Simulador computacional para auxiliar no desenvolvimento de estratégias de comportamento para futebol de robôs

Abstract

The humanoid robot soccer competition that happens in RoboCup gathers knowledge from various fields of technology. For the development of high-level behavior strategies we can use a computer simulator to eliminate the need to have a constantly available robot for the work to evolve. Besides eliminating this need, the ideal scenario would be that the simulator could be used as a development environment. This could be done by creating abstractions to the user so that the code developed in the simulator can be ported to the real robot transparently. The problem with this idea is the lack of a computer simulator applied to robotics that keeps both a high level of abstraction and a hardware compatibility commitment. On the one hand there are powerful simulators but with a low level of abstraction and on the other there are soccer simulators that simplify the problem by creating a good level of abstraction but are not related to robotics. The remaining question is: can we have a computer simulator applied to robotics that at the same time has a high level of abstraction and maintains hardware compatibility? To achieve this goal, we need to adopt a standard data communication between sensors and actuators and the agent who owns them. This paper suggests using the JSON format. JSON provides a generic representation of objects that is at the same time simple, compact and comprehensive. We can simplify the necessary information about the objects perceived in the environment and the commands to be passed to the actuators. With this goal in mind we created a 2D computer simulator using python as programming language and pygame as graphic library. The level of abstraction was achieved successfully making it possible for the behavior strategy programmer to keep focus on the logic of his algorithm, abstracting how the robot works. The simulator also provides an environment for creating and testing behavior strategies, but is pending on the verification of code portability. Nevertheless, it is easy to see that we can adapt the robot code so that communication with the behavior strategy is made following the standards specified in this paper, thus making it possible for the code created in the simulator to be ported seamlessly to the robot.