My current research focuses on the many challenging problems surrounding control of legged robots, from sensing and low-level control to high-level behavior development.
Legged Locomotion: Novel designs have allowed today's legged robots to rely upon "programmed" mechanical properties to ensure dynamic stability, increasing their performance over previous legged robots, while keeping computation and energy consumption in line with wheeled robots. I am researching methods to actively modify robot gaits, preprogrammed leg motions with little or no sensor feedback, to change the dynamic behavior of a moving robot, while remaining within the domain of dynamic stability, dictated by the mechanism.
Sensor-Based Behaviors: Within this framework for altering robot gaits, I am working to allow our legged robots to sense the ground they are walking or running over, actively modifying the gait in order to progress through complex environments, such as transitioning amongst different surfaces or navigating over and through geometric obstacles. Using a minimal amount of sensor data, these control policies can greatly increase performance without adversely affecting computation, unlike traditional legged walkers which deliberately choose each footfall.
Development Tools and Software Frameworks: Rapid development of robots and behaviors requires tools perfectly suited for each given task. I prefer to take a bottom-up approach to software development, noticing patterns that emerge throughout development and incorporating these commonalities into software tools for the user. This is opposed to a top-down approach, where an all-encompassing framework, as designed early on, tends to limit the perspicacity and freedom of the robot developer to "do what they want". As well, reducing the scope of each tool reduces the learning curve and speeds adoption amongst users, aiding rapid development.
|Research Interest Keywords|
|3-D perception, active perception, architectures, computer graphics, control, data visualization, field robotics, legged locomotion, manipulation, mobile robots, motion planning, obstacle avoidance, sensor fusion, teleoperation|
|The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.|
Contact Us | Update Instructions