2-D Pole Balancing with Recurrent Evolutionary Networks

The success of evolutionary methods on standard control learning tasks has created a need for new benchmarks. The classic pole balancing problem is no longer di cult enough to serve as a viable yardstick for measuring the learning e ciency of these systems. In this paper we present a more di cult version to the classic problem where the cart and pole can move in a plane. We demonstrate a neuroevolution system (Enforced Sub-Populations, or ESP) that can solve this di cult problem without velocity information.