Abstract
Purpose
The purpose of this paper is to establish a situation evaluation model of the robot and ball in SimuroSot5vs5 platforms and enhance the strength of the team in a SimuroSot5vs5.
Design/methodology/approach
This paper presents a mathematical model based on situation evaluations which can improve the strength of the team in SimuroSot5vs5. The situation evaluation focus on four aspects includes scores of both sides, possession of teams on ground, ball strategy, and treat. The statistical analysis of the score can verify validity and stability of current strategy in confrontation. To evaluate the situation more effectively without blindness, possession on both teams is, respectively, evaluated. Ball strategy analyzes coordinate transformation to the ball on the ground and illustrates the circumstance of both teams on the offensive position accurately in length and breadth. To know the circumstance on the field more completely and synthetically, a threat situation evaluation model is built. An effective and practical algorithm for comprehensive situation evaluation is successfully finished. The experiments prove validity and performance of the proposed situation evaluation model.
Findings
A mathematical model is designed to achieve situation evaluation, and the strategy can change in accordance with different situations on the ground.
Research limitations/implications
The system is specifically applied to SimuroSot5vs5 platform. The extensibility of the system is limited.
Practical implications
When the robot and ball is in high speed movement, a large calculated amount will slow the speed of the system.
Originality/value
The paper shows that situation evaluation in SimuroSot decision support systems will enhance the battle effectiveness of the soccer team.
Keywords
Citation
Hao‐bin, S., Zhu‐jun, Y., You‐feng, X. and Wei‐hua, L. (2012), "The study of situation evaluation in SimuroSot decision support systems", Kybernetes, Vol. 41 No. 9, pp. 1226-1234. https://doi.org/10.1108/03684921211275243
Publisher
:Emerald Group Publishing Limited
Copyright © 2012, Emerald Group Publishing Limited
1 Introduction
With the development of technology of Robot Soccer and daily perfection of the robot's motion design, research on the robot soccer diverts to achieve and improve the strategy design gradually, which is especially obvious at the 11th anniversary FIRA RoboWorld Cup. However, most teams adopt single strategy in the whole match at present. Obviously, the condition changes each time. With development of robot soccer, if the strategy cannot change in accordance with different conditions, the monotonous design (Park et al., 2001) will be eliminated inevitably.
In SimuroSot (Kong et al., 2007) diversity of strategy and high motion speed of both robot and ball, lead to the fact that the situation changes each time. So situation evaluation (for a different but relevant discussion on situation evaluation, see Liu et al. (2002), Lin et al. (2004) and Lin and Liu (2006)) of the robot and ball on the ground is necessary to render the robot players to finish the pre‐set action. Suppose a ball is moving fast toward the goal and a football robot is coming from the other direction to defend. The defender could take two actions at this moment (Li et al., 2001). One is to kick the ball to the opponent's goal straightly; another is to kick the ball to the sideline. Compared to the two strategies above, adopting a right strategy can make the robot kick the ball more effectively. We will choose the first way to change defense to attack when there are few opponents; on the contrary, the ball would be intercepted if we take the same action; we may take the second method before our player comes back. Obviously, the situation evaluation could used to evaluate the robot's choice in actions when it is under different conditions. This paper discusses the design and implementation of situation evaluation algorithms in SimuroSot5vs5 decision support systems[1].
2 Factors that affect situation assessment in simurosot5v5
The match platform adopted in SimuroSot5vs5 was developed by Australia (Wu et al., 2001). In order to simply analyze, we take ground center as the original point of the ground coordinate. The coordinate collected by the platform could transform to the corresponding ground coordinate. Suppose (a, b) is the coordinate collected by system platform, (X, Y) refers to the ground coordinate after calculating. With accurate test of the platform, transfer function could be defined as formula (1) (Wilsker, 2003; Chen et al., 2007; Kong and Zou, 2006): Equation 1
Based on this mechanism, the following four aspects could take field situation evaluation references.
