A genetic programming formulation to evade proportional navigation

Abstract

Genetic Programming evolves a manoeuvre plan for a single evader with constant velocity and a single pursuer employing proportional navigation with lag in a planar region. During evolution GP was required to discover the most effective evasion strategy which generalized across multiple training scenarios, i.e. different initial pursuer positions. GP was trained to evade interception for each of the multiple scenarios by controlling the evader's turn rate throughout pursuit.