Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Agent Coordination for a Partially Observable and Dynamic Robot Soccer Environment with Limited Communication

Daniele Affinita, Flavio Volpi, Valerio Spagnoli, Vincenzo Suriani, Daniele Nardi, Domenico D. Bloisi

TL;DR

Addressing coordination of fully autonomous humanoid robots in RoboCup SPL under limited communication, the paper presents a market-based distributed coordination framework with a distributed world model (DWM), distributed task assignment (DTA), and Voronoi-guided positioning. The approach combines local world modeling, context-driven task selection, and auction-like utilities to assign tasks without broad network exchange, using predictive models (Gaussian Mixture Model for obstacles, Kalman filter for the ball) to bridge data gaps. Experiments in real RoboCup matches and in SimRobot show reduced role overlaps and improved adaptability in low-rate communication environments. This work advances robust multi-agent coordination toward RoboCup's 2050 goals by enabling decentralized, anticipatory decision-making under partial observability.

Abstract

RoboCup represents an International testbed for advancing research in AI and robotics, focusing on a definite goal: developing a robot team that can win against the human world soccer champion team by the year 2050. To achieve this goal, autonomous humanoid robots' coordination is crucial. This paper explores novel solutions within the RoboCup Standard Platform League (SPL), where a reduction in WiFi communication is imperative, leading to the development of new coordination paradigms. The SPL has experienced a substantial decrease in network packet rate, compelling the need for advanced coordination architectures to maintain optimal team functionality in dynamic environments. Inspired by market-based task assignment, we introduce a novel distributed coordination system to orchestrate autonomous robots' actions efficiently in low communication scenarios. This approach has been tested with NAO robots during official RoboCup competitions and in the SimRobot simulator, demonstrating a notable reduction in task overlaps in limited communication settings.

Multi-Agent Coordination for a Partially Observable and Dynamic Robot Soccer Environment with Limited Communication

TL;DR

Addressing coordination of fully autonomous humanoid robots in RoboCup SPL under limited communication, the paper presents a market-based distributed coordination framework with a distributed world model (DWM), distributed task assignment (DTA), and Voronoi-guided positioning. The approach combines local world modeling, context-driven task selection, and auction-like utilities to assign tasks without broad network exchange, using predictive models (Gaussian Mixture Model for obstacles, Kalman filter for the ball) to bridge data gaps. Experiments in real RoboCup matches and in SimRobot show reduced role overlaps and improved adaptability in low-rate communication environments. This work advances robust multi-agent coordination toward RoboCup's 2050 goals by enabling decentralized, anticipatory decision-making under partial observability.

Abstract

RoboCup represents an International testbed for advancing research in AI and robotics, focusing on a definite goal: developing a robot team that can win against the human world soccer champion team by the year 2050. To achieve this goal, autonomous humanoid robots' coordination is crucial. This paper explores novel solutions within the RoboCup Standard Platform League (SPL), where a reduction in WiFi communication is imperative, leading to the development of new coordination paradigms. The SPL has experienced a substantial decrease in network packet rate, compelling the need for advanced coordination architectures to maintain optimal team functionality in dynamic environments. Inspired by market-based task assignment, we introduce a novel distributed coordination system to orchestrate autonomous robots' actions efficiently in low communication scenarios. This approach has been tested with NAO robots during official RoboCup competitions and in the SimRobot simulator, demonstrating a notable reduction in task overlaps in limited communication settings.
Paper Structure (8 sections, 7 equations, 4 figures)

This paper contains 8 sections, 7 equations, 4 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Method
  4. Distributed World Model
  5. Distributed Task Assignment
  6. Voronoi Diagram
  7. Experimental Results
  8. Conclusions

Figures (4)

  • Figure 1: Frame from the quarter-finals of the RoboCup SPL 2023 between SPQR and HTWK teams. Above the names of the teams, the counters of the exchanged packets for each team are shown.
  • Figure 2: The overall architecture of DWM and DTA. The input is represented by a network event. If the event does not occur, prediction models probabilistically extend the previously estimated models. Then all local models are merged into DWM, used to select the most valuable context and assign utility values to each <robot, task> pair. Finally, the optimal configuration V is employed to match the number of roles to the number of available robots. Roles are then assigned to maximize cumulative utilities.
  • Figure 3: Voronoi graph in 2D and 3D field view. In the 2D view (left), blue points represent the opponent robots and black connections depict the Delaunay Triangulation, while red points are the Voronoi nodes and red links are the Voronoi edges. In the 3D view (right), just the Voronoi nodes and edges are shown.
  • Figure 4: Role overlaps over time: for each role, the cumulative time (minutes) of role overlaps is shown. This demonstrates the improvements of the proposed approach (green) w.r.t. the baseline (blue).