2.1 Scores of both sides
Set the system and collect information once every 50 periods. The system runs every 0.016 s in one period. T is an evaluation period. The times of collect information in T period is n (T/(0.016*50) = n). It defines two mathematical variables, jg and sg, which refers to the scores of both sides at each information acquisition time, respectively. The following mathematical relation is designed to reflect the variation of the score. As formula (2): Equation 2 where s is the DIFF (difference in score), sk is the DIFF at information acquisition time k in an evaluation cycle. It could reflect the score in one evaluation cycle. It also could verify the validity and the stability of the current strategy in competition by statistic analysis (e.g. expectation and variance of s) of variable s in one cycle.
2.2 Control rate of the ball
Formulas (3) and (4) are used to judge whether a player is controlling the ball: Equation 3 Equation 4 where (xi,yi) is the coordinate of the robot i, (xq, yq) is the coordinate of the ball. R is the critical value of the distance between a player and ball, which is set according to common training. If the formula f is true this means that the player i is controlling the ball at the moment, adversely, player i is not controlling the ball.
Formulas (5)‐(7) present the control rate of the ball of the main player, a team and the ratio between the two above, respectively. From the three aspects, it can analyze the control rate of the ball effectively: Equation 5 Equation 6 Equation 7 In formulas (5)‐(7), where (x0, y0) is the coordinate of the main player, ∑j=1nf(x0,xq,y0,yq) is the sum total of the main player's control rate of the ball in an evaluation cycle (T), ∑i=15fm(xi,xq,yi,yq) and ∑i=15fe(xi,xq,yi,yq) present the number of players in our team and opponent who possesses the information acquisition time, respectively. ∑j=1n(∑i=15fm(xi,xq,yi,yq)+∑i=15fe(xi,xq,yi,yq)) is the sum total of the control rate of the ball of all players on the ground in time period T. u0 is the control rate of the ball of our main player in time period T. u is the control rate of the ball of our team in time period T. ua is the rate between control rate of the ball of the main player and our team. With these formulas, stability of current strategy could evaluate by offending and defending capability of the whole team. After all, in either the human soccer or the robot soccer, coordination of the whole team is one of the most important factors. So, the set of the control rate of the ball is done in order to evaluate the situation more effectively without blindness. The range of values ua∈(0.25, 0.45) and uo∈(0.5, 1) is set according to normal experience. We could recognize that the control rate of the ball is reasonable in this range.
2.3 Ball strategy
Ball strategy is the situation evaluation of the coordinate transformation to the ball on the ground. It could analyze with four conditions as follows:
- 1.
Ball on half field either, it is illustrated as follows: Equation 8 where gt is the sum of the coordinate of the ball on the field (only considering the x coordinate because of the players in half field at a moment, considering the y coordinate has no sense). This formula reflects the condition of the ball on half field. It is an important index that measures the drive capability of a team to the ball.
- 2.
For the situation when a team is on opponent half field, the mathematical definition is expressed as: Equation 9 In formula (9), where atm,ate present the sum coordinates of all players in a team when they are within the opponent half field (atm,ate present our team and the opponent team, respectively.). It called threat degree to the goal. If in the formula atm<0, it means that the opponent is in passive state and combat is concentrated on the opponent field; on the other hand, if the formula ate>0 then it means that our team is in a passive state and combat is concentrated on our field. Combat between two teams on the field is reflected with this group mathematical relation above.
- 3.
The threat degree of a team to the opponent forbidden zone, the mathematical definition is expressed as follows: Equation 10 where l is the transverse length of the field. If in the formula ate>l · n, this means that our goal is threatened by shooting by the opponent. On the contrary, the formula atm>l · n means that our team has a chance to shoot at the opponent goal.
- 4.
The offense and defense is concentrated on the central or the sideline, the mathematical definition is expressed as follows: Equation 11 Equation 12
In short, with the statistic analyses of these variables atm, ate, am, and ae, it could illustrate the circumstances of both teams on the offensive position accurately in length and breadth.
2.4 Threat
These parameters given above discuss several profiles dispersedly. In order to know the circumstance on the field more completely and synthetically, we build up a preliminary mathematical model as shown in Figure 1.
