Abstract:              Fully autonomous vehicles, equipped with the ability to perceive a dynamic environment and act intelligently within it, is a moonshot goal for the field of robotics. The disciplines of perception and planning are advancing at breakneck speeds, but there's been a bifurcation in development: perception is built upon statistical models for the environment thereby providing a measurement of uncertainty in generated representations, while planning is treated deterministically, often as an offline process. The real world however is inherently uncertain and these approaches in planning and control inevitably fail as a result. In this talk, I will discuss algorithms under development in my lab that are addressing these critical challenges to autonomy: visual-inertial perception with online statistical change detection, and adaptive simulation-in-the-loop model predictive control (MPC). Visual-inertial perception with change detection is a technique we have developed to both self-calibrate and navigate in unknown environments with a variety of sensors acting on a mobile platform. These methods rely on determining when calibrations have become stale and initiates a new self-calibration when they do. Nonlinear MPC on the other hand focuses on how to infer optimal control inputs of a high-dimensional, sophisticated system, and I will show how these methods apply to autonomous vehicles. Included in this work is the development of fast, accurate and reliable perception and planning algorithms for operating small-scale cars at high speeds.

Biography:          Chris Heckman is an Assistant Professor in the Department of Computer Science at the University of Colorado at Boulder. Professor Heckman earned his BS in Mechanical Engineering from UC Berkeley in 2008 and his PhD in Theoretical and Applied Mechanics from Cornell University in 2012, where he was an NSF Graduate Research Fellow. He had postdoctoral appointments at the Naval Research Laboratory in Washington, DC as an NRC Research Associate, and in the Autonomous Robotics and Perception Group at CU Boulder as a Research Scientist, before joining the faculty there in 2016.

His research focuses on developing mathematical and systems-level frameworks for autonomous control and perception. His work applies concepts of nonlinear dynamical systems to the design of control systems for autonomous agents, in particular ground and aquatic vehicles, enabling them to navigate uncertain and rapidly-changing environments. A hallmark of his research is the implementation of these systems on experimental platforms.