Professor Odest Chadwicke Jenkins
Odest Chadwicke Jenkins, Ph.D. is
Assistant Professor of Computer Science at Brown
Chad’s research group, Robotics, Learning and Autonomy at Brown, explores topics related to human-robot interaction and robot learning, with a specific focus on robot learning from human demonstration. His research strives towards realizing robots and autonomous systems as effective collaborators for humans to pursue their endeavors.
His research into “robot learning from demonstration”, or robot LfD, centers on the automated discovery of processes underlying human movement and decision making. In recent years, robot LfD has emerged as a compelling alternative, where robots are programmed implicitly from a user’s demonstration rather than explicitly through an intermediate form (e.g., hardcoded program) or task-unrelated secondary skills (e.g., computer programming). The role of learning, in this case, is the estimation of a human’s intended control policy or movement process from demonstrated examples.
Towards this goal, Chad’s work has contributed methods in manifold learning, a form of nonparametric dimension reduction, for uncovering dynamical processes and underlying structure in nonlinear time-series data. These methods have been applied to human motion for learning motion primitives, or predictive dynamical motion priors. He has learned and applied these primitives in several domains, including humanoid robot control, vision-based human tracking, and sparse user control of prosthetic devices.
More recently, his group has taken this work in a new direction for learning robot controllers from human demonstration through algorithms and models for nonparametric regression with infinite mixtures of experts. His goal for this work is to extend robot LfD beyond applicability to constrained scenarios and towards the ability to learn any finite state automata from users. As such, he aims to elevate robot LfD to be on par or better than with manual coding for developing robot controllers.
Chad also addresses research problems in robot/computer perception, humanoid robotics, machine learning, autonomous control, dexterous manipulation, computer animation, and game development.
He coauthored Creating Games: Mechanics, Content, and Technology. His papers include Automated Derivation of Primitives for Movement Classification, Deriving Action and Behavior Primitives from Human Motion Data, A Spatio-temporal Extension to Isomap Nonlinear Dimension Reduction, Automated Derivation of Behavior Vocabularies for Autonomous Humanoid Motion, Performance-Derived Behavior Vocabularies: Data-Driven Acquisition of Skills From Motion, and Dynamo: Dynamic, Data-driven Character Control with Adjustable Balance. Read the full list of his publications!
Chad earned his B.S. (cum laude) in Computer Science and Mathematics at Alma College in 1996. He earned his M.S. in Computer Science at the Georgia Institute of Technology in 1998 and his Ph.D. in Computer Science at the University of Southern California in 2003 with the dissertation “Data-driven Derivation of Skills for Autonomous Humanoid Agents”.