shows that the distance between team members marginally favours GOAP, the FSMs were substantially out-performed by GOAP in both the Capture the Flag and Domination games with squad behaviour enabled. GOAP without squads even scored higher against the FSM with squads in the Capture the Flag game. So what explanation is there for the GOAP superiority as shown in the results? The first possible reason goes back to the results from the Deathmatch and Last Man Standing tests. The GOAP system just acting without squad commands was able to stay alive longer and kill better than the FSM. So when attempting to satisfy a game mode’s goal using squad commands obviously an agent that can stay alive longer and kill better has a higher chance of successfully completing the order or goal.
Another reason that the GOAP system out-performed the FSM system is due to the fact that the FSM is reactive and if it receives an order, more often than not it changes state immediately to follow that order. However, the GOAP system weighs up possible options by evaluating the relevancy of all its goals, the order may affect the relevancy of a goal but the GOAP system may decide to take some other course of action instead and come back to the goal that satisfies the order at a later stage. This prioritisation of goals can bring about unexpected behaviour in the GOAP system and is one explanation as to why the GOAP system performs so well in the experiments.
Human error could be another factor as to why the results favour the GOAP system. Devising all the state transitions for the FSM system was tricky even for the few states that were in this game and it is a possibility that the state transitions were maybe incomplete or even incorrect. This is one of the most common complaints encountered by AI developers when using FSMs and becomes especially difficult to manage when the number of states increases (Isla, 2007). The state transitions didn’t need to be created for the GOAP system so it wasn’t an issue.
Orkin hints at one reason why the GOAP system could out-perform a FSM explaining that GOAP is declarative and FSMs are procedural (Orkin, 2005). The GOAP