Tactics/strategy Artificial Neural Network and Q Learning.



Tactics and Strategy

The subjects I will now talk about in relation to my game are tactics and strategy. These techniques are used by the AI's to overwhelm the player. Tactics are where the AI's make individual decisions to attack the player, for example they could hide in the shadows, snipe from afar or attack from behind cover. Strategy is when the AI characters work together to form an overall  plan to eliminate the player, this can be where the AI's form a large team to attack and overwhelm by volume. I do not think that either tactics or strategy would be particularly suitable for my football game, because there is no real need for tactics or strategies to eliminate or overwhelm AI characters in a sports game. However, I think these would be very useful in my shooter game idea because the AI characters would need to overwhelm the player to eliminate him. 




Image result for Tactics and strategy
      

      
Artificial Neural Network

The technique I will now talk about in connection with my game is Artificial Neural Network, this in basic terms is a program that is designed to simulate the human brain, which means that AI characters can learn from their experiences. They also learn how the player plays and can therefore adapt to counter the players play style, this then gives the AI's a great advantage. An example of this could be that the AI character has attacked the player from the front and if this has failed the AI learns from his mistake and next time will attack from behind as they have learnt that the player favours the front. This was used in the game Metal Gear Solid Phantom Pain when the AI's were shot in the head they learnt from this and stated to wear helmets, when attacked at night they used flashlights in order to see the player more easily and if they were attacked from the front and thrown to the ground they would use shields to protect themselves. Also the AI's can learn the ways and tactics that a player uses to play the game, such as using decoys to fool the enemy, they will learn this and implement it themselves to trick the player. In my football game this could be used effectively as the AI characters will learn the most effective way to attack the opposing side and gain control of the ball in order to score a goal. Also the AI's will learn the players play style and adapt to this giving them the upper hand. This will make the game much more realistic and more of a challenge to the player who in order to win will have to keep altering his playing style.



Image result for Artificial Neural Network



Q Learning

The last AI technique I will talk about is Q Learning, this is where AI characters or agents look back at their history having gone down all available paths repeatedly over a period of time, by this repetition they are able to find the best route possible. Each time that an AI character or agent gets to the target they are rewarded with a set value such as 100.  With regard to my game I think that Q Learning could be suitable because the AI character would learn through previous attempts the best route to get to the goal and once they have scored they will be rewarded with 2 points.

     






             

Comments

Post a Comment

Popular posts from this blog

Decision making concepts,State machines and Fuzzy logic

Image
Decision Tree The first thing I will talk about in this section of the assignment, in connection with my game, is Decision Tree, which can also be known as Classification Tree. This is a tree like graph that collects together all of the decisions that the AI has made and stores them in this tree graph.  These decisions are made up of parent and child nodes, the parent nodes represent various choices while the child nodes represent possible decisions, like yes, no or maybe.  For the tree to work properly the child nodes have to represent all of the possible decisions that can be taken.  I think that I could use the Decision Tree in my football game to  allow the players to make decisions such as attack from the left hand side, right hand side or pass to another AI player.     State Machines  The next subject I will talk about is State Machines, this involves AI characters switching between various states of play within a game, such as wandering, patrolling, attacking or
Read more

Assignment 2 Algorithms, heuristics and hacks

Image
Algorithms The first thing I will talk about is Algorithms, with regards to the 3D game I am planning  to make, I will explain how I will be using them. An Algorithm is a simple set of rules or instructions,  that apply to the AI character,  which a coder puts into a game. These Algorithms are the baseline for creating any type of game, therefore I think they would be suitable for the 3D game I am going to create.  Firstly I could use it to help the AI player or players navigate around the 3D map, this will enable them to move forwards, backwards, left and right.  An algorithm could be used to create a ragdoll affect so that when one player collides with another at speed, one or both players would fall to the ground as if injured.  If the collision is considered to be deliberate by the player not on the ball a free kick will be given to the player that was fouled.  I could also use this to define which of the AI characters is the referee, this would mean that he would not play the g
Read more

Assignment 2 Boids Flocking, A* Algorithm,Dijkstra Algorithm and Cost Driven

Image
Boids Flocking The next AI techniques I will talk about in connection with my game is Boids Flocking  and if I could use it  in any way in my football game. Boids Flocking is an AI techniques that is meant to simulate the way birds flock together in a group, the way they fly closely together avoiding each other when they get too near. This technique can also be used for groups of people such as in the game Assassin Creed, when the developers tried to create a city with the impression of large numbers of people. Other animals could also be used for example stampeding in fright or hunting in packs.  I think Boids Flocking would be suitable for my football game as it can simulate a group of players running after the ball and trying to score a goal, but a limit on the number of players grouping together would have to be put in place, as in a real football game not every player runs after the ball at the same time.   A* Algorithm The next AI techniques to talk about regarding m
Read more