This Faculty Early Career Development (CAREER) grant will create new ways to understand and control robots that walk, jump, or grasp. Specifically, the project will study robots that repeatedly make and break contact with the ground or some other surface. Planning and controlling motion involving contact is particularly challenging, because very small changes in, for example, the friction between the two surfaces or the points at which the contact occurs, may cause large changes in the robot's overall motion. For example, small changes in where a robot hand touches a tool may cause the tool to slip from the robot's grasp. Or small differences in where a robot places its foot on rocky ground may cause the robot to trip and fall. Robots are most successful when their workspace is carefully monitored and maintained. For robots to be equally successful in unstructured and unpredictable real-world settings requires new motion planning and control methods, such as this project will develop. These new capabilities promise benefits in applications ranging from home healthcare, to environmental monitoring, to manufacturing. The project also features robot activities for K-12, undergraduate, and graduate education.
This project will model and analyze the structure of systems that have uncertainty in their contact conditions. In particular, this work considers both local stability properties around uncertain contact timing and event sequences, as well as the global topological structure of the system. This project will discover properties of the dynamics and contact conditions that predict successful behaviors by extending notions of contraction or convergence to discontinuous contact systems as well as finding persistent topological features. The second part of this project will then use these properties to find new control generation strategies that are robust to uncertainty in the timing of planned contact events as well as anticipate and possibly avoid unplanned contact events. While many ad hoc solutions have been researched for different settings, this research will unify and extend these point solutions by leveraging these common structural properties. The result will be a set of methods for generating controllers that avoid problematic behavior when contact modes change unexpectedly, for example when feet scuff or slip on a surface, but still allow for rich changing contact conditions.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.