Online entertainment game difficulty self-adaptive adjustment method based on evolutionary algorithm

By applying evolutionary algorithms and adaptive feedback control systems in online entertainment games, combining parallel computing and machine learning technology to analyze player data in real time, the shortcomings of difficulty adjustment in the existing technology are solved, and dynamic and personalized game experience improvement is achieved and efficient calculation.

CN118949428AInactive Publication Date: 2024-11-15SHENZHEN LEYI TECH CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411066312.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing online entertainment game difficulty adjustment methods lack dynamic adaptability, low data processing efficiency, lack of response time delay and multi-dimensional data analysis, and cannot meet the needs of real-time and personalization.

Method used

Using an evolutionary algorithm-based method, dynamically adjust the game difficulty by collecting and analyzing player data in real time, combining adaptive feedback control systems, parallel computing and machine learning technology to optimize game difficulty adjustments to ensure fast response speed, efficient data processing and multi-dimensional Comprehensive analysis.

Benefits of technology

It realizes dynamic adjustment of game difficulty based on players' real-time performance, improves game challenge and user experience, improves the fun and continuous attractiveness of the game, and significantly improves computing efficiency and response speed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118949428A_ABST
    Figure CN118949428A_ABST
Patent Text Reader

Abstract

The invention discloses an online entertainment game difficulty self-adaptive adjustment method based on an evolutionary algorithm, which comprises the following steps: S1, initializing game difficulty at the beginning of a game, and collecting player data in real time; s2, based on the collected player data, evaluating the performance of the player by using a data analysis method; s3, adjusting the game difficulty by applying an evolutionary algorithm; s4, evaluating and screening the candidate game difficulty according to the real-time feedback of the player through a self-adaptive feedback control system; s5, predicting the ability and behavior mode of the player by using real-time data analysis and machine learning technologies, and adjusting the game difficulty in advance; and S6, through parallel computing, a multi-thread technology, GPU acceleration and a distributed computing architecture, the response speed and the computing efficiency of game difficulty adjustment are improved. The method has the advantages of high dynamic real-time adjustment capability, high data processing efficiency, short response time and multi-dimensional data comprehensive analysis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of adaptive adjustment technology, and in particular to an adaptive adjustment method for difficulty of an online entertainment game based on an evolutionary algorithm. Background Art

[0002] With the rapid development of information technology and the online entertainment industry, adaptive adjustment of game difficulty has become an important technology to improve user experience and game stickiness. Especially in online entertainment games, the reasonable adjustment of game difficulty plays a vital role in the sustained interest and participation of players. However, the existing technology still has many deficiencies in adaptive adjustment of game difficulty, and it is difficult to meet the requirements of real-time and personalization.

[0003] In the prior art, traditional methods for adjusting game difficulty mostly rely on preset rules and fixed difficulty levels. These methods are powerless in the face of diverse player behaviors and ever-changing game environments. Specifically, traditional methods have obvious deficiencies in the following aspects:

[0004] 1. Lack of dynamic adaptability: Traditional difficulty adjustment mechanisms are usually based on fixed rules or simple dynamic adjustment strategies, and cannot be flexibly adjusted according to the player's real-time performance, resulting in the game difficulty being either too simple or too complex.

[0005] 2. Low data processing efficiency: Faced with massive amounts of player data, traditional methods often lack efficient data processing and analysis capabilities and are unable to extract useful information from the data in a timely manner, resulting in delayed difficulty adjustments.

[0006] 3. Response time delay: Due to its reliance on preset rules and static models, traditional difficulty adjustment systems have significant delays in detecting and responding to changes in player behavior, and cannot achieve real-time adjustment and rapid response.

[0007] 4. Lack of comprehensive perspective: Traditional methods mostly focus on a single indicator (such as player score or game time), ignoring the comprehensive analysis of multi-dimensional data such as player operation speed and number of incorrect operations, making it difficult to fully reflect the player's true ability and gaming experience.

[0008] Therefore, how to provide a method for adaptively adjusting the difficulty of online entertainment games based on an evolutionary algorithm is a problem that those skilled in the art urgently need to solve. Summary of the invention

[0009] One purpose of the present invention is to propose a method for adaptively adjusting the difficulty of online entertainment games based on evolutionary algorithms. The present invention makes full use of technologies such as evolutionary algorithms, data analysis and mining, real-time computing and response, feedback control systems and parallel computing, and describes in detail an algorithm for dynamically adjusting the difficulty of games by collecting and analyzing player data in real time. The method has the advantages of strong dynamic real-time adjustment capability, high data processing efficiency, fast response time and comprehensive analysis of multi-dimensional data.

[0010] A method for adaptively adjusting the difficulty of an online entertainment game based on an evolutionary algorithm according to an embodiment of the present invention is characterized by comprising the following steps:

[0011] S1. Initialize the game difficulty at the beginning of the game, and collect the player's score, game time, operation speed and number of wrong operations in real time;

[0012] S2. Based on the collected player data, use data analysis methods to evaluate the player's performance and generate a preliminary model of the player's current gaming ability;

[0013] S3, applying an evolutionary algorithm to adjust the game difficulty, selecting the current game difficulty and several candidate game difficulties as the initial population, performing a cross operation on the selected candidate game difficulties to generate new candidate difficulties, performing a mutation operation on the cross-generated candidate difficulties, and generating a variety of difficulty adjustment schemes;

[0014] S4, through the adaptive feedback control system, evaluate and screen the difficulty of candidate games according to the real-time feedback of players, and select the game difficulty that suits the current ability of players;

[0015] S5. Use real-time data analysis and machine learning technology to predict players’ abilities and behavior patterns and adjust game difficulty in advance;

[0016] S6. Through parallel computing, multi-threading technology, GPI acceleration and distributed computing architecture, the calculation process of the evolutionary algorithm is optimized to improve the response speed and computing efficiency of game difficulty adjustment.

