Implementing Lego Agents Using Jason

TL;DR

This paper presents a method to integrate the Jason multi-agent framework with Lego robots, enabling simple robots to perform complex multi-agent tasks with minimal coding, despite some delays in sensor processing.

Contribution

It introduces a novel extension of Jason to control Lego robots, allowing for multi-agent system implementation with physical hardware.

Findings

Robots can follow lines and communicate to avoid obstacles

The system is robust for simple multi-agent tasks

Sensor-dependent agents experience delays in reasoning cycle

Abstract

Since many of the currently available multi-agent frameworks are generally mostly intended for research, it can be difficult to built multi-agent systems using physical robots. In this report I describe a way to combine the multi-agent framework Jason, an extended version of the agent-oriented programming language AgentSpeak, with Lego robots to address this problem. By extending parts of the Jason reasoning cycle I show how Lego robots are able to complete tasks such as following lines on a floor and communicating to be able to avoid obstacles with minimal amount of coding. The final implementation is a functional extension that is able to built multi-agent systems using Lego agents, however there are some issues that have not been addressed. If the agents are highly dependent on percepts from their sensors, they are required to move quite slowly, because there currently is a high delay in the reasoning cycle, when it is combined with a robot. Overall the system is quite robust and can be used to make simple Lego robots perform tasks of an advanced agent in a multi-agent environment.