24/06/2008, 11:28
|
| | | Fecha de Ingreso: septiembre-2006
Mensajes: 639
Antigüedad: 18 años, 1 mes Puntos: 25 | |
Respuesta: Empire: Total War EMPIRE: TOTAL WAR AI DIARY.
Hi, I’m Jack Lusted a Games Tester at the Creative Assembly UK and this blog will detail my part in the Battle AI development process and how the AI testing works. This blog compliments an upcoming video development diary on the Battle AI.
First a bit of background on myself. Now I’m sure many of you from the community will know me (so you can just skip the next bit), but for those who don’t I used to be an admin of Total War Center, one of the biggest Total War fan sites. I’ve been playing the series since Medieval: Total War was released back in 2002. I’ve been working here at the Creative Assembly UK since June last year as a Tester and secondee to the Empire design team.
As a tester, I have a wide range of tools available to me with which I can see exactly what the AI is doing and thinking. This makes it easy to spot problems with the AI, and helps Richard Bull (Battle AI programmer) more quickly fix any problems that are found.
There are a variety of battles that we have set up that we use to test the AI on. Most battles will consist of me trying a variety of tactics against the AI to see how it reacts, and noting what it does and doesn’t do well. Other times we will run the game so the AI controls both sides and fights itself, to see how well it does in that situation. That is also one of the most useful ways of exposing flaws in the AI. As the project goes on, the AI will be tested under a wider variety of situations to make sure it plays well in all battles.
During each battle, we can play with all the AI debug information being shown. This lets us see exactly what each AI unit is doing and what tactics it’s engaging. This allows us to get a very clear picture of the AI’s overall strategy and so makes it easy to see where it could be improved and where it isn’t quite behaving right.
Every fortnight myself, Richard and one of the Designers will meet up and discuss the progress of the AI. We’ll talk about the issues with it, new things that have been added in and other progress that has been made since the last meeting. We’ll also make suggestions for improvements by discussing how the AI should react in certain situations and how it should play. For instance some new code has recently gone in based on an idea to try and improve the organization of the AI during the later stages of battle. After each meeting we’ll generally have a few specific areas of the AI that we’ve been asked to test before the next meeting.
The close link between myself as a tester and the battle AI programmer is new for Empire and this process of regular meetings and constant testing helps ensure that the battle AI is always moving forwards. The process will carry on right up until release, and things are looking very promising already.
Also for Empire, I’ve written a series of design documents on how the AI should deal with certain situations and general things it should do based on how myself and others play the game. These documents are constantly updated as more is added to them, new tactics are included and situations arise which have AI design implications. This ensures that there are always up to date documents detailing how the AI should play both generally in battles, and more specifically for certain situations. This helps give clear goals for how the AI should progress, and means we are aiming towards an AI that plays a lot more like a human and so should provide more of a challenge to both our experienced players and those new to the series.
If you’ve got any questions on this blog, please do send them in.
Jack.
Os resumo lo que dice aquí: Que Jack Lusted es uno de los encargados de "testear" la IA, aunque también se encarga de alguna que otra cosilla. También se dice que están provando distintas variables tácticas y estratégicas, y que tienen buenas esperanzas en que esto va por el buen camino. Y punto.
Ahora, ¿Hay algo REALMENTE NUEVO? Os lo voy a decir yo: NO. Más palabrería que lo único que consigue es que pierdas 5 minutos leyéndolo para quedarte con la misma información que tenías desde hace 5 meses, o al menos si eres una persona confiada, te puede servir para tranquilizarte porque le están dedicando un apartado especial a la IA, cosa que si no me equivoco nunca se le había dado tanta importancia como ahora.
Yo estoy en un punto medio en el que sé que es un TW en el que la IA está por detrás de la "impresibilidad" del juego (oséase que sea un juego que impresione a la gente para que lo compre); aunque el hecho de que sean tan pesados con que van a mejorar la IA, me hace contener un atisbo de ilusión en que quizás tengan algo de verdad y tengamos una IA decente.
Así que, en resumen, antes y después de leer cada una de las últimas "novedades" que sacan me quedo con lo mismo: LA GRAN DUDA QUE QUIZÁS ALGÚN DÍA NOS RESUELVAN. Maldita palabrería inglesa.
__________________ Enseñar no es conseguir que otros aprendan a defender sus intereses. Enseñar es que los jóvenes aprendan a tener intereses. R. Spaemann.
Última edición por Chow_Total_War; 24/06/2008 a las 11:50 |