Application of a Factorial Designed Experiment to Optimize Selection of Reinforcement Learning Observations for a Hexapod Trajectory Following Task