[0017] Optionally, the S1 specifically includes:

[0018] S11. At the beginning of the game, the number of enemies, enemy attack power, number of obstacles and game progress speed are initialized according to a preset game difficulty model;

[0019] S12. Collect the player's score data in real time. The score data S is the game score obtained by the player within a specific time:

[0020]

[0021] Where t is time, s is iis the score in the i-th time period;

[0022] S13. Collect the player's game duration data in real time. The game duration data T is the duration of the player in a specific game stage:

[0023]

[0024] Among them, t i is the length of the i-th time period;

[0025] S14. Collect the player's operation speed data in real time. The operation speed V is the number of operations per unit time of the player:

[0026]

[0027] Among them, O is the total number of operations, and T is the total game time;

[0028] S15. Collect the number of wrong operations of the player in real time, where the number of wrong operations E is the number of wrong operations that occurred during the game:

[0029]

[0030] Among them, e i is the number of wrong operations in the i-th time period;

[0031] S16, standardize the collected player data to generate a standardized player data set D:

[0032]

[0033] Among them, μ is the mean of the data set, σ is the standard deviation of the data set, and x i is the i-th item in the original data, x′ i is the standardized data;

[0034] S17. Construct the player's real-time performance matrix M based on the standardized player data set D:

[0035]

[0036] Among them, the matrix element M ij Represents the standardized value of the j-th data in the i-th time period.

[0037] Optionally, the S2 specifically includes:

[0038] S21. Based on the collected player data, use data analysis methods to evaluate the player's performance in multiple dimensions and build a dynamic player performance evaluation model. The model input is the standardized player data set D, and is updated in real time in combination with the time decay factor and weight adjustment mechanism;

[0039] S22. Calculate the player's time-weighted average score Introduce the time decay factor λ:

[0040]

[0041] Among them, s i is the score in the i-th time period, n is the total number of time periods, and λ is the time decay factor;

[0042] S23. Calculate the time-weighted average game duration of players

[0043]

[0044] Among them, t i is the length of the i-th time period, λ is the time decay factor;

[0045] S24. Calculate the time-weighted average operation speed of the player

[0046]

[0047] Among them, v i is the operating speed in the i-th time period, λ is the time decay factor;

[0048] S25. Calculate the time-weighted average number of wrong operations of the player

[0049]

[0050] Among them, e i is the number of misoperations in the i-th time period, and λ is the time decay factor;

[0051] S26, using the time-weighted average value calculated above, to generate a dynamic preliminary model of the player's current gaming ability, the preliminary model including the player's comprehensive performance index I t The comprehensive performance index is calculated by dynamic weighted average. Assuming the weight coefficients are α, β, γ, and δ respectively, the comprehensive performance index I t for:

[0052]

[0053] Among them, α+β+γ+δ=1, and α, β, γ, δ can be dynamically adjusted according to the player's performance.

[0054] Optionally, the S3 specifically includes:

[0055] S31, select the current game difficulty and several candidate game difficulties as the initial population, and set the current game difficulty to be D c , the difficulty of the candidate games is {D1,D2,...,D k}, the population size of the initial population is set to N, including the current game difficulty and the candidate game difficulty, and the initial population is expressed as:

[0056] P0={D c ,D1,D2,...,D k};

[0057] Among them, the initial population size N = k + 1;

[0058] S32. Evaluate the fitness of the initial population. The evaluation criteria include the player's comprehensive performance index I t , game fluency G and player satisfaction S, the fitness function F is defined as:

[0059] F(D i )=w1·I t (D i )+w2·G(D i )+w3·S(D i );

[0060] Among them, w1, w2, w3 are weight coefficients, satisfying w1+w2+w3=1, I t (D i ) is the comprehensive performance index of the player corresponding to the i-th game difficulty, G(D i ) is the game fluency score corresponding to the i-th game difficulty, S(D i ) is the player satisfaction score corresponding to the i-th game difficulty;

[0061] S33, based on the fitness evaluation results, select several game difficulties with higher fitness to enter the crossover operation, and set the selection probability to be P s , the selection operation adopts the roulette selection mechanism, and the selection probability P s Defined as:

[0062]

[0063] Among them, F(D i ) is the fitness value of the i-th game difficulty, and N is the size of the initial population;

[0064] S34, perform a cross operation on the difficulty of the selected candidate games, using a single-point crossover method to generate a new candidate game difficulty, assuming that the crossover point is C and the difficulty of the two selected games is D i and D j , the new game difficulty D′ after crossover is expressed as:

[0065] D′=Crossover(D i ,D j ,C);

[0066]

[0067] Among them, k is the index of the game difficulty parameter, D i (k) and D j (k) are the parameter values ​​of the i-th and j-th game difficulties at index k, respectively, and C is the intersection position;

[0068] S35, perform mutation operation on the difficulty of the candidate games generated by crossover, and set the mutation probability to be P m :

[0069] D″=Mutate(D′,P m );

[0070]

[0071] Where D″ is the game difficulty after mutation, D′ is the game difficulty after crossover, k is the index of the game difficulty parameter, Δ is a small random change value, rand(0,1) is a random number function generated between 0 and 1, P m is the preset mutation probability;

[0072] S36. Re-evaluate the fitness of the candidate game difficulty after mutation, and compare it with the initial population, select several game difficulties with higher fitness to enter the next generation population, and assume that the candidate game difficulty after mutation is D″, and the fitness function is F:

