Managing Coordination

    To MRT scientific achievements
    To MRT home page


SCARE (SCARE Coordinates Agents in Robotic Environments) implements a general architecture for coordination in multi-robot domains.

SCARE is able to deal with:

when a MAS is made up of agents with different skills, our architecture exploits these differences in order to improve the overall performance
coordination policy may change according to the amount of information that can be exchanged among agents and according to the network connectivity
in order to grant the autonomy of the system, the coordination mechanism is able to dynamically modify its parameters in reaction to environment changes
Using SCARE, the MAS application developer has to identify the macro-activities that the agents can carry out; we call jobs such macro-activities. Besides jobs, we have defined in the architecture other macro-activities named schemata. The activity assignement process is carried out by a special kind of agent called meta-agent, through two phases: decision making and coordination.


In the decision making phase, the meta-agent gathers up all the available information about the current situation, and evaluates the parameters associated to the activities. By employing multi-objective analysis techniques, the meta-agent obtains an ordered list of activities: the agenda.

In the coordination phase, the meta-agent searches for the best job allocation to agents by observing coordination constraints, such as cardinality (for each job the designer sets the minimum and maximum cardinality, i.e., the minimum and maximum number of agents that can be assigned to the job at the same time) and schema coverage (a schema can be assigned only if there is a suitable agent for each functionality).

Localization process

In this figure, on the left the zoom of one of the SCARE implementations shown on the right. A set of activities is considered and evaluated, and the agenda is proposed to the assignment module.

Once an agent is assigned to an activity, SCARE interacts with BRIAN sending to it coordination parameters, which influence the enabling conditions, and perceptual paramters, which are used by the parametric behavioral moduels.

Research contributors
A. Bonarini, F. Calvi, G. Fontana, M. Restelli.

Related Papers

  1. Bonarini, A., Restelli M. (2002) An Architecture to Implement Agents co-operating in Dynamic Environments Proceedings of AAMAS 2002 - Autonomous Agents and Multi-Agent Systems, ACM Press, New York, NY, 1143-1144.

  2. Bonarini, A., Restelli M. (2002) An architecture to implement adaptive cooperative strategies for heterogeneous agents Proceedings of IROS 2002 Workshop on Cooperative Robotics, IEEE Press, Piscataway, NJ, on CD.