4AI for Fighting - Analysis Attack Point – High, Medium, LowDefense Point – High, Medium, LowProfessional Attack Distance – Long, Middle, ShortMovement ability – Dash, Run, Walk, Jump
5AI for Fighting – Attack / Defense CaseHP / StatusAttack Type Range: ShortAttack Type Range: MidAttack Type Range: LongRival DistanceOutput Action1High / StandHigh Damage Long IdleNILMid – Range1. Dash front 2. Attack in short distance2Low / StandLow Damage Very Short IdleMedium Damage Long IdleFar1. Stay 2. Attack in long distance3Medium / StandHigh Damage Short IdleLow Damage Short IdleMedium Damage Very Long IdleNear1. Dash Back2. Attack in mid distance4Medium /Be Attack1. Blocking2. Combo Attack
6AI for Action Game Move Closer To Player Dash Attack Combo Attack TacticRankAISkillTacticBig BossSmartwalk, run, dash attack, normal attack, combo attack, special attackFor example, move to player backward when player is fighting with others.Small Bosswalk, run, dash attack, normal attack, combo attackNoNormal SoldierNormalwalk, normal attack, combo attack
10A* (A Star) Algorithm Heuristic Estimation Weight of path Cost Comparing
11Concept of A* (A Star) Add the starting node (tile) to the open list. Look for the lowest F cost node on the open list. We refer this as the current node.The node move from open list to close list.For each of the 4 nodes, top, right, down and left adjacent to this current node. If nodes note closed or not reachable then skip them. If the node is in the open list already, check the G cost. Select the linked note of lowest G with cost 1. If the node is not in the open list, then add to open list.Repeat step 2 to step 4 until found the target node (tile) and return the path. But if the open list is empty and not reaching the target tile yet that means you can't reach the target node and return false.
16Fuzzy Logic Degrees of value Fuzzy State Machine If True is 1, False is 0. We should have 0.1 fuzzy values, which near false but not absolutely false.Fuzzy State MachineWhen design AI State machine, we should apply fuzzy values for state controlling and state presentation
17Traditional fuzzy state machine Number of VariableNumber of value exist in each variableTotal rules generation2552 = 25353 = 125454 = 62555 = 3125656 = 15625
18Traditional fuzzy state machine (Example) My Team ScaleEnemy Team ScaleMy Team Attacking StatusStrongAttack CarefullyAverageAttackWeakFull AttackRunaway
19Combs methodWilliam E. Combs designed a method to prevent combinatorial explosion in fuzzy logic rules. The method describes the direct relationship between every variable and the fuzzy value of state. AI system then combines all results to generate the final result.
20Combs method (Example) Variable : ValueAttacking StatusMy Team Scale: StrongFull Attack (90%)My Team Scale: AverageAttack (60%)My Team Scale: WeakAttack Carefully (30%)Enemy Team Scale : StrongEnemy Team Scale : AverageEnemy Team Scale : Weak
21Combs method (Example) My Team ScaleEnemy Team ScaleMy Team Attacking StatusStrong (90%)Strong (30%)Attack Carefully (27%)Average (60%)Attack (54%)Weak (90%)Full Attack (81%)Runaway (18%)Attack Carefully (36%)Weak (30%)Runaway (9%)