[0073] F(D″)=w1·I t (D″)+w2·G(D″)+w3·S(D″);

[0074] Among them, w1, w2, w3 are weight coefficients, satisfying w1+w2+w3=1, I t (D″) is the comprehensive performance index of players corresponding to the difficulty of the game after mutation, G(D″) is the game fluency score corresponding to the difficulty of the game after mutation, and S(D″) is the player satisfaction score corresponding to the difficulty of the game after mutation;

[0075] Select the game difficulty with higher fitness to enter the next generation population, and set the new generation population as P next :

[0076] P next ={D i ∈P∪D″|TopNbyF(D i )};

[0077] Among them, P is the initial population, D″ is the difficulty of the candidate game after mutation, and N is the size of the new generation population;

[0078] S37, let the current generation be t, and the maximum evolution generation be T max , the fitness value convergence judgment condition is ΔF<∈, where ΔF is the fitness change and ∈ is the convergence threshold:

[0079] ΔF=|F avg (P t+1 )-F avg (P t )|;

[0080] Among them, F avg (P t ) is the average fitness value of the t-th generation population:

[0081]

[0082] Among them, if ΔF<∈ or t≥T max , then stop evolving and output the optimized game difficulty D * , D * is the individual with the highest fitness in the final population;

[0083] S38, the optimized game difficulty D * Applied to the actual game, the game difficulty is adjusted in real time to adapt to the player's current ability and performance. The game difficulty after real-time adjustment is set to D real :

[0084] D real (t+1)=α·D real (t)+(1-α)·D * ;

[0085] Among them, α is the adjustment smoothing coefficient, which ranges from 0 to 1 and is used to control the smoothness of real-time adjustment.

[0086] Optionally, the S4 specifically includes:

[0087] S41, monitoring the player's game performance in real time through an adaptive feedback control system, wherein the game performance includes the player's score S, game time T, operation speed V, and number of erroneous operations E;

[0088] S42, the monitored player performance data is associated with the current game difficulty to construct a real-time feedback matrix f, the matrix element F ij Represents the real-time feedback value of the jth data in the i-th time period:

[0089]

[0090] S43. Calculate the comprehensive performance index I of the real-time feedback matrix F f :

[0091]

[0092] in, and are the average values ​​of player score, game time, operation speed and number of wrong operations respectively, α, β, γ, δ are weight coefficients, satisfying α+β+γ+δ=1;

[0093] S44. Comprehensive performance index I f Perform dynamic adjustments, set the adjustment coefficient to K, and adjust the game difficulty D based on the real-time feedback from players:

[0094] D new =D crrent +K·(I target -I f );

[0095] Among them, D current is the current game difficulty, D ne w is the adjusted game difficulty, I target is the target comprehensive performance index;

[0096] S45. Introduce an integral control mechanism into the real-time feedback control system, and set the integral time as T i :

[0097]

[0098] Among them, ΔD is the integral control adjustment amount, and t is the time;

[0099] S46, combining proportional control and integral control to generate the final adjusted game difficulty D final , and the final adjusted game difficulty D fainal Applied to the game:

[0100] D funal =D new +ΔD.

[0101] Optionally, the S5 specifically includes:

[0102] S51. Predicting the player's ability and behavior pattern using real-time data analysis and machine learning technology, wherein the player data includes score S, game time T, operation speed V, and number of erroneous operations E;

[0103] S52. Input the player data into the prediction model. The prediction model uses the long short-term memory (LSTM) neural network to construct the input sequence X and output sequence Y of the prediction model:

[0104]

[0105] Y=[S n+1 T n+1 V n+1 E n+1 ];

[0106] S53. Train the LSTM model and use the time series cross-validation method to evaluate the model performance. The loss function is defined as the mean square error (MSE):

[0107]

[0108] Among them, Y i is the true value, is the predicted value;

[0109] S54. Use the trained LSTM model to predict the player's future performance and generate a prediction value in Including future scores Game duration Operation speed and number of misoperations

[0110]

[0111] S55. According to the prediction results, the game difficulty is adjusted in advance, and the adjustment coefficient is set to K′:

[0112] D pred =D current +K′·(I target -I pred );

[0113] Among them, D current is the current game difficulty, D pred To predict the game difficulty after adjustment, I target is the target comprehensive performance index, I pred According to the predicted value Calculated Composite Performance Index;

[0114] S56. In the actual game process, the final game difficulty D is generated by combining real-time feedback adjustment and prediction adjustment. final :

[0115] D final =β·D real +(1-β)·D pred ;

[0116] Among them, D realThe game difficulty is adjusted according to real-time feedback, and β is the weight coefficient, ranging from 0 to 1, which is used to balance the real-time adjustment and predictive adjustment.

[0117] Optionally, the S6 specifically includes:

[0118] S61. In the selection, crossover and mutation process of the evolutionary algorithm, multiple candidate game difficulties are processed in parallel using multi-threading technology. Assume that the number of threads for parallel processing is M, and the set of candidate game difficulties processed by each thread is {D i1 ,D i2 ,...,D ik}:

[0119]

[0120] Among them, P t is the t-th generation population;

[0121] S62, in the fitness evaluation process, GPU acceleration technology is used to parallel calculate the fitness function, the fitness function F(D i )for:

[0122] F(D i )=w1·I t (D i )+w2·G(D i )+w3·S(D i );

[0123] Among them, D i is the difficulty of the i-th candidate game, w1, w2, w3 are weight coefficients, satisfying w1+w2+w3=1, I t (D i ) is the player's comprehensive performance index, G(D i ) is the game fluency score, S(D i ) to rate player satisfaction;

