Our group has two symbiotic research goals.
- Apply engineering and biological topics to build and control walking robots
- Use the resulting robots to better understand how animals control their locomotion.
Developing robots that can walk will enable humans to explore and traverse extreme terrains everywhere, including farms and orchards, mines, underwater on the sea floor, and on other planets. Working with neuroscientists to better understand
Our research touches three fields.
Bioinspired robotics is heavily interdisciplinary, drawing from robotics, computational neuroscience, and neurobiology.
Robot mechanics and sensing
Many robots have centralized controllers that infer the robot's state with as little sensory feedback as possible. However, animals control behaviors using a large amount of redundant sensory feedback. We are investigating strategies for integrating multiple sensory systems to produce one coordinated behavior. Animals simplify control tasks by relying on their passive mechanics to solve some problems for them. We engineer our robots with animal-like mechanics and sensory dynamics to achieve similar benefits.
Our computational neuroscience modeling framework is called
Synthetic Nervous Systems (SNS). With SNS, we design and implement dynamic
neural controllers for real-time, closed-loop robot operation. To achieve this
goal, we developed the
Functional Subnetwork Approach (FSA), which enables us to design nonspiking,
spiking, and oscillatory networks without machine learning or optimization.
Robotic modeling of animal systems
Dynamic scaling is a means by which to model the mechanics of an animal with a robot of a different size. The motions of differently-sized animals are dominated by different forces. The same constraints apply to robots, meaning that animal-like robots must be designed with the mechanics of the model animal in mind.