Large-Scale Flexible Actuation and Sensing

Abstract

Humans have many more sensors and actuators within their bodies than artificial systems to help them learn from the environment and complete complex tasks. This thesis develops platforms for large-scale sensing and actuation systems. The many interfaces (wires) from sensors transmitting data to the Si-COMS processor in large-scale sensing systems lead to integration challenges. On the other hand, large-scale soft actuation systems demonstrated by piezoelectric soft robots have three main challenges: (1) soft body static and dynamic modeling; (2) advanced motion mechanisms; and (3) powering and control electronics integration for untethered operation. This thesis tackles the above challenges by developing research platforms that include models, algorithms, and prototypes. For large-scale actuation demonstration, this thesis uses soft robots containing five piezoelectric actuators to demonstrate various motions: bio-inspired inchworm motion, in-place jumping, and forward/backward jumping. This thesis develops models for our soft robots to tackle the modeling challenge. They include: (1) an analytical static model for the robot's shape, contact, and friction; (2) a simulation framework for dynamics; (3) an analytical dynamic model for understanding the jumping motion. Further is a model-based shape controller for target shape match and crawling under various tight overhead barriers. The models and control algorithms are validated by experiments in varying cases with good agreement. For electronics integration, this thesis develops an untethered soft robot with embedded high-voltage circuity powered by batteries and controlled through a wireless link. The robot exhibits a range of controllable motions, including bidirectional crawling, turning, and in-place rotation. It can jump 20x high of its own thickness, and its speed can reach 6 cm/s with specific payload distribution. To tackle the interfacing challenge in large-scale sensing, this thesis develops an architecture that exploits sensing signal sparsity exhibited in various applications. The architecture implements compressed sensing via thin-film-transistor (TFT) switches and is demonstrated in a force-sensing system. The number of interfaces is reduced to the sparsity level instead of the number of sensors. This prototype needs 161 TFTs per sensor. TFT yield is analyzed and improved by preventing process-induced gate dielectrics breakdown with layout modifications and temporary shorting of connections on key components.