[0124] S63. In the crossover operation and mutation operation, the distributed computing architecture is used for parallel processing. Assume that the number of distributed computing nodes is N, and the set of crossover and mutation operations processed by each node is {C i1 ,C i2 ,...,C ik} and {M i1 ,M i2 ,...,M ik}:

[0125]

[0126] Among them, P t+1 is the t+1 generation population, Crossover(D i1,D i2 ,C ij ) is the jth crossover operation in the i-th node, Multate(D ij ,M ij ) is the jth mutation operation in the i-th node;

[0127] S64, introduce a dynamic load balancing mechanism to dynamically adjust task allocation according to the real-time load of each computing node. Suppose the node load is L i , the total load is L total :

[0128] L i,new =L i +ΔL i ;

[0129] in, Balanced distribution of computing resources is achieved by dynamically adjusting the load of each node;

[0130] S65. At each stage of the evolution process, optimize the allocation strategy of computing resources by monitoring computing performance indicators:

[0131]

[0132] Among them, T compute is the total computation time, is the utilization of the i-th GPU, is the utilization rate of the i-th CPU thread, by optimizing P eff Improve overall computing efficiency.

[0133] The beneficial effects of the present invention are:

[0134] (1) The present invention provides a deep understanding and dynamic adaptation of player behavior by combining evolutionary algorithms, data analysis and mining, real-time computing and response, feedback control systems, and parallel computing technologies, so that the system can continuously adjust the game difficulty according to the player's real-time performance, thereby effectively improving the challenge of the game and the user experience. Through the selection, crossover, and mutation operations of the evolutionary algorithm, the game difficulty can be gradually optimized among a variety of candidate solutions, ensuring that each player can obtain the gaming experience that best suits his or her ability.

[0135] (2) The present invention uses an adaptive feedback control system to monitor and analyze the player's game performance in real time, combines the score control mechanism and the proportional control mechanism, quickly responds to the player's feedback, and adjusts the game difficulty in a timely manner to avoid excessive or untimely adjustment of the game difficulty. The multi-dimensional data comprehensive analysis is used to fully reflect the player's real ability and ensure the accuracy and stability of the game difficulty adjustment.

[0136] (3) The present invention utilizes parallel computing, multi-threading technology, and GPU acceleration technology to significantly improve the efficiency of data processing and computing. By processing multiple candidate game difficulties in parallel during the selection, crossover, and mutation process of the evolutionary algorithm, and by using GPU acceleration technology to parallelly calculate the fitness function during the fitness evaluation process, the rapid response and efficient execution of the game difficulty adjustment are ensured. At the same time, a dynamic load balancing mechanism is introduced to dynamically adjust task allocation according to the real-time load conditions of each computing node, optimize the utilization of computing resources, and further improve the overall computing efficiency.

[0137] (4) The present invention uses machine learning technology to predict the player's ability and behavior pattern, adjust the game difficulty in advance, and provide a personalized gaming experience. The long short-term memory (LSTM) neural network model is used to predict the player's future performance in real time, and the game difficulty is dynamically adjusted according to the prediction results to ensure that the game difficulty matches the player's expected performance, thereby improving the fun and continued appeal of the game. BRIEF DESCRIPTION OF THE DRAWINGS

[0138] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0139] Figure 1 This is a general framework diagram of a method for adaptively adjusting the difficulty of online entertainment games based on an evolutionary algorithm proposed by the present invention;

[0140] Figure 2 A flowchart of a method for adaptively adjusting the difficulty of an online entertainment game based on an evolutionary algorithm proposed by the present invention;

[0141] Figure 3 A schematic diagram of real-time collection of player data and initialization of game difficulty of an online entertainment game difficulty adaptive adjustment method based on an evolutionary algorithm proposed by the present invention;

[0142] Figure 4 This is a schematic diagram of real-time monitoring and adjustment of game difficulty by an adaptive feedback control system of an online entertainment game difficulty adaptive adjustment method based on an evolutionary algorithm proposed by the present invention. DETAILED DESCRIPTION

[0143] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0144] refer to Figure 1-Figure 4 , a method for adaptively adjusting the difficulty of an online entertainment game based on an evolutionary algorithm, characterized in that it comprises the following steps:

[0145] S1. Initialize the game difficulty at the beginning of the game, and collect the player's score, game time, operation speed and number of wrong operations in real time;

[0146] In this implementation, S1 specifically includes:

[0147] S11. At the beginning of the game, the number of enemies, enemy attack power, number of obstacles and game progress speed are initialized according to a preset game difficulty model;

[0148] S12. Collect the player's score data in real time. The score data S is the game score obtained by the player within a specific time:

[0149]

[0150] Where t is time, s is i is the score in the i-th time period;

[0151] S13. Collect the player's game duration data in real time. The game duration data T is the duration of the player in a specific game stage:

[0152]

[0153] Among them, t i is the length of the i-th time period;

[0154] S14. Collect the player's operation speed data in real time. The operation speed V is the number of operations per unit time of the player:

[0155]

[0156] Among them, O is the total number of operations, and T is the total game time;

[0157] S15. Collect the number of wrong operations of the player in real time, where the number of wrong operations E is the number of wrong operations that occurred during the game:

[0158]

[0159] Among them, e i is the number of wrong operations in the i-th time period;

[0160] S16, standardize the collected player data to generate a standardized player data set D:

[0161]

[0162] Among them, μ is the mean of the data set, σ is the standard deviation of the data set, and x i is the i-th item in the original data, x′ i is the standardized data;

[0163] S17. Construct the player's real-time performance matrix M based on the standardized player data set D:

[0164]

[0165] Among them, the matrix element M ij Represents the standardized value of the j-th data in the i-th time period.

