ArtsAutosBooksBusinessEducationEntertainmentFamilyFashionFoodGamesGenderHealthHolidaysHomeHubPagesPersonal FinancePetsPoliticsReligionSportsTechnologyTravel
  • »
  • Technology»
  • Computers & Software»
  • Computer Science & Programming

Visualization module behavior of intelligent agents in a developing of controlling system

Updated on March 30, 2016

Abstract

Visualization behavior of intelligent agents is useful in the development process during debugging multi-agent system. We have to visualize the coordinates of agents, the travelling direction, the search direction, the current action ( "think" , search, selection , returns ) , as well as custom debugging information in the form of graphics primitives and text messages . This paper considers to develop the module of visualization behavior of intelligent agent. Attention is given to methods of controlling agents of a group, the mechanisms of interaction between agents. Visualized parameters are characterized the state of the agent at a particular time.

1. Introduction

Multi-agent approach - it's one of the modern and active-developing approaches in the field of artificial intelligence. Currently, multi-agent systems are used in many fields, especially in computer games. Under the intelligent agent (AI) in games understand the agents that make the decision on implementation of actions based on the results of the analysis of the state of the environment - the virtual world game[1][2][4].

In [3] described a system for game AI. The agent is built on a system similar to a human player, so he has "intelligence". The system solves the problems that relate to behavior of management agency. Agents in the system have sufficient intellectual behavior. However, in view of the current version of the constructed model is carried out.

Visualization AI behavior is useful in the development process during debugging multi-agent system. Visualized coordinates agents, the direction, the direction of the search, the current action ("think", search, selection, returns), as well as custom debugging information in the form of graphics and text messages primitives.

Work is devoted to development of methods for managing agents, mechanisms of interaction between agents and the development of the visualization module.

2. Analysis of imaging systems behavior of intelligent agents (AI)

Today there are a lot of behaviors of modeling agency. By Including can be enumerate the following systems: modeling “NetLogo”[5] , the simulator "MASON" [6] , the simulator “Ascape” [7] , the simulator "Repast" [8]. Such systems are implemented based on the multi-agent approach.

All current supported 2D- graphics, besides modeling system "MASON". Experienced programmer can create functions for the model agent, configure or monitor settings agents.

In all systems, there are renderer behaviors of AI. But the debugging information is shown in format text and graphic , and not attached to the agent.

These systems modeling visualized state of the virtual world and the behavior of agents. The system «NetLogo» visualized on the number of parameters and energy agents using graphics. Other parameters agents are not shown , for example , speed, direction , behavior, etc. This is lack of system.

System modeling the behavior of agents can solve the problem , according to user's requirement .

The module was developed in order to improve the quality of debugging multi-agent system using the debugging information, which took place in the above systems.

3. Implementation method module

The algorithms control the behavior of AI (Steering behavior), find the shortest path algorithm and the method of synchronization agent components.

The Algorithms of management of conduct AI is done optimally. Each behavior AI developed by a separate class. When starting a new behavior, you just called new instance of the class.

The algorithm of finding the shortest path is realized on the basis of the algorithm, which is widely used in computer games.

Method synchronization agent component was implemented, like it says in section 5.

4. System Architecture

The system consists of 2 main components:

- Rendering subsystem is the component to draw the scene. Such a component can be implemented by a variety of means: by using a low-level graphics APIs(DirectX, OpenGL, GDI++) or high-level graphics and games libraries(ORGE, Havok, Irrlicht Engine).
- Subsystem scene construction is the main component that converts the parameters in the multi-agent system into a format suitable for use graphics library. This subsystem contains algorithms of data processing and control algorithms of conduct AI (Steering Behavior).

1 - Input data: the Current settings of the model agent: coordinates, linear velocity and angular velocity, acceleration, the status of each manipulator agent) and information for control algorithms behavior (Steering Behavior.
2 - The Sequence of commands to render the world.
3 - Output:

- The image of the agent in the virtual world.

- Table of parameters of the model agent bounds to the agent.

