Multi-Agent Systems

Multi-Agent Systems

By Daniel Copeland, C.J. Dopheide, Charles McQueen, and Mantej Singh

Multi-agent Systems


A Multi-agent system is a system in within an environment can be composed of intelligent agents. These multi-agent systems can help solve problems. It can solve difficult or impossible for a single agent to figure out and solve. Multi-agent systems are distributed computing systems. Just like all the other distributed systems, these are composed of many amounts of interacting computational entities. The difference between these systems and the normal distributed systems is that these are smart.  The modernization is useful when designing things for applications. Things like medical diagnosis with these systems. These systems are very important in computer science. These systems have been studied as part of a fake version of smart.  These systems have come up as one of the very important areas of research in computer science. The goal of this system is to find methods that let us to build systems that are composed of autonomous agents. These agents are capable of enacting the desired global behaviors. The whole idea behind the multi agent system is a very simple one. An agent like this is a system on a computer that allows of things to happen that its user wants it to. It can figure out on its own what it has to do to make its design and objectives. This multi agent involves multiple agents. These agents interact with each other to exchange messages through computer networks. They respond to users and owners with different behaviors.  These current systems have several decades been investigating systems that contain multiple interacting components. A multi agent system is a system that is made up of multiple agents that communicate with each other. The easiest cases will, the same people program all of these agents so they know what they are doing together. To successfully communicate with each other, they will have to cooperate, coordinate, and negotiate with other agents. These multi agents make the most important decisions and are very important to computer science and computers in general.

Multi-agent systems in Computers

A multi agent system is a collection of intelligent, interacting parts. A computer multi agent system is a collection of many separate computers that operate with each other to accomplish complex tasks.  These parts communicate and work together with each other in order to perform complex operations that may be too difficult for one operating system to handle alone. A multi agent system has the capability of dividing the work load of a system so that each part system can function at full capacity. In computer multi agent systems, there are three characteristics that distinguish the individual parts. These characteristics are “autonomy, local views, and decentralization.” An autonomous unit is capable of acting on its own without the need for external prompts to begin taking action. Each of the parts of the system has only the information necessary to complete its designated task in the system. The individual system agent parts are not required to see or know the overall application of the whole system. They need only to have a clear understanding of their specific task and the pass along that information to the next agent in the system. This is what is meant by local view. The agents of a system must also be capable of acting without the aid of a central control system. Each agent is programmed to know its actions and implement them without direction from a central system. This allows multi agent systems to function and communicate effectively and efficiently. Multi agent systems have the ability to the take simple operations of individual units and combine them into massively complex operations. These systems are capable of self-organizing and self-steering when the individual parts function together correctly. Multi agent systems are used in computer simulations, and operate by moving the simulation along in specifically timed intervals. Wikipedia offers an explanation of how components of multi agent systems “communicate typically using a weighted request matrix, e.g.

 Speed-VERY_IMPORTANT: min=45 mph,

 Path length-MEDIUM_IMPORTANCE: max=60 expectedMax=40,


 Contract Priority-REGULAR

    and a weighted response matrix, e.g.

 Speed-min:50 but only if weather sunny, 

 Path length:25 for sunny / 46 for rainy

 Contract Priority-REGULAR

 note - ambulance will override this priority and you'll have to wait

    A challenge-response-contract scheme is common in MAS systems, where

 First a "Who can?" question is distributed.

 Only the relevant components respond: "I can, at this price".

 Finally, a contract is set up, usually in several more short communication steps between sides,


    also considering other components, evolving "contracts", and the restriction sets of the component algorithms.”


    Multi agent systems also perform operations using the pheromone system, where information is left behind for the next agent in the line to use. Multi agent systems can be applied in computer to accomplish a wide verity of tasks such as computer video games, defense technology coordination; computer generated images in movies, and multi agent systems can be implemented in logistics and transportation use as well.  By having simple systems communicate to accomplish complex tasks the work load is divided between coordinating parts of the system and problems that once seemed impossible to solve can be analyzed and completed with relative ease, as long as each system is programmed properly that is. 

All-Robot Multi-Agent Systems

                        As has been stated before, Multi-Agent systems are systems where the interacting parts each have a mind of their own.  In layman’s terms, a multi-agent system could be anything where two independent people/animals/things work together to do...something (what that something is, exactly, can vary quite a bit).  In the most abstract, human society could be classified as a Multi-Agent system as people interact in their daily lives.  However, since this is computer science, and not sociology, this “chapter” (though I’m sure it’s far too short to truly merit the title) will focus on a more traditional definition of a Multi-Agent System: systems of robots working together to accomplish a specific goal.