[0166] S2. Based on the collected player data, use data analysis methods to evaluate the player's performance and generate a preliminary model of the player's current gaming ability;

[0167] In this implementation, S2 specifically includes:

[0168] S21. Based on the collected player data, use data analysis methods to evaluate the player's performance in multiple dimensions and build a dynamic player performance evaluation model. The model input is the standardized player data set D, and is updated in real time in combination with the time decay factor and weight adjustment mechanism;

[0169] S22. Calculate the player's time-weighted average score Introduce the time decay factor λ:

[0170]

[0171] Among them, s i is the score in the i-th time period, n is the total number of time periods, and λ is the time decay factor;

[0172] S23. Calculate the time-weighted average game duration of players

[0173]

[0174] Among them, t i is the length of the i-th time period, λ is the time decay factor;

[0175] S24. Calculate the time-weighted average operation speed of the player

[0176]

[0177] Among them, v i is the operating speed in the i-th time period, λ is the time decay factor;

[0178] S25. Calculate the time-weighted average number of wrong operations of the player

[0179]

[0180] Among them, e i is the number of misoperations in the i-th time period, and λ is the time decay factor;

[0181] S26, using the time-weighted average value calculated above, to generate a dynamic preliminary model of the player's current gaming ability, the preliminary model including the player's comprehensive performance index I t The comprehensive performance index is calculated by dynamic weighted average. Assuming the weight coefficients are α, β, γ, and λ respectively, the comprehensive performance index I t for:

[0182]

[0183] Among them, α+β+γ+δ=1, and α, β, γ, δ can be dynamically adjusted according to the player's performance.

[0184] S3, applying an evolutionary algorithm to adjust the game difficulty, selecting the current game difficulty and several candidate game difficulties as the initial population, performing a cross operation on the selected candidate game difficulties to generate new candidate difficulties, performing a mutation operation on the cross-generated candidate difficulties, and generating a variety of difficulty adjustment schemes;

[0185] In this implementation, S3 specifically includes:

[0186] S31, select the current game difficulty and several candidate game difficulties as the initial population, and set the current game difficulty to be D c , the difficulty of the candidate games is {D1,D2,...,D k}, the population size of the initial population is set to N, including the current game difficulty and the candidate game difficulty, and the initial population is expressed as:

[0187] P0={D c ,D1,D2,...,D k};

[0188] Among them, the initial population size N = k + 1;

[0189] S32. Evaluate the fitness of the initial population. The evaluation criteria include the player's comprehensive performance index I t , game fluency G and player satisfaction S, the fitness function F is defined as:

[0190] F(D i )=w1·I t (D i )+w2·G(D i )+w3·S(D i );

[0191] Among them, w1, w2, w3 are weight coefficients, satisfying w1+w2+w3=1, I t (D i ) is the comprehensive performance index of the player corresponding to the i-th game difficulty, G(D i ) is the game fluency score corresponding to the i-th game difficulty, S(D i ) is the player satisfaction score corresponding to the i-th game difficulty;

[0192] S33, based on the fitness evaluation results, select several game difficulties with higher fitness to enter the crossover operation, and set the selection probability to be P s , the selection operation adopts the roulette selection mechanism, and the selection probability P s Defined as:

[0193]

[0194] Among them, F(D i ) is the fitness value of the i-th game difficulty, and N is the size of the initial population;

[0195] S34, perform a cross operation on the difficulty of the selected candidate games, using a single-point crossover method to generate a new candidate game difficulty, assuming that the crossover point is C and the difficulty of the two selected games is D i and D j , the new game difficulty D′ after crossover is expressed as:

[0196] D′=Crossover(D i ,D j ,C);

[0197]

[0198] Among them, k is the index of the game difficulty parameter, D i (k) and D j (k) are the parameter values ​​of the i-th and j-th game difficulties at index k, respectively, and C is the intersection position;

[0199] S35, perform mutation operation on the difficulty of the candidate games generated by crossover, and set the mutation probability to be P m :

[0200] D″=Mutate(D′,P m );

[0201]

[0202] Where D″ is the game difficulty after mutation, D′ is the game difficulty after crossover, k is the index of the game difficulty parameter, Δ is a small random change value, rand(0,1) is a random number function generated between 0 and 1, Pm is the preset mutation probability;

[0203] S36. Re-evaluate the fitness of the candidate game difficulty after mutation, and compare it with the initial population, select several game difficulties with higher fitness to enter the next generation population, and assume that the candidate game difficulty after mutation is D″, and the fitness function is F:

[0204] F(D″)=w1·I t (D″)+w2·G(D″)+w3·S(D″);

[0205] Among them, w1, w2, w3 are weight coefficients, satisfying w1+w2+w3=1, I t (D″) is the comprehensive performance index of players corresponding to the difficulty of the game after mutation, G(D″) is the game fluency score corresponding to the difficulty of the game after mutation, and S(D″) is the player satisfaction score corresponding to the difficulty of the game after mutation;

[0206] Select the game difficulty with higher fitness to enter the next generation population, and set the new generation population as P next :

[0207] P next ={D i ∈P∪D″|TopNbyF(D i )};

[0208] Among them, P is the initial population, D″ is the difficulty of the candidate game after mutation, and N is the size of the new generation population;

[0209] S37, let the current generation be t, and the maximum evolution generation be T max , the fitness value convergence judgment condition is ΔF<∈, where ΔF is the fitness change and ∈ is the convergence threshold:

[0210] ΔF=|F avg (P t+1 )-F avg (P t )|;

[0211] Among them, F avg (P t ) is the average fitness value of the t-th generation population:

[0212]

[0213] Among them, if ΔF<∈ or t≥T max , then stop evolving and output the optimized game difficulty D * , D * is the individual with the highest fitness in the final population;

[0214] S38, the optimized game difficulty D * Applied to the actual game, the game difficulty is adjusted in real time to adapt to the player's current ability and performance. The game difficulty after real-time adjustment is set to D real :

[0215] D real (t+1)=α·D real (t)+(1-α)·D * ;

[0216] Among them, α is the adjustment smoothing coefficient, which ranges from 0 to 1 and is used to control the smoothness of real-time adjustment.

