Decoupling Behavior, Control, and Perception in Affordance-Based Manipulation

A novel mechanism is introduced by which a robot can connect the general notion of an affordance of an object to specific behaviors by which the robot can achieve the desired action. We achieve this by decomposing an affordance-drive behavior into three components. We first define controllers that specify how to achieve a desired change in object state through changes in the agent’s state. For each controller we develop at least one behavior primitive that determines how the controller outputs translate to specific movements of the agent. Additionally we provide at least one perceptual proxy that defines the representation of the object that is to be computed as input to the controller during execution. A variety of proxies may be selected for a given controller and a given proxy may provide input more than one controller. Decoupling these components allows the systematic exploration of a variety of strategies when evaluating the affordances of novel objects. We demonstrate the approach using a PR2 robot that executes different combinations of controller, behavior primitive, and proxy to perform a push positioning behavior on a selection of household objects.