Fusing multiple sensors through behaviors with the distributed architecture

This paper describes a new action selection mechanism that allows the robot controller to dynamically determine which goals (tasks) are most applicable in the current state of the environment. This allows the robotic controller to resolve conflicting goals that arise, allow multiple non-conflicting goals to run simultaneously, and dynamically switch to a new goal if the target for the goal has been found in the environment. This is accomplished by using a behavior based paradigm and allowing each behavior to calculate its own activation level (run level). The behaviors then use the activation level to compete for control in a new distributed control architecture to allow execution of multiple (and sometimes conflicting) goals on a robotic system.