[0217] S4, through the adaptive feedback control system, evaluate and screen the difficulty of candidate games according to the real-time feedback of players, and select the game difficulty that suits the current ability of players;

[0218] In this implementation, S4 specifically includes:

[0219] S41, monitoring the player's game performance in real time through an adaptive feedback control system, wherein the game performance includes the player's score S, game time T, operation speed V, and number of erroneous operations E;

[0220] S42, the monitored player performance data is associated with the current game difficulty to construct a real-time feedback matrix F, the matrix element F ij Represents the real-time feedback value of the jth data in the i-th time period:

[0221]

[0222] S43. Calculate the comprehensive performance index I of the real-time feedback matrix F f :

[0223]

[0224] in, and are the average values ​​of player score, game time, operation speed and number of wrong operations respectively, α, β, γ, δ are weight coefficients, satisfying α+β+γ+δ=1;

[0225] S44. Comprehensive performance index I f Perform dynamic adjustments, set the adjustment coefficient to K, and adjust the game difficulty D based on the real-time feedback from players:

[0226] D new =D current +K·(I target -I f );

[0227] Among them, Dcurrent is the current game difficulty, D new The game difficulty is adjusted. target is the target comprehensive performance index;

[0228] S45. Introduce an integral control mechanism into the real-time feedback control system, and set the integral time as T i :

[0229]

[0230] Among them, ΔD is the integral control adjustment amount, and t is the time;

[0231] S46, combining proportional control and integral control to generate the final adjusted game difficulty D final , and the final adjusted game difficulty D final Applied to the game:

[0232] D final =D new +ΔD.

[0233] S5. Use real-time data analysis and machine learning technology to predict players’ abilities and behavior patterns and adjust game difficulty in advance;

[0234] In this implementation, S5 specifically includes:

[0235] S51. Predicting the player's ability and behavior pattern using real-time data analysis and machine learning technology, wherein the player data includes score S, game time T, operation speed V, and number of erroneous operations E;

[0236] S52. Input the player data into the prediction model. The prediction model uses the long short-term memory (LSTM) neural network to construct the input sequence X and output sequence Y of the prediction model:

[0237]

[0238] Y=[S n+1 T n+1 V n+1 E n+1 ];

[0239] S53. Train the LSTM model and use the time series cross-validation method to evaluate the model performance. The loss function is defined as the mean square error (MSE):

[0240]

[0241] Among them, Y i is the true value, is the predicted value;

[0242] S54. Use the trained LSTM model to predict the player's future performance and generate a prediction value in Including future scores Game duration Operation speed and number of misoperations

[0243]

[0244] S55. According to the prediction results, the game difficulty is adjusted in advance, and the adjustment coefficient is set to K′:

[0245] D pred =D current +K′·(I target -I pred );

[0246] Among them, D current is the current game difficulty, D pred To predict the game difficulty after adjustment, I target is the target comprehensive performance index, I pred According to the predicted value Calculated Composite Performance Index;

[0247] S56. In the actual game process, the final game difficulty D is generated by combining real-time feedback adjustment and prediction adjustment. final :

[0248] D final =β·D real +(1-β)·D pred ;

[0249] Among them, D real The game difficulty is adjusted according to real-time feedback, and β is the weight coefficient, ranging from 0 to 1, which is used to balance the real-time adjustment and predictive adjustment.

[0250] S6. Through parallel computing, multi-threading technology, GPU acceleration and distributed computing architecture, the calculation process of the evolutionary algorithm is optimized to improve the response speed and computing efficiency of game difficulty adjustment.

[0251] In this implementation, S6 specifically includes:

[0252] S61. In the selection, crossover and mutation process of the evolutionary algorithm, multiple candidate game difficulties are processed in parallel using multi-threading technology. Assume that the number of threads for parallel processing is M, and the set of candidate game difficulties processed by each thread is {D i1 ,D i2 ,...,D ik}:

[0253]

[0254] Among them, P t is the t-th generation population;

[0255] S62, in the fitness evaluation process, GPU acceleration technology is used to parallel calculate the fitness function, the fitness function F(D i )for:

[0256] F(D i )=w1·I t (D i )+w2·G(D i )+w3·S(D i );

[0257] Among them, D i is the difficulty of the i-th candidate game, w1, w2, w3 are weight coefficients, satisfying w1+w2+w3=1, I t (D i ) is the player's comprehensive performance index, G(D i ) is the game fluency score, S(D i ) to rate player satisfaction;

[0258] S63. In the crossover operation and mutation operation, the distributed computing architecture is used for parallel processing. Assume that the number of distributed computing nodes is N, and the set of crossover and mutation operations processed by each node is {C i1 ,C i2 ,...,C ik} and {M i1 ,M i2 ,...,M ik}:

[0259]

[0260] Among them, P t+1 is the t+1 generation population, Crossover(D i1 ,D i2 ,C ij ) is the jth crossover operation in the i-th node, Mutate(D ij ,M ij ) is the jth mutation operation in the i-th node;

[0261] S64, introduce a dynamic load balancing mechanism to dynamically adjust task allocation according to the real-time load of each computing node. Suppose the node load is L i , the total load is L total :

[0262] L i,new =Li +ΔL i ;

[0263] in, Balanced distribution of computing resources is achieved by dynamically adjusting the load of each node;

[0264] S65. At each stage of the evolution process, optimize the allocation strategy of computing resources by monitoring computing performance indicators:

[0265]

[0266] Among them, T compute is the total computation time, is the utilization of the i-th GPU, is the utilization rate of the i-th CPU thread, by optimizing P eff Improve overall computing efficiency.

