POMDP Solutions for Autonomous Vehicles

Collaborators: Stefan J. Witwicki and Shlomo Zilberstein
Support: Nissan Research Center - Silicon Valley

We design a collection of POMDPs and implement them on Nissan's fully operational autonomous vehicle prototype for urban intersection decision-making with vehicles and pedestrians.

Relevant Paper: "Online Decision-Making for Scalable Autonomous Systems" IJCAI 2017. [pdf]


Safety in Semi-Autonomous Systems

Collaborators: Luis Pineda and Shlomo Zilberstein
Support: National Science Foundation

We build a hierarchical approach to safely transfer control between agents and humans with both application to semi-autonomous vehicle route planning and theoretical guarantees for a notion of provable saftey within the model.

Relevant Paper: "Hierarchical Approach to Transfer of Control in Semi-Autonomous Systems" IJCAI 2016. [pdf]


Multi-Objective Reasoning With Preferences

Collaborator: Shlomo Zilberstein
Support: National Science Foundation

We solve rich real-world problems using multiple objectives in a natural manner by sequentially solving each objective and allowing for slack--allowable deviation from optimal--to be applied at each step, enabling us to solve complex multi-objective autonomous vehicle decision-making with a human-understandable meaning.

Relevant Paper: "Multi-Objective POMDPs with Lexicographic Reward Preferences" IJCAI 2015. [pdf]