Simulators: RSoccer and Grsim

Utama‑Core includes a lightweight simulator adapted from RobôCIn’s rSoccer project. The simulator implements SSLStandardEnv and other environments for reinforcement learning. Key points:

• The environment is initialised with parameters such as field type (Division A/B/HW), number of robots per team, time step and optional starting formations. The observations include FrameData (positions of ball and robots) and RobotInfo lists for each team. Actions are 5‑dimensional vectors specifying translational velocity, angular velocity, kick power and whether the dribbler is active.
• The simulator inverts the y‑axis and orientation relative to grSim, so inputs and outputs must be transformed accordingly.
• The API exposes teleport functions to reposition the ball and robots, useful for resetting scenarios during training.
• Overlay drawing functions allow rendering points, lines and polygons for debugging and visualisation.

These environments enable reinforcement‑learning research and training for robot soccer skills. The rSoccer simulator is documented in a TDP published in the RoboCup 2021: Robot World Cup XXIV proceedings. The abstract notes that rSoccer provides an open‑source simulator for small‑size robot soccer, supports OpenAI Gym environments and benchmark tasks, and helps teams explore reinforcement‑learning approaches.