Producing Complex Behaviours on Trajectory Velocity Learning Mobile Robots

This paper describes a novel and effective method for devising complex behaviours on a mobile robot that is considerably easier to implement than most existing robot control methods. Instead of designing multiple behaviours and a subsumption layer for selecting behaviours, this method provides the robot with only one behaviour (goal seeking) and produces more complex behaviours by altering the robot's perception of the environment so that significant objects appear to have different shapes. For example, a robot playing robot soccer can be made to always take shots at the goal by placing an illusionary object around the ball which has an open end away from the goal. Thus, by being attracted to the ball and repelled by the sides of the illusionary object the robot is directed onto the ball in the direction of the goal. This effectively causes the robot to always line up and strike the ball toward the enemy goal. Experimental, results of this goal shooting behaviour are provided. These results demonstrate both the simplicity and effectiveness of this approach at controlling a robot to perform a task that would be considerably more difficult to implement with other methods.