CEDEC 2011 Report has been translated "ARMORED CORE V" I realized in that? Advanced AI challenge to their goals as human action plan, aims to combat flexibility NPC Place between days 6 and August 9 Venue: Pacifico Yokohama Graphics, animation side by side, the game AI has become not remove elements to achieve the game. From Software has continued in the progressive approach in this area is a mechanism that is customized action plan to launch in January 2012 "ARMORED CORE V (below" ACV ")" Toward evolved further to realize that the game AI. Second day session of the CEDEC 2011, "Hierarchical control of aircraft and goal-oriented planning AI matchup ARMORED CORE V" titled, was presented the results of AI research that research is underway right now. The announcement was made to belong to Mr. Nobuyuki Okamura from software engineering department. Future look, the game's AI to achieve a more sophisticated gaming experiences. I want to focus on both fun and that's the difficulty. ■ work on the From Software "Goal-oriented planning" tiered approach From Software Engineering Department, Nobuyuki Okamura "ACV" Requirements of AI. Pretty "smart" you are trying something Lecture by Mr. Okamura Game AI research is carried out in two sessions, the first part is "ACV" approach was shown in complex terrain in the search path used by the three-dimensional. In the second half and covered in this paper, control of aircraft on the correct path was given, it was based on "hierarchical goal-oriented planning" was introduced to the efforts. "ACV" and the concept of race as one of the strategic team to focus in, and strategic decision to NPC, several layers of intelligence is needed to implement the decision. To do so, "goal-oriented planning" tiered approach to the handling of it was that AI's history in the first place. In this approach, the treatment of type AI to react to a simple conditional to changes in the state, unlike their own actions and set targets AI, it will flow to a concrete action plan to achieve goals , and is realized close to the mindset of human behavior. For example, when playing against players NPC, NPC AI is a strategic goal of "survival of the ship," "destruction of the enemy" and be something. The AI at the same time, it must respond to what happens locally. Avoid enemy's attack or move in order to hit an enemy attack, and that's part of attitude control. Performed in parallel in order to determine the particle size is very different this time as "hierarchical" is not required for processing. Actually, this "goal-oriented hierarchical planning", was released in 2006 for the Xbox 360 "Chrome" has been attempted at. Okamura's mission, while based on technology developed there, more complex actions, with a more complex spatial configurations "ACV" Toward said that it can evolve to its contents. The challenge has become a particularly large "aircraft control" part of it. The speed at which robot action "ACV". In this work the "goal-oriented hierarchical planning" is trying to achieve the strategic AI teams do battle with. It's flexible approach to changes in the assembly to withstand the mechanical parts The concept of goal-oriented planning Hierarchical planning goals. To achieve top-tier planning structure is organized in smaller units are planned for the lower layer First, "hierarchical goal-oriented planning" is trying to introduce rough handling that is what actually went AI. "ACV" AI for the "strategic layer" layer "tactics", "aircraft control layer," which consists of three layers. Action Plan in the upper layers (goal) is set at the lower layer can take steps to realize their idea that they will be assembled as individual goals. Part of the treatment goal to make the appropriate settings in each tier, "Planning" called. Here are a number of possible characters from the current state of "one hand" and then go to virtually run to predict the state of the result. This process is performed in countless combinations to reach the goal state, given by the upper tier "steps" if you find a posting that they do it. The AI of chess and Go, find the best move the process closer to defeat the opponent As an implementation of the AI for this game is not very rich, active in more AI, there is merit to act flexibly line wither. Looking to the player, decent behavior of NPC "intent" feel, the greater the fun it. Attaching a function of change in the future to predict the state of AI, but we have attractive characteristics that are unique to the behavior of the NPC. Much will approach the game that could be achieved with feedback Layer consisting of 3 "ACV" goal-oriented planning in detail ■ Too Rich?! With the action and complex terrain "ACV" in the experimental solutions Working with inertial systems, which followed a route determined in cases of unexpected difficult You do not fall from scaffolding, the move operation, "planning" to create an out how to do that? Planning to take advantage of the configuration process control function is used to move aircraft operation "Goal-Oriented Hierarchical Planning", while the AI behavior can be expected in a very interesting, but there is a weakness of the technical difficulty that is very easy to increased memory and CPU utilization. In particular, "ACV" complex terrain and is expressed in a variety of actions that ruffian. Because of the enormous spread of the possible actions AI, an explosive increase in the future state to be predicted, but because of the increased cost projections. So that was current rumor later in the session layer "body control" is part of the implementation. For example, in order to achieve the tactical goal of moving to a point layer "body control" to move the case to assemble. Movement path (path) search has been resolved in three dimensions AABB method. That path, in combination with the ship moving action should Nazoranakere properly. FPS based games, such as the human hero character, is relatively simple because the movement control may be just a vine on a path tracing Innovation. However, "ACV" deals with the movement of robots working in a strong inertia. And there could be more complex actions such as kicking the wall to jump. To trace the route "steps" I find the challenge process. Mr. Okamura is on trial and error, using the state update function is used to move aircraft, took an approach to predict the transition. Give input to the function of the same players move operations to simulate the future, from the combinatorial explosion, but finding a solution to Trace a path to reach the goal. That way, all terrain, all expectations, in all circumstances, may be planning a move to flexible operation. As with human players, but not strong AI is achieved to effectively utilize the various forms of terrain. Implemented layer "body control" produced by the movement. That can be traced with great precision the complex movement paths. Which also input from the controller, of course, the same condition as when you operate the player. And various other bogus or will be forced to move or warp is not Processing task too heavy, too. Close that if the "ACV" is used in it? The problem is "too heavy calculations" that. CPU cores (or any model untouched) one, it was occupied only would result in this process. Given too much too heavy to move as a simple AI. We optimized in any way, it is essential for practical applications can dramatically reduce the processing load. Okamura said, "Optimization of collision data," while another mentioned as a solution to a fundamental performance improvement is a challenge in the future, and reported the situation. In other words, this technique is still not a developing research and development. What's interesting about this announcement is not about the game worked with elements of pre-release, it became the theme of advanced-stage studies of not knowing whether the game actually used. Game AI research and development that while Mr. Okamura is a very sophisticated, has faced challenges with resolve towards commercialization. Soredakeni, some of the feedback that I not get a big harvest. Possibly, generations of game machines may now come to the conclusion that it is difficult to accomplish. However, players like to think, if implemented to fight the AI and game play alone, play in multi-player game mixed with NPC, should become more interesting. Really, to be released in January 2012 "ACV" I realized that what enjoyment, and hope to expect. □ Computer Entertainment Association (CESA) homepage Http://Www.cesa.or.jp/ □ "Computer Entertainment Developers Conference 2011 (CEDEC 2011)" You are Http://Cedec.cesa.or.jp/2011/ □ From the home page of the software Http://Www.fromsoftware.jp/ □ "CORE Armored V" page http://www.armoredcore.net/acv/
