Thursday, 4 December 2008
Presentation Feedback
Question is slightly unclear. The word better should not be used as it is not objective. Also the possibility that a more complex usage of G.A. may increase overall complexity quite a considerable amount.
Presentation
How can Genetic Algorithms best be used to create non-playable characters which are capable of adapting to better compete against different types of player in a real time game environment and what are the issues associated with their use?
Matthew Lock
Technical Supervisor: Dave Ellison
Create A.I. entities that are capable of adapting
Entities must remain flexible
Entities need to improve while remaining realistic
I.E. should not be “overpowered”
Performance hindrance should be minimal
Entity can learn at runtime
Makes play more player specific
Makes entities more human like
Multiplayer game – Based on Bomberman
Implement Genetic Algorithm for A.I. players
Use kills and survival as measure of success
Compare with a Basic “A.I.”
Fairly advanced entities.
Matthew Lock
Technical Supervisor: Dave Ellison
Create A.I. entities that are capable of adapting
Entities must remain flexible
Entities need to improve while remaining realistic
I.E. should not be “overpowered”
Performance hindrance should be minimal
Entity can learn at runtime
Makes play more player specific
Makes entities more human like
Multiplayer game – Based on Bomberman
Implement Genetic Algorithm for A.I. players
Use kills and survival as measure of success
Compare with a Basic “A.I.”
Fairly advanced entities.
Worksheet 4 Feedback
(Also old)
An interesting topic.
I think you understand the issues very well with this topic.
The question is a little long. I would truncate the last part as it should be implied by answering the first part of the question, so it would become:
How Can Genetic Algorithms best be used to create non-playable characters which are capable of adapting to better compete against different types of player in a real time game environment?
I think you might be taking on a lot of work here. Not only do you have to create a game, but also several different genetic algorithms. I would maybe just choose to of the most common genetic algorithms and do a comparison between them.
It is very good that you have recognised the judging criteria of the AI, though this might not be enough. You might want to have an aggressiveness factor included too. The reason is an AI could survive for a very long time by hiding from all opponents. This would make it a good strategy though a boring one which could infuriate a player.
Carry on the good work, I hope it goes well.
An interesting topic.
I think you understand the issues very well with this topic.
The question is a little long. I would truncate the last part as it should be implied by answering the first part of the question, so it would become:
How Can Genetic Algorithms best be used to create non-playable characters which are capable of adapting to better compete against different types of player in a real time game environment?
I think you might be taking on a lot of work here. Not only do you have to create a game, but also several different genetic algorithms. I would maybe just choose to of the most common genetic algorithms and do a comparison between them.
It is very good that you have recognised the judging criteria of the AI, though this might not be enough. You might want to have an aggressiveness factor included too. The reason is an AI could survive for a very long time by hiding from all opponents. This would make it a good strategy though a boring one which could infuriate a player.
Carry on the good work, I hope it goes well.
Worksheet 4
Worksheet 4 (old adding now as had forgotton to upload it)
Introduction
The topic of the project is the usage of genetic algorithms. The aim of the project is to use genetic algorithms to create A.I. entities capable of altering behaviour in a real time game environment to better compete against players or other A.I. entities, while remaining relatively flexible. Also numerous different variations will be implemented to allow further insight into what works best in a given situation as well as into the issues involved with the methods.
Issues
The issues to be addressed are how best genetic algorithms can be used to create A.I. entities that can adjust to perform at a better standard in a real time game environment. It is vital that such entities remain flexible and are realistic in there playing ability, i.e. they should not have super-human ability at the game. Also because of the real time aspect it is important to consider any impact on performance, alongside any other issues that may be involved in the implementation.
Research question
How Can Genetic Algorithms best be used to create non-playable characters which are capable of adapting to better compete against different types of player in a real time game environment and what are the issues associated with their use?
Addressing the Question
I will write a multiplayer game, likely to be based on Bomberman, that will contain enough decision making and strategy for the computer controlled players to be able to learn, through playing, which techniques and strategies are the best way to play the game. I will then use several varying techniques to control this A.I. learning and judge them on success, based primarily on number of kills but also on time survived, as well as on flexibility and effect on performance. This will allow me to figure out which methods work best as well as evaluate any issues with the techniques.
Progress
So far I have researched around the area of genetic algorithms as a means of adaptive A.I. to help redefine exactly what I’m doing. This has enabled me to better aim my work and to understand what I am doing and what I need to do from now on.
Introduction
The topic of the project is the usage of genetic algorithms. The aim of the project is to use genetic algorithms to create A.I. entities capable of altering behaviour in a real time game environment to better compete against players or other A.I. entities, while remaining relatively flexible. Also numerous different variations will be implemented to allow further insight into what works best in a given situation as well as into the issues involved with the methods.
Issues
The issues to be addressed are how best genetic algorithms can be used to create A.I. entities that can adjust to perform at a better standard in a real time game environment. It is vital that such entities remain flexible and are realistic in there playing ability, i.e. they should not have super-human ability at the game. Also because of the real time aspect it is important to consider any impact on performance, alongside any other issues that may be involved in the implementation.
Research question
How Can Genetic Algorithms best be used to create non-playable characters which are capable of adapting to better compete against different types of player in a real time game environment and what are the issues associated with their use?
Addressing the Question
I will write a multiplayer game, likely to be based on Bomberman, that will contain enough decision making and strategy for the computer controlled players to be able to learn, through playing, which techniques and strategies are the best way to play the game. I will then use several varying techniques to control this A.I. learning and judge them on success, based primarily on number of kills but also on time survived, as well as on flexibility and effect on performance. This will allow me to figure out which methods work best as well as evaluate any issues with the techniques.
Progress
So far I have researched around the area of genetic algorithms as a means of adaptive A.I. to help redefine exactly what I’m doing. This has enabled me to better aim my work and to understand what I am doing and what I need to do from now on.
Subscribe to:
Posts (Atom)