Chad Jenkins is primarily interested in the development of methods for autonomous control and perception through leveraging human performance from the real world. His work furthers the idea that robot control and computational perception are better learned from human demonstration rather than explicit computer programming.
Prof. Jenkins' work strives to address three basic questions. First, how can we capture data from the world that is representative of human performance? Second, how can machine learning and data analysis be used to extract dynamical structure from performance data? Lastly, how can we utilize learned dynamics for building autonomous robot controllers and perception mechanisms?
His previous efforts were mostly geared towards humanoid robotics with respect to learning primitive behaviors for robot control through imitation. More generally, he addresses perception, control, and learning issues at the intersection of robotics, computer vision, computer animation, machine learning and interactive systems.