Principal Investigator: Ronald S. Fearing
Compensation for camera motion on unsteady robots for optical flow – Improved quality of on-board camera information on unsteady, chaotic robots by developing solutions using sensor fusion, mechanical isolation, novel optics, and sensory adaptation. Used biologically inspired approaches to create low-complexity algorithms appropriate for our small robotic platforms.
Efficient locomotion over rough terrain – Implemented policy gradient reinforcement learning to search for efficient gaits of the VelociRoACH legged robot on three distinct terrains. Quantified the robot’s rotational dynamics to understand how they affect locomotion and exteroceptive sensing, including data from an on-board camera.
Designed a 1 gram board with integrated sensing and control electronics, which has evolved to become the de facto electronics platform of every mobile robot project in our lab for the past 4 years, enabling dozens of publications.
Set up and maintained public GitHub repositories for the lab’s embedded code projects and its PCBs. Championed code sharing while reducing fragmentation of the lab’s codebase, which prevented duplication of effort across projects.