off-center kinematic control, using a differential drive by Leopold Armesto.