Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Improving Dribbling, Passing, and Marking Actions in Soccer Simulation 2D Games Using Machine Learning

Nader Zare, Omid Amini, Aref Sayareh, Mahtab Sarvmaili, Arad Firouzkouhi, Stan Matwin, Amilcar Soares

TL;DR

This study advances RoboCup 2D Soccer Simulation by introducing three core ML-driven capabilities—Multi-Action Dribble (MAD), Pass Prediction, and an optimized Marking Decision framework (OMAM)—to enhance dribbling safety, teammate passing coordination, and defensive synchronization. MAD integrates a DNN-predicted opponent movement within a chain-action dribble strategy, plus a MAD generator to disrupt opponents; combined with a DNN-based position predictor, it yields notable win-rate gains against strong opponents. Pass Prediction leverages a full-state predictor trained on noisy observations and a large engineered feature set to forecast the ball holder's actions, achieving up to around $80\%$ accuracy in clean data and maintaining substantial predictive power under noise. OMAM improves multi-agent marking by pruning candidate solutions and structuring defender roles, reducing goals conceded from $0.9$ to $0.33$ per main round between 2019 and 2021. Collectively, these contributions demonstrate that integrated ML modules can significantly elevate both offensive cooperation and defensive coordination in stochastic, partially observable soccer environments, with practical impact on competition performance and open-source ML toolchains for the SS2D community.

Abstract

The RoboCup competition was started in 1997, and is known as the oldest RoboCup league. The RoboCup 2D Soccer Simulation League is a stochastic, partially observable soccer environment in which 24 autonomous agents play on two opposing teams. In this paper, we detail the main strategies and functionalities of CYRUS, the RoboCup 2021 2D Soccer Simulation League champions. The new functionalities presented and discussed in this work are (i) Multi Action Dribble, (ii) Pass Prediction and (iii) Marking Decision. The Multi Action Dribbling strategy enabled CYRUS to succeed more often and to be safer when dribbling actions were performed during a game. The Pass Prediction enhanced our gameplay by predicting our teammate's passing behavior, anticipating and making our agents collaborate better towards scoring goals. Finally, the Marking Decision addressed the multi-agent matching problem to improve CYRUS defensive strategy by finding an optimal solution to mark opponents' players.

Improving Dribbling, Passing, and Marking Actions in Soccer Simulation 2D Games Using Machine Learning

TL;DR

This study advances RoboCup 2D Soccer Simulation by introducing three core ML-driven capabilities—Multi-Action Dribble (MAD), Pass Prediction, and an optimized Marking Decision framework (OMAM)—to enhance dribbling safety, teammate passing coordination, and defensive synchronization. MAD integrates a DNN-predicted opponent movement within a chain-action dribble strategy, plus a MAD generator to disrupt opponents; combined with a DNN-based position predictor, it yields notable win-rate gains against strong opponents. Pass Prediction leverages a full-state predictor trained on noisy observations and a large engineered feature set to forecast the ball holder's actions, achieving up to around accuracy in clean data and maintaining substantial predictive power under noise. OMAM improves multi-agent marking by pruning candidate solutions and structuring defender roles, reducing goals conceded from to per main round between 2019 and 2021. Collectively, these contributions demonstrate that integrated ML modules can significantly elevate both offensive cooperation and defensive coordination in stochastic, partially observable soccer environments, with practical impact on competition performance and open-source ML toolchains for the SS2D community.

Abstract

The RoboCup competition was started in 1997, and is known as the oldest RoboCup league. The RoboCup 2D Soccer Simulation League is a stochastic, partially observable soccer environment in which 24 autonomous agents play on two opposing teams. In this paper, we detail the main strategies and functionalities of CYRUS, the RoboCup 2021 2D Soccer Simulation League champions. The new functionalities presented and discussed in this work are (i) Multi Action Dribble, (ii) Pass Prediction and (iii) Marking Decision. The Multi Action Dribbling strategy enabled CYRUS to succeed more often and to be safer when dribbling actions were performed during a game. The Pass Prediction enhanced our gameplay by predicting our teammate's passing behavior, anticipating and making our agents collaborate better towards scoring goals. Finally, the Marking Decision addressed the multi-agent matching problem to improve CYRUS defensive strategy by finding an optimal solution to mark opponents' players.
Paper Structure (23 sections, 1 equation, 3 figures, 2 tables)

This paper contains 23 sections, 1 equation, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Previous Works
  3. Release
  4. Cyrus 2014 Source.
  5. CppDNN.
  6. Pyrus - Python 2D Soccer Simulation Base.
  7. Multi-Action Dribble
  8. Chain Action algorithm
  9. Agent 2D Dribble Action Generator
  10. MAD Generator
  11. Opponent movement prediction
  12. Results
  13. Pass Prediction
  14. Experimental Setup
  15. Result
  16. ...and 8 more sections

Figures (3)

  • Figure 1: Effecting pos-count on dribbling. Blue dots are possible candidates, red dots are candidates removed since they are impossible to be reach, and yellow dots are the possible candidates which are removed incorrectly because of the pos-count.
  • Figure 2: Types of MAD and basic dribble, blue dots stands for possible actions generated by Basic Dribble, red dots show impossible dribbles, and orange dots demonstrate possible actions that have been added after using MAD.
  • Figure 3: Improper defense strategy, leaving the leftmost player empty for a longer time, while T1 should mark O1 and T2 should mark O2 for an optimal solution.