In this model, we calculate the sum of the players' abscissa in a team and absolute value of the ordinate's sum, and take the two values as the right angle legs. This method could be defined as trigonometry. Hypotenuse length is φ(∑xi,|∑yi|) and the sine value is sin γ=(|∑yi|/∑xi). The two parameters are very useful references to integrated evaluation of the whole team's ability. With the trigonometric function, it could know that the general position of the whole team. In order to illustrate the essential function of the trigonometry better, some cases are briefly explained as shown in Figure 2:
- •
Case A. The team is in an offensive state which could be seen from the sine value, the angle γ≥135°, the offensive concentrated on central line. Judge from the length of the hypotenuse, the threat degree of the opponent goal is greater than ours.
- •
Case B. The angle 90°≤γ<135°. Although our team keeps ahead, the threat degree of the opponent goal is not enough and there is not enough chance to score.
- •
Case C. The angle 45°≤γ<90°, the combat is concentrated on our field, but it is unlikely for the opponents to make a breakthrough to the goal at a short time.
- •
Case D. The angle γ<45°, offense is within the forbidden zone, the opponent will score at anytime.
3 Design and implementation of synthetically situation evaluation module
In short, the synthetically situation evaluation is based on four distinctive parameters given above, combining them in design of the algorithm program to get a final reference value, which can be taken as a final reliance of the situation evaluation.
The chief reference of the situation evaluation is “DIFF”. When the score falls behind, some corresponding offensive strategy will be adopted. The second important reference is the control rate of the ball, which will be evaluated when a new strategy starts being used. Obviously, it is meaningless to compare the control rate of the balls between different strategies. At the micro level, ball strategy and threat could be used as a guide in a small range of a specific period. Two modules have been designed to achieve situation evaluation and support the function of the whole situation evaluation. The two modules could implement the situation evaluation algorithm in SimuroSot5vs5.
Module 1. T is an evaluation cycle. Apply modules implements once in every T period. We first arrange the priority level of the evaluation parameter. We have definitions as follows. First, the score is the premier reference factor. If the score manifests that our team fall behind, we must take the offensive strategy and sign it with red. Second, the condition of the control rate of the ball and the ball in field could be divided into four cases:
- 1.
low control rate of the ball and longer time the ball in our field, it must take positively defensive strategy and the formation should take corresponding adjustment;
- 2.
low control rate of the ball and longer time the ball in opponent field, it could take intensively offensive strategy; and
- 3.
high control rate of the ball and longer time the ball in our field, then intensive offensive strategy;
- 4.
high control rate of the ball and longer time the ball in opponent field, then keep the present strategy continuously.
In the module, we input the coordinates of the ball and the players, then get the current class by data preprocessing of coordinate and return it to the evaluation system. Pseudo code is given as follows:
// module of the class evaluation function
int SituJud_1()// set the external variable, directly call, accomplish the calculation of the coordinate above
{
if(GoalKickJud(int score))
//alternative statement, return value per count cycle, judge whether its class is Red which is the most dangerous state, if the //conditions given above unsatisfied, then enter the follow judgment.
else{
//set several conditional statements, judge the current //situational class with the input value, then return with SituJud_1()
}
}//SituJud_1
Module 2. The main module's design is for the aid that we can call different strategy at specific class. According to the class given in module 1, call the corresponding strategy in strategy warehouse. The corresponding pseudo code is shown as follows:
// module of strategy transformation function
void SituJud_2()
{
if(SituJud_1()== Red)
{//The census of the score indicates that the current strategy is in passive state completely, and the score fall behind, we //should take the offensive formation 3‐1‐1
StraSele();
// It unnecessary to consider about the problems of the role's assignment and selection of the offensive area temporarily as //the strategy has been modified
}
if(SituJud_1()== ORANGE)
{ // judge the offensive area of opponent
AttAre();
// intensify the defense in the right area, select the conservatively defensive formation 1‐3‐1
StraSele();
}
if(SituJud_1()== YELLOW)
{ // judge the offensive area of our team
AttAre();
//Call the substrative strategy function (casting system). Since the current feedback, our team has some prevalence, but //the control rate of the ball is not high. It is illustrate that some risk exist in the strategy and it require the relevant casting //system to accomplish.
