Advanced Applied Probabilistic Programming

Autonomous cars are one example of a compelling next-generation artificial intelligence technology. In order to safely navigate through the world, cars must plan long-range routes and short-range paths, perceive the world around them, and act according to a safety-first policy that takes into account the intent of agents in their surrounding world. While not strictly AI-complete, the challenge of autonomous driving in urban and unstructured environments is substantial, as-yet unsolved, and of paramount economic importance. This research is relating to the most significant challenges remaining to be solved before it becomes possible to make unrestricted autonomous cars a reality. In particular, this research will focus on theory of mind, inference of driver, pedestrian, and cyclist intent, and robust, computationally efficient solutions to these and other inference problems. The aim of this research will be to build the probabilistic programming software systems and tools that will make it possible to efficiently build models that predict what the various agents near and on the roadway will do up to two to three seconds into the future, or at least long enough to allow for contingencies that ensure that the controlled vehicle behaves safely.

Faculty Supervisor:

Frank Wood

Student:

William Harvey;Jonathan Wilder Lavington;Saeid Naderiparizi;Peyman Bateni;Renhao Wang

Partner:

Inverted AI Ltd

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

Program: