The Effects of Human-like Modifications to Heuristic Action Evaluation in Video Game Pathfinding

We present a series of parameterizable modifications to heuristic evaluation of actions in the A* algorithm, designed to create more human-like and dexterity-robust paths through games in the 2 dimensional platformer style. We attempt to create paths at various levels of player skill by imposing constraints onto the timing and duration of actions designed to mimic human reaction times and ability. We show that these action value modifications result in the A* search algorithm producing smoother paths, taking safer routes to avoid danger, and requiring fewer actions to be performed in a given amount of game time.