PELEA in PlanInteraction

The Control module contains a Session Controller agent (SC) that will perform the tasks of registration of agents, set the clock of the simulation and manage the information of the different ongoing planning domains and the system finalization conditions. We assume that each pair of problem-domain is a different session, so agents belonging to a session could not interact with those from other session. The simulation module contains a Simulation Agent (SA) for each running session.

 

We inherit four of the PELEA main modules for high level planning and execution: EXECUTING (EX), MONITORING (MO), DECISION SUPPORT (DS) and HIGH LEVEL REPLANNER (HLR).

An agent can include any combination and number of these modules but we considered two kind of agents, an Execution Agent (EA) and a Planning Agent (PA).

An EA is an agent with EX and MO capabilities.
A PA is an agent with DS and HLR capabilities.

 

 

The architecture allows different kinds of configurations according to the number of agents present in the system.

The simplest configuration consist of an EA, a PA and a SA, all of them in the same session. A simulation has the next steps:

  1. 1.- the EA process his domain and problem files.
  2. 2.- he registers with the SC.
  3. 3.- asks the SA for the initial state.
  4. 4.- asks the PA for a plan providing the plan, problem and state.
  5. 5.- executes the plan actions one by one checking the resulting states.
  6. 6.- when goals are achieved, the EA informs the SC and waits.
  7. 7.- the SC can send a finalization message or assign a new goal to the EA.

 

In a real execution, with robots, the SA is not needed, so the EA must be able to communicate with the physical actuator, to get the state of the world and to provide the actions to perform.

In a problem we usually find more than one actuator, so the architecture allow other configurations for the number of agents. A typical configuration is a set of n EA agents joining the same session with just one PA. In this approach the PA generates a coordinated plan and distributes the corresponding actions to each EA.