EditorialStrategy planner: Graphical definition of soccer set-plays
Introduction
Artificial Intelligence and Robotics have been two areas of research which have received a great deal of attention over the past few years.
These areas of research have been fostered particularly by international initiatives like RoboCup which accommodates many challenging competitions. From these competitions the one with the most fans is undoubtedly the soccer competition due to its wide acceptance over the world. This competition places two teams of robotic agents up against each other to dispute victory in a soccer match. Teams have been improving performance by creating new strategies that currently consider the definition of strategic positioning [1], [2], [3], [4], [5], [6] based on formations, tactics and set-plays [7], [8], [9].
A set-play can be part of a team's strategy and is a widely known concept in real soccer as well as in other cooperative sports to leverage a competitive advantage against an opposing team. A set-play can be described as a structured plan that describes courses of actions that a subset of players in a team should take based on the current state of a game. Some attempts to make use of set-plays have already been made in the robotic soccer domain, however the knowledge for their definition and execution is tightly coupled (hard-coded) with the soccer player agent internal implementation. A framework that promotes the decoupling of the knowledge of set-plays from the soccer player agents internal implementation using a s-expression language has recently been developed [9]. However, writing set-play definitions manually is a harsh, error-prone and time consuming process. For these reasons, a graphical tool, named Strategy Planner (SPlanner), is proposed to speed-up the definition of set-plays and reduce the amount of errors committed by abstracting the complexity of the grammar from the end users.
The rest of this article is organized in the following manner. Section 2 describes some of the related work done in the context of strategy definition, with a particular emphasis on the definition of set-plays. Section 3 describes the functionality and some usage examples of the Set-play framework used as the basis of this work. Section 4 presents the developed graphical user interface (GUI) of the SPlanner tool, focusing on its integration with the Set-play framework. Section 6 describes the methodology used to perform experiments in order to validate the usefulness of the developed tool. Section 7 presents an analysis of the results obtained from the experiments performed to assess the usefulness of SPlanner. Section 8 draws the main conclusions from the developed work and establishes some pointers for future work.
Section snippets
Related work
The general concept of strategy can be described as a previously planned and typically complex behavior whose goal is to make use of available resources in the most efficient and effective way [10]. The concept of strategy has been widely adopted in several domains and its definition has evolved to match the specificities of each domain. In collective sports, particularly in the soccer domain, the concept of strategy has been the main driver for the improvement of the game quality of teams over
Set-play framework
The set-play framework [9] provides a language specification for defining plans (set-plays) for the soccer domain, a built-in parser and an engine that allows them to be interpreted and executed at run-time.
Set-plays can be particularly useful in some situations (better exploit empty spaces in the opponent's goal area) to help the team achieve a competitive edge. Moreover, the continuous improvement of soccer agents tactics and skills requires the development of new strategies to counter them
Graphical definition of set-plays
The SPlanner tool was developed in C++ for the Linux platform and makes use of Qt1 graphical libraries to ease the integration with the majority of tools produced by the RoboCup Soccer community. An overview of the SPlanner tool architecture is presented in Fig. 1.
The development of the tool was carried out with modularity in mind, in order to easily allow the future integration of new
Promoting the usability of the tool
Several precautions were taken to promote a good usability of the tool. The process of designing the GUI was guided by the set of well known heuristics [15] described in Section 2.
To avoid unnecessary interactions for expert-users some keyboard accelerators were defined in the tool to speed-up the definition of actions for the selected player. For instance, if the user presses the letter P in the workspace when a player that owns the ball is selected, a pass action is initiated from that player
Tool validation methodology
In order to validate the developed tool four experiments were performed.
The first experiment consisted on testing the correctness and robustness of the SPlanner import process and consequently the visual representation of previously defined set-plays. The set-plays used in this test have been previously exported from SPlanner or manually defined by FC Portugal team members using a text editor.
The second experiment consisted on measuring the effectiveness of two of the previously created
Robustness of the set-plays import process
In the first experiment, the import process of all the set-plays that had been previously developed in SPlanner was always executed without any errors. When trying to import some of the set-plays defined manually by FC Portugal team members some exceptions occurred. The cause of these exceptions was identified as the definition of multiple actions for a player in a step transition in these set-plays. As previously mentioned in Section 5, some measures were implemented in SPlanner to simplify
Conclusions
The interaction with SPlanner was not equal for all test users but, as demonstrated by the results, users were greatly satisfied with its use having ranked it with an average score of 77 (out of 100). The use of this tool also allowed typical users (FC Portugal team members) to significantly reduce the time required (up to 90% on average) to perform set-play related tasks. Furthermore, users with little knowledge of the domain were able to use the tool with relative ease thus widening the range
João Cravo holds an MSc in Informatics Engineering and Computation from the Faculty of Engineering of the University of Porto. He is currently a working at Blip.pt, a software company located in Porto, Portugal, and as a Build Master. His other interests include (but are not limited to) Artificial Intelligence, Software Engineering, Graphical User Interface Design and Robotic Soccer.