            All-Robot Multi-Agent systems are relatively new, not necessarily a new idea, but we have only recently been able to begin to construct practical fully autonomous robotic systems due to recent advances in robotics and artificial intelligences.  The most important characteristic of an all robot Multi-Agent system is that it doesn’t have a centralized intelligence; there is no person at the controls or big computer keeping track of each of the individuals in the system.  Al of the robots in such a system have their own (generally rudimentary) intelligence and can only have a local knowledge of what is going on (they only know what they see, they probably don’t have an exact knowledge of what every single other agent  in the system is doing).  This is both the most important aspect of a Multi-Agent system and its biggest drawback.  The drawback comes from the fact that A.) There is no “guiding hand” that can tell if a robot is doing something wrong and B.) They’re really tough to program, getting each robot (which, keep in mind, probably doesn’t have the greatest computational capacity) to work with every other robot in the system in order to achieve something is hard.  The upside to a Multi-Agent system such as this is the flexibility that it allows, since there is no over-seeing intelligence, you can add as many robots as you want (well, not technically, but as many as you’ll ever need, probably) to the system without either straining a Human overseer or having to teach a central processor how to control the new additions.  This scalability allows whoever is using the robots to only use as many as they need (which is both efficient and probably saves money in the long run).

            As of today, most all-robot Multi-Agent systems are in the development phase, such as the U.S. Army’s Unmanned Ground Vehicle program, one of its goals being to delegate the process of advance scouting to, well, Unmanned Ground Vehicles in order to protect the lives of the advance scouts who would otherwise be driving ahead of the main army (Arkin).  Another future application for an All-Robot Multi-Agent system would be in disaster response search-and-rescue scenarios, where a bunch of autonomous robots that are capable of navigating a dangerous environment without tiring or requiring any oversight would be incredibly useful.  A design that illustrates this perfectly would be the Swarming Micro Air Vehicle Network (SMAVNET) being developed by the French Federal Institute of Technology which consists of a number of unmanned aerial vehicles that would survey a disaster area for survivors and dangerous conditions (Schwartz).

            While most practical All-Robot Multi-Agent systems are still in development, they promise to revolutionize industry, disaster relief, and many other aspects of society that would take me a very long time to list, and so I won’t



 Human-Robot Multi-Agent systems

Robots are experiencing new program places in support industry everyday. There is a need of robots for healthcare surgery treatment , resort business, cleaning of bedrooms or family. In contrast to commercial robots, which will work in a well known atmosphere and used by expertise providers, it is not the situation for support robots. Service robots have to work in a ”complex” atmosphere, where objects are not actually known or have not a known places nor trajectories. In revenge of appealing studies in the area of synthetic intellect, independent robots have still complications to execute complex projects in ”complex” atmosphere. The more independent the program is, the more arranged the globe design should be, and the more particular its projects are. A little difference between the globe design and the actual life causes the system break down. Consequently independent robots experience from common industry limitation . Up to now it is still not known how to accomplish finish independent performance of complicated projects in unstructured atmosphere .A possible bargain to prevent the industry limitation issue is to discover the right stability between software independence and human-robot-interaction

Usually the human-robot-”interaction” is restricted to a human-robot connections through key pad and joystick .But this kind of client customer interface is not user-friendly for personal, does not operate the assisting capabilities of personal and robot ,especially the great short-term and spatial knowing capabilities of humans ; provides connections possibilities only from the person to the application. Another common user-robot client customer interface is an haptic client customer interface like a 6D Mouse with bilateral feedback. Thus the proprietor gets an information feed returning about the causes used from the environment to the manipulator. But in common it is not always possible for an proprietor to control a manipulator with a lot of stages of freedom with such devices. These methods are not appropriated for individuals who have no information about automated methods.

 In the higher part, these primary habits are mixed in a successive or contingency manner in order to apply more complicated habits known as capabilities. The capabilities designed allow the robot to identify an interested customer who wishes to communicate with the robot; (ii) to keep a record of the customer while he/she goes in the environment; and (iii) to follow the customer along the surroundings preventing possible hurdles in the way. The three capabilities represent a minimum set of capabilities useful for many cellular automatic programs that needs to communicate with human customers. Besides, the multi-agent approach employed allows the program to be quickly extended with further overall performance. The program has been analyzed in different real-life tests with several customers, accomplishing great results and real-time overall performance.



Works Cited

Arkin, Ronald C., and Tucker Balch. Cooperative Multiagent Robotic Systems. Georgia Institute of Technology. N.d. Web. 30 Nov. 2012. <           //>

Schwartz, Ariel.  ”Six Rescue Robots That Could Save Your Life”. Fastcoexist. Mansueto Ventures LLC. N.d. Web. 30 Nov. 2012. <>