DIAC a distributed actuator control

In behavior-based control systems, several processes need to access the actuators of the system. On the other hand, an actuator needs to be controlled through unambiguous commands. Therefore the outputs of several processes have to be fused to form an unambiguous command. In this paper, DIAC, a method to control actuators using distributed systems is proposed, where utility distributions are used to perform action selection. DIAC allows the combination of several mechanisms, each producing utility distributions to control the actuators according to a special task. The overall behavior of the system emerges from the fusion of these mechanisms by fusing the distributions to form a specific command. Each mechanism is assigned a special task and tries to fulfil this task as well as possible in cooperation, and competition, with the other mechanisms. DIAC takes into account that avoiding bad situations for one mechanism is more important than optimizing the result for many other tasks.