SituJud_3();
}
if(SituJud_1()== GREEN)
{ // judge the offensive area of opponent
AttAre();
//The time on our field is longer but our players have strong ability to control the ball and the ossession is high, the //opponent's offense have no threat to us.
SituJud_3();
}
If(SituJud_1()== BLUE)
{//our team is in advantageous position,take no action}
}
With the two modules above, the design and implementation of situation evaluation algorithm in SimuroSot5vs5 DSS is fulfilled essentially.
4 Result analysis
Test platform: SimuroSot5vs5 developed by Australia. Test method: use the version DEMO developed by Wuhan Institute of Technology, our old version, and our new version with the addition of situation evaluation. Our new version and old version each competed with the version DEMO 40 times. In addition the match played in accordance with the normal match rule, Test results of the match are shown as Table I.
As shown in Table I, the fighting power of the team has been visibly strengthened after the situation evaluation added. Either the offensive or the defensive owns quite high performance, the score rate increased by 75.6 percent, while the loss rate decreased by 32.8 percent. Moreover, the control rate of the ball increased from 45.0 percent to 63.0 percent. These statistics fully illustrate that the technology of our team has made great improvement. The situation evaluation is of great importance in decision support systems.
In addition to normal tests, our team made excellent achievements in domestic and abroad matches with the situation evaluation used in decision system. Also, the version won many famous teams frequently. The rationality of the situation evaluation algorithm used in DSS has been verified again. The result also indicates that the algorithm can be widely applied and is of reference significance.
5 Conclusion
This paper presents and analyses the mathematical model based on situation evaluation of the robot and ball on the SimuroSot5vs5 platform. Then, the author puts forth a new mathematical model and practical mathematical algorithm. As a result of experiments, the method proves both practical and effective, especially for its advantage in the aspect of ball control rate of the ball. We can tell that the application of the situation evaluation will influence and improve the intelligence of the robot in SimuroSot. Moreover, it will bring far‐reaching influence on intelligence development of robot soccer match in the future.
Notes
Project supported by the fund of national ministries and the Young Teacher innovation fund of Northwestern Polytechnical University.
Corresponding author
Shi Hao‐bin can be contacted at: [email protected]
References
Chen, C., Li, H. and Ding, H. (2007), “A simple model‐based approach for time‐pressure dispensing control”, Advances in Systems Science and Applications, Vol. 7 No. 1.
Kong, F., Zhang, G. and Wang, G. (2007), “Martingale method of optimization design in investment and effectiveness model*”, Advances in Systems Science and Applications, Vol. 7 No. 1.
Kong, X. and Zou, K. (2006), “Evaluation method of fuzzy integral for the complex development level of the society*”, Advances in Systems Science and Applications, Vol. 6 No. 3.
Li, S., Cheng, J. and Sun, Z.Q. (2001), “Structural design and implementation of Tsinghua robot soccer team”, Journal of Tsinghua University (Natural Science), Vol. 41 No. 7, pp. 94‐7.
Lin, Y. and Liu, S. (2006), “Solving problems with incomplete information: a grey systems approach”, Advances in Imaging and Electron Physics, Vol. 141, Elsevier, Oxford, pp. 77‐174.
Lin, Y., Chen, M.‐Y. and Liu, S.‐F. (2004), “Theory of grey systems: capturing uncertainties of grey information”, Kybernetes: The International Journal of Systems & Cybernetics, Vol. 33 No. 2, pp. 196‐218.
Liu, K.D., Lin, Y. and Gao, L.G. (2002), “Unascertained rationals and subjective uncertain information”, Systems Analysis Modelling Simulation, Vol. 42, pp. 343‐58.
Park, K.‐H., Kim, Y.‐J. and Kim, J.‐H. (2001), “Modular Q‐learning based multi‐agent cooperation for robot soccer”, Robotics and Autonomous Systems, Vol. 35, pp. 109‐22.
Wilsker, B. (2003), A Study of Multi‐Agent Collaboration Theories, Information Science Institute Research Report No. ISI/RR‐961149, November.
Wu, L.J., Zhang, C.H. and Xu, X.H. (2001), “Robot soccer reasoning model of decision‐making system”, Journal of Northeastern University (Natural Science), Vol. 22 No. 6, pp. 597‐9.