Automatic code generation for new behaviors in a tool for set plays design

Abstract

Coordinated plans in Multi-Agent Systems (MASs) are essential for the success of the system. In MultiRobot Systems (MRSs), collective sports consist of an important testbed for MASs. In special, robotic soccer is a well-known challenge for researchers in MRSs. Set-plays are collective plays that correspond to coordinated plans in the robotic soccer domain. Some tools for designing new set-plays were developed. This article proposes a solution to support the implementation of new behaviors for simulation and execution of cooperative MASs. The manual development of these behaviors requires prior knowledge of the source codes of these tools, and a high development effort. Our solution uses Model-Driven Development (MDD), which different from a low-level manual programming paradigm, enables the definition of a behavior in high abstraction level and subsequent (semi) automatic transformation of it in the source code for specific tools. The solution uses templates from the target platforms, and automatically accesses tools of simulation and execution of a cooperative MAS, updating its files, inserting the codes of new behavior. Experiments in a real soccer team of robots demonstrated their feasibility in generating source code for the FCPortugal SPlanner and Framework Tools for creating set-plays in robotic