Speaker
Description
Recent breakthroughs in artificial intelligence have greatly increased the capabilities of robots. These methods, however, require large amounts of memory, computation, and training data. These resources are not always available or accessible. In this talk I discuss three examples of alternative forms of robotic intelligence. In the first example, low-level intelligence is built into the physical structure of the robot while the high-level planning is handled by a human operator. In the second example, a robot dog learns to correct a physics-based control law using a linear policy. The final example shows a swarm of low-capability agents collectively estimating environmental quantities without any individual being required to maintain a global picture of the estimate or data.