[0267] Embodiment 1:

[0268] In order to verify the feasibility and effectiveness of the present invention, we applied the present invention to a large online entertainment game platform. The platform has millions of active users, with various game types and a wide range of difficulty. The traditional fixed difficulty adjustment mechanism cannot meet the needs of all players, causing some players to lose interest because they think the game is too simple, and some players to feel frustrated because they think the game is too difficult. Therefore, the platform decided to deploy the online entertainment game difficulty adaptive adjustment method based on evolutionary algorithm of the present invention.

[0269] The gaming platform selected a representative online multiplayer game for testing. The game includes multiple levels, each with different difficulty levels. To ensure the comprehensiveness of the test, the platform selected 1,000 active players for the experiment, covering all levels from novices to veteran players.

[0270] First, at the start of the game, the system initializes the game difficulty and begins to collect the player's game data in real time, including the player's score, game time, operation speed, and number of wrong operations. The system uses this data and data analysis methods to evaluate the player's performance and generate a preliminary model of the player's current game ability.

[0271] Next, the system applies an evolutionary algorithm to adjust the game difficulty. The initial population consists of the current game difficulty and several candidate game difficulties. The system continuously optimizes the game difficulty through selection, crossover, and mutation operations. Specifically, based on the fitness evaluation results, the system selects several game difficulties with higher fitness for crossover operations, and mutates the candidate game difficulties generated by the crossover. The fitness of the mutated candidate game difficulties is re-evaluated and compared with the initial population, and several game difficulties with higher fitness are selected to enter the next generation population.

[0272] Under the action of the adaptive feedback control system, the system evaluates and screens the difficulty of candidate games based on the real-time feedback of the players, and selects the game difficulty that suits the current ability of the players. Real-time data analysis and machine learning technology are used to predict the ability and behavior patterns of players, adjust the game difficulty in advance, and provide a personalized gaming experience. Through parallel computing, multi-threading technology, GPU acceleration and distributed computing architecture and other technologies, the system optimizes evolutionary algorithms and other computing processes to improve response speed and computing efficiency. Specific data proves our beneficial effects: In order to verify the effectiveness of the present invention, we collected a large amount of data during the experiment and conducted a detailed analysis.

[0273] Table 1 Game platform difficulty adaptive adjustment effect report

[0274]

[0275]

[0276] It can be seen from these data that after implementing the present invention, the average game time of players increased significantly, indicating that players' interest and involvement in the game increased. The average score of players also increased significantly, indicating that the system can effectively adjust the difficulty of the game to match the ability level of the players. The player's operation speed increased and the number of wrong operations decreased, indicating that the system can help players improve their gaming skills and reduce frustration when adjusting the game difficulty. The increase in satisfaction scores further proves the players' recognition of the new system.

[0277] In addition, the system demonstrated extremely high response speed and computing efficiency throughout the experiment. Through parallel computing and GPU acceleration technology, the system's average response time when detecting and responding to changes in player behavior was shortened from 10 seconds to 2 seconds, significantly improving the real-time adjustment of game difficulty. The introduction of a dynamic load balancing mechanism enables efficient use of computing resources, more balanced load distribution of each computing node, and a 30% increase in overall computing efficiency.

[0278] The adaptive difficulty adjustment method for online entertainment games based on evolutionary algorithms of the present invention collects and analyzes player data in real time, uses evolutionary algorithms and parallel computing technology to dynamically adjust the game difficulty, and significantly improves the game experience and satisfaction of players. The excellent data obtained in the experiment proves the effectiveness and superiority of the present invention, and provides a powerful and flexible difficulty adjustment solution for online entertainment games.

[0279] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for adaptively adjusting the difficulty of online entertainment games based on evolutionary algorithms, characterized in that: The steps include: S1. Initialize the game difficulty at the beginning of the game, and collect the player's score, game time, operation speed and number of wrong operations in real time; S2. Based on the collected player data, use data analysis methods to evaluate the player's performance and generate a preliminary model of the player's current gaming ability; S3, applying an evolutionary algorithm to adjust the game difficulty, selecting the current game difficulty and several candidate game difficulties as the initial population, performing a cross operation on the selected candidate game difficulties to generate new candidate difficulties, performing a mutation operation on the cross-generated candidate difficulties, and generating a variety of difficulty adjustment schemes; S4, through the adaptive feedback control system, evaluate and screen the difficulty of candidate games according to the real-time feedback of players, and select the game difficulty that suits the current ability of players; S5. Use real-time data analysis and machine learning technology to predict players’ abilities and behavior patterns and adjust game difficulty in advance; S6. Through parallel computing, multi-threading technology, GPU acceleration and distributed computing architecture, the calculation process of the evolutionary algorithm is optimized to improve the response speed and computing efficiency of game difficulty adjustment.

