the goal of the lagr program is to develop a new generation of learned perception and control algorithms for autonomous ground vehicles, and to integrate these learned algorithms with a highly capable robotic ground vehicle.
an overview of the net-scale/new york university robotics software solution can be found here.