The Group of Bidding Agents

In the model reported in this thesis we present a group of agents that take care of different tasks that, when coordinated through the bidding mechanism, provide the overall desired behavior of leading the robot to a target landmark. The tasks are:

• to keep the information on the map consistent and up-to-date,
• to keep the target located with minimum imprecision and move towards it,
• to keep the risk of losing the target low,
• to recover from blocked situations.

Four agents have been designed to fulfill each one of these goals (Map Manager, Target Tracker, Risk Manager and Rescuer, respectively), plus a communicator agent that is the responsible for communicating the Navigation system with the other robot systems (Pilot and Vision).

The actions that agents can bid for are:

• $\mathrm{Move}(\mathit{direction})$, instructs the Pilot system to move the robot in a particular direction,

• $\mathrm{Stop}$, instructs the Pilot system to stop the robot,

• $\mathrm{Look}(\mathit{angle})$, instructs the Vision system to identify all the possible landmarks that can be found in the area at $\mathit{angle}$ radians from the current body orientation.

Finally, agents may ask one another with respect to the different knowledge they have. For instance, any agent in the society may request from the Map Manager to compute the location of the target or of a diverting target. Agents may also broadcast messages to the rest of the agents in the society. For example, the Rescuer informs about the target to be reached, and the Target Tracker informs about the imprecision on the target's location.

In the next sections we describe each of the agents, and their code schemas can be found in Section 4.4.6.