Obstacle-avoiding robot

Abstract

An autonomous mobile robot can navigate around an unseen space without any human intervention with the help of obstacle avoidance techniques. Such a robot will interpret information acquired by its sensors to detect obstacles in its way and navigate around them. Typical algorithms for this include Bug, Artificial Potential Field, Dynamic Window and Vector Field Histogram. In this project, we built a wheeled differential drive robot using a Lego Mindstorms EV3 Kit. Since the robot is required to avoid obstacles without colliding into them, we initially equipped the robot with an ultrasonic and an infrared sensor at its front to measure the distance from any obstacles. Tests to find the distance error and variance for both sensors showed that the ultrasonic sensor is much more reliable than the infrared and hence, we only used the ultrasonic sensor for our implementation. We investigated the Bug2 algorithm for obstacle avoidance. Here, the robot follows a goal line representing the target path and whenever it detects an obstacle, it follows the obstacle’s perimeter until it encounters the goal line to follow once again. This process is repeated until the robot reaches the desired target location. We then investigated the Artificial Potential Field algorithm. This involves attractive and repulsive fields in order to plan a path from start to goal. The attractive field is used to attract the robot towards the goal while the repulsive fields repel the robot away from obstacles. We tested both algorithms using a 10 x 10 grid as the robot’s environment. The grid included a goal location and different obstacles were randomly placed in the grid in order to interfere with the path which the robot was planning to take. In each of the ten repeated tests for each of the eight environments, the robot reached its goal without ever colliding with obstacles. However, although Bug2 presents a better way of avoiding obstacles since it detects them in real time, Artificial Potential Field required the robot to traverse less distance than Bug2 to avoid obstacles. The results showed that on average, Bug2 travelled 123.62 ± 52.88% extra distance with original start and goal and 77.06 ± 26.88% extra distance with them inverted. On the other hand, APF travelled 43.78 ± 10.22 % extra distance with original start and goal and 75.78 ± 21.99 % extra distance with them inverted. Therefore, for the original environment, APF travelled an average of 79.84 % extra distance less than Bug2 and for the inverted environment it travelled an average of 1.28 % extra distance less.