Effects of Different Optimization Formulations in Evolutionary Reinforcement Learning on Diverse Behavior Generation