The Toyota Research Institute (TRI) has announced its next phase of its collaborative research program, selecting 13 research institutions to receive more than $75 million over the course of the next five years. As one of these institutions, Penn Engineering is home to two researchers who will conduct projects related to robotics, computer vision and safety under this program: Michael Posa, Assistant Professor in the Department of Mechanical Engineering and Applied Mechanics, and Jianbo Shi, Professor in the Department of Computer and Information Science.
Posa’s project aims to enable control of complex robotic motions, whether that be in-hand manipulation or legged locomotion over rough terrain. By algorithmically uncovering simplified models that distill the key features of the high-dimensional, multi-contact dynamics, robots will be able to make real-time decisions on how to move and react to their changing environments. This project will leverage and contribute to the Drake open-source software library. Posa’s lab employs a bipedal robot, known as Cassie, to improve the walking skills that toddlers naturally master but robots continue to struggle with.
Shi’s project, which will be conducted in collaboration with Professor Hyun Soo Park of the University of Minnesota, Twin Cities, aims to use computer vision algorithms to better understand what a car’s driver and passengers are doing. Gathering data with multiple cameras positioned within a car’s cabin, machine learning techniques can be used to predict when a driver is distracted or falling asleep, or learn how and where passengers are sitting in relation to the car’s seatbelts and airbags. By modeling the interior of a car and its occupants from multiple angles and lighting conditions, the computer systems that govern a car’s safety features would be better prepared to respond in a wide range of situations.
Posa and Shi are both members of Penn Engineering’s GRASP Lab, which includes around two dozen professors who are increasingly developing robotics research with industrial ties.