Approximate Soft Policy Iteration Based Reinforcement Learning for Differential Games with Two Pursuers versus One Evader | IEEE Conference Publication | IEEE Xplore