Extending Online Planning for Resource Production in Real-Time Strategy Games with Search

Planning in domains with temporal and numerical properties is an important research problem. In prior work, we presented an online planning architecture for resource production problems in RTS games. At every decision point, our planner constructs plans that achieve a certain fixed set of intermediate renewable resource goals. It then uses the plan with the smallest makespan to choose an action at the current decision point. While each plan it considers is suboptimal, we showed empirically that this action selection strategy was competitive with human players in solving large resource goals. In this work, we investigate the effect of reducing one source of suboptimality in the plan generation step. Instead of considering a fixed set of intermediate resource goals, we allow the planner to search over a variable but bounded set of goals. We investigate empirically the plan quality of three different problems for our planning architecture.