MSE 298 Seminar: A Memristor-Based Hybrid Analog-Digital Cerebellum (Small Brain) for Mobile Robotics

McDonnell Douglas Engineering Auditorium (MDEA)
Wei Wu, Ph.D.

Associate Professor 
Ming Hsieh Department of Electrical and Computer Engineering 
University of Southern California

Abstract: Algorithms for mobile robotic systems are generally implemented on purely digital computing platforms. As complementary metal-oxide-semiconductor (CMOS) scaling approaches the end of the road, the improvement of the throughput of digital processors and computing power efficiency is nearing its end. This issue not only affects the power requirements of large data centers but also limits the performance of mobile robotic systems with perception and actuation. Developing alternative computational platforms may lead to more energy-efficient and responsive mobile robotics. Inspired by how human and animal brains work, we report a hybrid analog-digital computing platform enabled by memristors on a mobile inverted pendulum robot. The “cerebellum” (sensor fusion + motion control) of this mobile robotic system is implemented in memristor-based analog circuits and the rest of the system is implemented in digital circuits. Such a platform can perform computation in the analog domain and thus removes the speed and energy efficiency bottleneck. Using a model-free optimization method, the mobile robotic system can tune the conductance states of memristors adaptively to achieve optimal control performance. The robot using the hybrid analog-digital platform demonstrated more than one order of magnitude enhancement of speed and energy efficiency over the traditional digital platform. A demo video ( also shows that after an impact, the robot using our hybrid analog-digital platform recovers balance much faster than the same robot with the traditional digital platform. The technology can also be used in more sophisticated robotic systems. The implementation of this technology on micro air vehicles (MAV) will be presented too.

Bio: Wei Wu graduated from Peking University with a bachelor's degree in physics in 1996 and received a doctorate in electrical engineering from Princeton University in 2003. He is an associate professor at the Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California. Before joining USC in 2012, he had worked as a research associate, scientist and senior scientist at HP labs. His work includes nano imprint lithography and applications in nano-electronics, nano-photonics, plasmonics, chemical sensing and nano-electrochemical cells. He co-authored 125 peer-reviewed scientific journal papers with 11,447 citations, two book chapters and more than 150 conference presentations including 16 keynote and invited presentations. He has 118 granted U.S. patents. Half of them were also filed internationally. His H-index is 50. He is a co-editor of Applied Physics A, an associate editor of IEEE Transactionson Nanotechnology and a regional editor (North America) of Nanomanufacturing and Metrology. He was also an IEEE Nanotechnology Council 2015 and 2016 distinguished lecturer and a recipient of the USC Stevens Center for Innovation Commercialization Award 2020.