michaelg.info: Michael Goldborough

What interests me about AI in sports games such as this is that given the presence of full behavioural independency of player controllable entities, the limitations of the programmed AI provides a space for players to adopt weakly dominant strategies based upon learning the predictable behaviour of the entities that the AI controls. We can also consider that NPC AI characters can also play a role in teaching players the rules of the game. I will start below by analysing the higher level manifestations of AI in Bot Arena 3 and then proceed to those at lower levels.

Heuristics and Learning

In consideration of the 'evade' challenge level in which the only dominant tactic is to run away along the limits of the polygonal map, someone without access to source-code can conclude that the AI entities have no functions for learning which would enable them to adopt a tactic clear to sentient beings of the human variety, that intercepting the path of the player by moving diagonally is possible when the player adopts the previously mentioned tactic. Once the player determines this evasive strategy, the only ongoing challenge becomes one of reflex.

Behaviour States

The AI always adopts one of a subset of behaviours and switches behaviour automatically based upon the state of the game and the in match objectives. For reference, those behaviours are:
• Follow (player assignable)
• Close-in
• Regular
• and Wander
Given the above list, it must be stated that the list above is only a list of behaviours that the player is made aware of and there may possibly be others. Player notification of AI behaviour and the ability for players to not only delineate between behaviours but also to understand the consequences and characteristics each behaviour presents is well applied in this game.

Full Spatial Awareness

Given this game is based upon a real world concept "Robotwars" (TV) in which humans have full spatial awareness over the domain of play which is a small flat 2 dimensional plane, I don't particularly consider full spatial awareness a design flaw.

Simple Pathfinding and Interaction

Bots under AI control show the function of strait line motion in which when their aim is to approach an objective by setting a target and following a path to the target location at which point they reassess. Low level AI that controls the bots understanding and response to spatial limits such as walls, other bots and NPC husks also exist and interrupt path finding which is I consider pretty fundamental to suspending disbelief in respect to the relation of AI to the physical rules of the space represented in this game.

Another point of interest is the methodology with wich bots peform their main function. Target selection and preference (or lack of) for healing or destroying are quite simple but yet play an important part in making this game challenging and enjoyable.