S.A. Karpov - Post-graduate Student, Moscow State University of Information Technologies, Radio Engineering and Electronics (MIREA). E-mail: s_karpov@mirea.ru P.E. Tripolsky - Ph. D. (Eng.), Associate Professor, Moscow State University of Information Technologies, Radio Engineering and Electronics (MIREA). E-mail: tripolski@mirea.ru

This paper investigates the ways of motion planning for intelligent agents, combined into a single multi-agent robotic system. Motion planning should give the way to maintain the coherence of the system. We simulate performance of a multi-agent system, which consist of universal intelligent robots, number 5−100 units, in this paper. The environment of simulated MARS is presented as a 2D grid. Modeling communication between agents in this approach serves to produce using graphs in which peaks correspond to the robots, and the presence / absence of the edges there between indicates the presence / absence of connection between these agents. The task of Mars can be split into sub-tasks. These tasks can be delegated to a separate intelligent agents. We can calculate the op-timum alignment of agents for each step breaking the task at a certain stage. But the task of maintaining the coherence of the system requires us also to consider the possibility of «special situations». These situations include: «Tears» - a situation in which there is a relationship between two robots, but robots are moving in the opposite direction «Connection» - a situation in which there is no connection between the two robots, but robots are moving towards each other «overload» - a situation where passing through the unit delays occur because of the large number of messages «downtime» - a situation in which the unit does not pass through any communication, except posts this site Obviously, these situations require additional constructions. We know the graph describing the system and the connection adjacency matrix, defines it at each step. So we can determine the movement of agents and predict the availability of communication with each other just comparing the adjacency matrix at different steps of the task. The proposed method consists of three steps: analysis of the matrices and data verification identified specific situations, the decision to change the sequence of execution tasks. We can use these amendments during optimal arrangement formation. Determination of the optimal arrangement based on our research in other papers [1]. Thus, the sequence of action to maintain the coherence of the system should be as follows: 1) Delegation subtasks for intelligent agent 2) Analysis of the current arrangement, the search for «special situations», making appropriate adjustments 3) The calculation of the optimal placement agents 4) Issuance of targeting the tactical level. The result of described in this article algorithm will be an array of edges and vertices. We need add these vertexes to the graph of the system to make the network full mesh. And it is necessary to locate coordinates of the robots, which will form the best arrangement for this stage of the task MARS, to complete the algorithm. The resulting algorithms can be the foundation of software that allows you to simulate the behavior of multi-agent robotic system with a given number of robots. Software allows to evaluate the link at any time the performance of MARS tasks and to determine the coordinates of auxiliary robots to be added (or reassign available) to recover connectedness. We can use results of this research in modeling the behavior of real multi-agent systems, planning the movements and distribution of tasks, and incorporated in the related tasks, such as the task of assessing the effectiveness of the number of groups of agents in multi-agent robotic system. Experiments on this issue suggest the need for further researches of the behavior of multi-agent systems in an environment with ob-stacles and the development of algorithms to maintain connectivity in a negative radio environment, as well as algorithms for predicting the following system.

1. Karpov S.A., Tripolskijj P.EH. Modeli i algoritmy obespechenija svjazannosti agentov multiagentnojj robototekhnicheskojj sistemy v besprovodnojj seti // Uspekhi sovremennojj radioehlektroniki. № 3. S. 201−206.
2. Lokhin V.M., Manko S.V., Romanov M.P., Diane S.A.K., Karpov S.A., Tripolskijj P.EH. Modeli i algoritm ocenki chislennosti sostava multiagentnykh robototekhnicheskikh sistem // Mekhatronika, avtomatizacija i upravlenie. 2014. № 3. S. 20−23.
3. JohannesStrom, EdwinOlson.Multi-sensorATTenuationEstimation (MATTE): Signal-strengthpredictionforteamsofrobots. U.S. DoDGrantFA2386-11-1-4024. DepartmentofComputerScienceandEngineering. UniversityofMichigan. AnnArbor, MI 48109 USA.
4. Sheng W., Yang Q., Tan J., Xi N.Distributedmulti-robotcoordinationinareaexploration // RoboticsandAutonomousSystems. 2006. V. 54. P. 945−955.
5. Olfati-Saber R., Fax J.A., Murray R.M.Consensusandcooperationinnetworkedmulti-Agentsystems // ProceedingsoftheIEEE. 2007. V. 95. P. 215−233.
6. B.D.O. Anderson, Yu C., Fidan B., Hendrickx J.Rigidgraphcontrolarchitecturesforautonomousformations // IEEEControlSystemMagazine. 2008. V. 28. P. 48−63.
7. Karpov S.A. Programmnye i apparatnye sredstva besprovodnojj seti mobilnykh robotov na baze tekhnologii ANT. Diplomnyjj proekt. MIREHA. Moskva.2013.