SCS Undergraduate Thesis Topics
|Dilan Edirisinghe||Dave Touretzky||Perception of Affordances by a Mobile Robot|
In this project, we present a novel approach to robot mobile programming using the idea of affordances. Coined by the psychologist J. J. Gibson, affordances are descriptions of relationships between the robot, the object, and the environment they share that determine sets of possible actions that can be performed within that environment. In this framework, which is implemented using the Tekkotsu framework on the Sony AIBO dog, objects such as balls and lines are represented as primitives that can be seen, touched and manipulated with. The manipulation of the object then becomes a straightforward application of affordances. For example, a ball affords different pushing schemes based on its size: a large ball affords pushing using the dog.s chest, while a small ball affords pushing using the dog.s paw. Internally, each affordance is implemented using the following components: a finite state machine, an achievability code, and a return code. An achievability code informs the programmer how realizable an affordance.s action is prior to the programmer executing the action, while the return code informs the programmer how well the affordance perfomed its action. A set of affordances is generated by an affordance recognizer at each time step. The programmer only has to select which affordance to execute, alleviating him of the low-level details such as coordinate systems and mechanical limits associated with the robot.