References (29)
- et al.
Multi-robot coordination using setplays in the middle-size and simulation leagues
Mechatronics
(2011) - et al.
Situation based strategic positioning for coordinating a team of homogeneous agents
- et al.
Dynamic Positioning based on Voronoi Cells (DPVC)
- et al.
Multi-agent positioning mechanism in the dynamic environment
- et al.
Dynamic positioning method based on dominant region diagram to realize successful cooperative play
- et al.
Pareto-optimal offensive player positioning in simulated soccer
- et al.
Pareto-optimal collaborative defensive player positioning in simulated soccer
- et al.
Scenario-based teamworking, how to learn, create, and teach complex plans?
- et al.
Flexible coordination of multiagent team behavior using HTN planning
- et al.
Co-ordination in RoboCup's 2D simulation league: setplays as flexible, multi-robot plans
What is Strategy? Definition and Meaning
A survey on coordination methodologies for simulated robotic soccer teams
Coaching: Learning and using Environment and Agent Models for Advice
COACH UNILANG — a standard language for coaching a (robo)soccer team
Cited by (12)
BahiaRT Setplays Collecting Toolkit and BahiaRT Gym
2022, Software ImpactsCentralized Robot Soccer Architecture Based on Roles
2016, RIAI - Revista Iberoamericana de Automatica e Informatica IndustrialTowards Automatic Code Generation for Robotic Soccer Behavior Simulation
2024, Journal of Intelligent and Robotic Systems: Theory and ApplicationsFiltering active moments in basketball games using data from players tracking systems
2023, Annals of Operations ResearchGenerating a dataset for learning setplays from demonstration
2021, SN Applied Sciences
João Cravo holds an MSc in Informatics Engineering and Computation from the Faculty of Engineering of the University of Porto. He is currently a working at Blip.pt, a software company located in Porto, Portugal, and as a Build Master. His other interests include (but are not limited to) Artificial Intelligence, Software Engineering, Graphical User Interface Design and Robotic Soccer.
Fernando Almeida is a Lecturer at the Polytechnic Institute of Viseu, Portugal. He is currently taking his PhD which focuses on Automatic Plan Extraction, Recognition and Optimization from collective sports games, with a particular emphasis in soccer. His other interests include (but are not limited to) Artificial Intelligence, Autonomous Agents, Multi-Agent Systems (MAS), Coordination in MAS, Automated Reasoning and Inference, Game Analysis and Robotic Soccer.
Pedro Henriques Abreu is an Assistant Professor at Coimbra University, Portugal. He obtained his PhD in soccer teams' Modeling from the University of Porto (2011). His interests include (but are not limited to) Artificial Intelligence, automatically analysis in Game Analysis, Tactical Modeling, and Data Mining Techniques Applied to sport collective games.
Luis Paulo Reis is an Associate Professor at the University of Minho, member of the Directive Board of LIACC Laboratory and coordinator of the Human–Machine Intelligent Cooperation Research Group in Portugal.
He was principal investigator of more than 10 research projects in the areas of Artificial Intelligence and Robotics including FC Portugal, three times World Champion and eight times European Champion at RoboCup. He also won more than 30 other scientific awards. He supervised 13 PhD theses and 80 MSc theses to completion and is the author of more than 250 publications in international conferences and journals. He is the president of the Portuguese Robotics Society.
Nuno Lau is an Assistant Professor at Aveiro University, Portugal.
He got is Electrical Engineering Degree from Oporto University in 1993, a DEA degree from Claude Bernard Lyon 1 University in 1994 and the PhD from Aveiro University in 2003.
His research interests include Intelligent Robotics, Artificial Intelligence, Multi-Agent Systems and Simulation. He has lectured courses at Phd and MSc levels on Distributed Artificial Intelligence, Intelligent Robotics, Computer Architecture, Programming, etc.
Nuno Lau has participated, often with the coordination role, in several research projects that have been awarded international prizes. Nuno Lau is the author of more than one hundred publications in international conferences and journals.
Luís Mota is an Auxiliary Professor at ISCTE — IUL, a university in Lisbon. He got his BSc and MSc from Instituto Supeior Técnico (IST, Lisbon) and his PhD from the University of Porto. He has been researching mainly in the MAS and Robotic Soccer areas.