Formation maintenance and control of mobile robots

This thesis investigates the problem of formation maintenance and control of mobile robots. The goal of this study was to develop a method that successfully allows mobile robots to successfully navigate in formation to a goal while avoiding obstacles. The method was based upon the leader-follower formation technique and is meant to be decentralized; all inputs to the follower are from passive observations only. There is no communication between the vehicles. For this investigation an overhead camera and a color recognition algorithm were used for localization of the robots, obstacles, and goal position. The two vehicles used were iRobot Creates and both were controlled by a central PC using a Bluetooth network. All software including the color recognition was written in Visual Basic. Since a leader-follower formation technique is used a proven hybrid potential field/fuzzy logic navigation system was used for the leader and for a secondary navigation system for the follower. This allowed the leader to navigate successfully to the goal location and the follower to navigate towards the desired formation position if not in position. The primary navigation system of the follower robot involved two separate aspects. First was the formation keeping method, for which a fuzzy logic method was chosen. Once the secondary navigation method brought the robot within a specified distance the formation keeping method was able to keep it in position. The position error was used as inputs two fuzzy logic controllers, one for steering and one for speed. The second focus was how to avoid obstacles. A potential field method was chosen to alter the leader-follower formation. The potential field method uses virtual repulsive forces away from obstacles and an attractive force toward a goal. This thesis used these forces to alter the desired follower angle for the leader-follower formation. The goal towards which the attractive forces act towards was chosen as the position which the follower would be in if there were no obstacles present. The findings of this thesis present many successful trials using this method. The experimental set up proved very reliable for testing and data acquisition. The camera and localization system were very accurate (+/- 2.5mm). The Creates were very simple to use and the Bluetooth network was very reliable. These trials show that the decentralized method devised can navigate a leader-follower formation to a goal while performing obstacle avoidance