5. Description of the agent and the virtual world

All objects of the virtual world are represented as composites graphics, physical and behavioral models. Passive objects do not have behavioral model. The structure of the active object (agent) can be presented as follows:


Agent = {graphical component, physical component, AI component}

Graphical component uses data physical component, which examined the AI component for rendering on the stage. Such data are the new coordinates, the new state of each arm of the agent and primitives debugging information.

AI component uses data physical component for decision making and navigation. This component is used for modeling the behavior of active objects in the future active object has "intelligence". The agent has knowledge of a virtual environment (the state of the environment, the state and behavior of other agents in the environment). AI component analyzes such knowledge and process them. In AI component were developed algorithms of processing the events of the world and algorithms of management of behavior AI.

The physical component stores all of the physical parameters of the agent in the virtual world. Was developed synchronization physical and AI components.
The virtual world of any game consists of many areas or locations and objects placed in the locations. The structure of the virtual world game in General is as follows:

W = (,),

Where:

- set of regions

- set of objects.

Objects can be passive and active. The area is also released into 2 types: one type is the area through which the agent cannot pass, and the other type is the area through which the agent can take place.

6. Testing module

To check the main provisions of the concept, was designed and implemented a prototype system that simulates the behavior of agents in a three-dimensional world.

Physics of the virtual world represents the physics of rigid bodies in a three-dimensional world. There are many similar agents. Agents can move around in a virtual world. Agent randomly is looking for flower. If the flower is from the agent at a certain distance, then sees it and starts move to the flower. He takes the flower and go "home" to a specific area on a map.


There were developed the following basic tactics:

- Avoidance of obstacles in the way of our character's something to be and the character will need to do something to this, something to avoid.

- Wandering - agent moves around the screen without any purpose to find flower.
- Arrival - the agent on arrival at a certain point gradually slowed down and stop, when it reaches to this point.

- Selection of flower - agent takes the flower.

7. Conclusions

In this paper we have developed a module on the basis of multi-agent approach to visualize the behavior of AI. There are rendered not only the state of the world, but also the parameters of each agent in the certain moment of time. During the inspection test example agents demonstrated the behavior accordingly reality.

The module allows you to visualize the set of fixed parameters and not an arbitrary set of parameters specified by the user.

In future, you can modify the module to fix all the problems and use the module for the development of model intellectual agents during debugging of multi-agent systems.

References

1 Tarasov V.B. have been Agents, multi-agent systems, virtual communities: a strategic direction in computer science and artificial intelligence/ V.B. have been Tarasov //news of artificial intelligence. - 1998. -№2. - P.6-9.
2 C. M. Macal Tutorial on agent-based modeling and simulation / C. Macal, M.J. North // In the Proceedings of the 2005 Winter Simulation Conference. - 2005. - P. 2-15.
3 Alimov, A.A. Artificial intelligence in video games. Multi-level planning and reactive behavior of agents / A. Alimov, O.A. Shabalin, " Izv. VSTU. Series "Actual problems of management, computer engineering and computer science in technical systems". Vol. 10 : meiwes. collected scientific articles. Art. / VSTU. - Volgograd, 2011.-№3.C.90-94.
4 Shabalin O.A. System's artificial intelligence / A. Alimov, O.A. Shabalin // proceedings of VSTU. Series "Actual problems of management, computer engineering and computer science in technical systems". Vol. 13 : meiwes. collected scientific articles. Art. / VSTU. - Volgograd, 2012. - № 4 (91). - C. 166-169.
5 modeling System "NetLogo" Mode of access: http://ccl.northwestern.edu/netlogo
6 simulation System "MASON"/ access Mode: http://cs.gmu.edu/~eclab/projects/mason
7 modeling System "Ascape" / access Mode: http://ascape.sourceforge.net/index.html%23Contact
8 System modeling "REPAST"/ access Mode: http://repast.sourceforge.net/repast_3/index.html

Comments

    0 of 8192 characters used
    Post Comment

    No comments yet.