Control of Robust Micro-Robots in Uncertain Environments

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

Autonomous micro-robots a few millimeters in size have transformative potential for applications ranging from infrastructure monitoring to disaster response to exploration of hostile or inaccessible locales. Such tasks will require robots that can successfully navigate complicated and unpredictable terrain and environments. Given limitations on moving parts in micro-scale systems, legged locomotion offers the best combination of adaptability and efficiency. Standard features of full-scale walking robots, such as motorized and instrumented knee and ankle joints, are impossible or impractical to realize with current micro-fabrication techniques, making flexible linkages the most promising approach to articulated limbs. But, because of differences between the ways that leg stiffness and body mass scale as robot dimensions decrease, the natural springiness of the legs cannot be relied on to keep all feet in contact with uneven ground, which can greatly degrade gait effectiveness. These challenges make legged locomotion an essentially different and more difficult problem at the micro-scale. This project will advance the fundamental state of the art towards high-speed, high-mobility legged micro-robots navigating in an uncertain environment. It will develop models of the leg-surface interaction, and use these models to derive efficient and effective walking gaits. To maximize adaptability in changing conditions, internal sensors capable of monitoring gait effectiveness will be added to existing micro-robot leg designs. This project will study the complex contact interaction of walking micro-robots with multiple feet moving at high-speeds on varying surfaces. The resulting dynamic models will be used to identify walking gaits that produce efficient forward motion over a range of environmental conditions. The investigators have previously combined thin-film lead-zirconate-titanate (PZT) with resilient polymer materials to demonstrate micro-fabricated legs with an unusually large range of motion and actuation speed. In this project, sensing elements will be integrated into the thin-film PZT/polymer legs to allow basic estimation of robot motion. Actuation and estimator design will meet constraints imposed by small-scale power sources, through input sequences based on a limited number of switched input voltages and leg and robot movement estimates based on limited sensor samples. Once estimation of robot walking effectiveness has been demonstrated using on-chip measurements, adaptive control algorithms will be tailored to respond to environmental conditions that change over time.