Robotics

Robotics requires continuous decision making under a short feedback loop from reality: decisions are real-world motions, immediately followed by a flow of sensory updates that must be processed in the nick of time before a new decision is required. Making sense of these sensory updates to achieve meaningful tasks is a key challenge, and a forefront where artificial intelligence is accelerating. From humanoid robots that could only walk on flat floors in the 2000’s, we have stepped within two decades to humanoid athletes that can perform backflips and enduring quadrupeds that hike mountains.

The two forefronts of the current research landscape in robotics, model-based and end-to-end architectures, are represented within the Prairie Institute. On the model-based front, we prioritize the development of advanced prediction capabilities based on mathematical modeling and dedicated optimization algorithms that leverage the mathematics of motion. On the end-to-end front, we design and study architectures that can integrate a wide spectrum of modalities, with a particular emphasis on the connection of natural language and vision to robotic motion.