Abstract: Autonomous vehicles and systems promise transformative benefits in transportation, scientific exploration, and other applications, but fielding such systems has proven extraordinarily difficult. One particular challenge is handling uncertainty. If an autonomous system is overconfident about its surroundings, it will be efficient in accomplishing goals most of the time, but will lack safety. On the other hand, overly conservative assumptions about uncertainty can make systems safer but may also make the system impractically inefficient in achieving other goals. This talk will present a simple taxonomy for classifying uncertainty consisting of the following classes: aleatoric, epistemic static, epistemic dynamic, and interaction. With this framework, the talk will discuss tradeoffs inherent in different modeling and solution approaches and give an overview of work at the Autonomous Decision and Control Laboratory (ADCL) at CU Boulder to develop scalable algorithms and fast Julia reference implementations for safe and efficient autonomy in the face of uncertainty.
About the speaker: Zachary Sunberg is an Assistant Professor in the Ann and H.J. Smead Aerospace Engineering Sciences Department. He earned Bachelors and Masters degrees in Aerospace Engineering from Texas A&M University and a PhD in Aeronautics and Astronautics at Stanford University in the Stanford Intelligent Systems Lab funded by an NSF Graduate Research Fellowship. Before joining the University of Colorado faculty, he served as a postdoctoral scholar at the University of California, Berkeley in the Hybrid Systems Lab. His research is focused on safe and efficient operation of autonomous vehicles and systems on the ground, in the air, and in space. A particular emphasis is on handling uncertainty using the partially observable Markov decision process and game formalisms.