The perceiving robot: What does it see? What does it do?

The nature of robots in the future is examined, and it is proposed that they should fundamentally function as responsible agents for people and not merely as programmed artifacts. Thus besides having extended powers of perception, they will need to deal with their own purposes, embedded in purpose structures, and with the ways of modifying and optimizing their purposes in parallel. The primary purpose of robotic perception is to see how well the robot is performing on a current task (or subtask). Being responsible means being able to take care of more than one task at once: for example, a cleaning robot must be able not only to clean, but to do it efficiently and to conserve power and itself; it must be able to modify what it tries to do in order to take care of new requirements. Thus the robot must be able to modify its own understanding of what clean means in various circumstances. This viewpoint of robotics extends the current approaches in a new direction