2. The method for adaptively adjusting the difficulty of online entertainment games based on evolutionary algorithms according to claim 1, characterized in that: The S1 specifically includes: S11. At the beginning of the game, the number of enemies, enemy attack power, number of obstacles and game progress speed are initialized according to a preset game difficulty model; S12. Collect the player's score data in real time. The score data S is the game score obtained by the player within a specific time: Where t is time, s is i is the score in the i-th time period; S13. Collect the player's game time data in real time. The game time data T is the duration of the player in a specific game stage: Among them, t i is the length of the i-th time period; S14. Collect the player's operation speed data in real time. The operation speed V is the number of operations per unit time of the player: Among them, O is the total number of operations, and T is the total game time; S15. Collect the number of wrong operations of the player in real time, where the number of wrong operations E is the number of wrong operations that occurred during the game: Among them, e i is the number of wrong operations in the i-th time period; S16, standardize the collected player data to generate a standardized player data set D: Among them, μ is the mean of the data set, σ is the standard deviation of the data set, and x i is the i-th item in the original data, x′ i is the standardized data; S17. Construct the player's real-time performance matrix M based on the standardized player data set D: Among them, the matrix element M ij Represents the standardized value of the j-th data in the i-th time period.

3. The method for adaptively adjusting the difficulty of online entertainment games based on evolutionary algorithms according to claim 2, characterized in that: The S2 specifically includes: S21. Based on the collected player data, use data analysis methods to evaluate the player's performance in multiple dimensions and build a dynamic player performance evaluation model. The model input is the standardized player data set D, and is updated in real time in combination with the time decay factor and weight adjustment mechanism; S22. Calculate the player's time-weighted average score Introduce the time decay factor λ: Among them, s i is the score in the i-th time period, n is the total number of time periods, and λ is the time decay factor; S23. Calculate the time-weighted average game duration of players Among them, t i is the length of the i-th time period, λ is the time decay factor; S24. Calculate the time-weighted average operation speed of the player Among them, v i is the operating speed in the i-th time period, λ is the time decay factor; S25. Calculate the time-weighted average number of wrong operations of the player Among them, e i is the number of misoperations in the i-th time period, and λ is the time decay factor; S26, using the time-weighted average value calculated above, to generate a dynamic preliminary model of the player's current gaming ability, the preliminary model including the player's comprehensive performance index I t The comprehensive performance index is calculated by dynamic weighted average. Assuming the weight coefficients are α, β, γ, and δ respectively, the comprehensive performance index I t for: Among them, α+β+γ+δ=1, and α, β, γ, δ can be dynamically adjusted according to the player's performance.

4. The method for adaptively adjusting the difficulty of online entertainment games based on evolutionary algorithms according to claim 3, characterized in that: The S3 specifically includes: S31, select the current game difficulty and several candidate game difficulties as the initial population, and set the current game difficulty to be D c , the difficulty of the candidate games is {D1,D2,...,D k }, the population size of the initial population is set to N, including the current game difficulty and the candidate game difficulty, and the initial population is expressed as: P0={D c ,D1,D2,...,D k }; Among them, the initial population size N = k + 1; S32. Evaluate the fitness of the initial population. The evaluation criteria include the player's comprehensive performance index I t , game fluency G and player satisfaction S, the fitness function F is defined as: F(D i )=w1·I t (D i )+w2·G(D i )+w3·S(D i ); Among them, w1, w2, w3 are weight coefficients, satisfying w1+w2+w3=1, I t (D i ) is the comprehensive performance index of the player corresponding to the i-th game difficulty, G(D i ) is the game fluency score corresponding to the i-th game difficulty, S(D i ) is the player satisfaction score corresponding to the i-th game difficulty; S33, based on the fitness evaluation results, select several game difficulties with higher fitness to enter the crossover operation, and set the selection probability to be P s , the selection operation adopts the roulette selection mechanism, and the selection probability P s Defined as: Among them, F(D i ) is the fitness value of the i-th game difficulty, and N is the size of the initial population; S34, perform a cross operation on the difficulty of the selected candidate games, using a single-point crossover method to generate a new candidate game difficulty, assuming that the crossover point is C and the difficulty of the two selected games is D i and D j , the new game difficulty D′ after crossover is expressed as: D′=Crossover(D i ,D j ,C); Among them, k is the index of the game difficulty parameter, D i (k) and D j (k) are the parameter values ​​of the i-th and j-th game difficulties at index k, respectively, and C is the intersection position; S35, perform mutation operation on the difficulty of the candidate games generated by crossover, and set the mutation probability to be P m : D″=Mutate(D′,P m ); Where D″ is the game difficulty after mutation, D′ is the game difficulty after crossover, k is the index of the game difficulty parameter, Δ is a small random change value, rand(0,1) is a random number function generated between 0 and 1, P m is the preset mutation probability; S36. Re-evaluate the fitness of the candidate game difficulty after mutation, and compare it with the initial population, select several game difficulties with higher fitness to enter the next generation population, and assume that the candidate game difficulty after mutation is D″, and the fitness function is F: F(D″)=w1·I t (D″)+w2·G(D″)+w3·S(D″); Among them, w1, w2, w3 are weight coefficients, satisfying w1+w2+w3=1, I t (D″) is the comprehensive performance index of players corresponding to the difficulty of the game after mutation, G(D″) is the game fluency score corresponding to the difficulty of the game after mutation, and S(D″) is the player satisfaction score corresponding to the difficulty of the game after mutation; Select the game difficulty with higher fitness to enter the next generation population, and set the new generation population as P next : P next ={D i ∈P∪D″∣TopNbyF(D i )}; Among them, P is the initial population, D″ is the difficulty of the candidate game after mutation, and N is the size of the new generation population; S37, let the current generation be t, and the maximum evolution generation be T max , the fitness value convergence judgment condition is ΔF<∈, where ΔF is the fitness change and ∈ is the convergence threshold: ΔF=|F avg (P t+1 )-F avg (P t )|; Among them, F avg (P t ) is the average fitness value of the t-th generation population: Among them, if ΔF<∈ or t≥T max , then stop evolving and output the optimized game difficulty D * , D * is the individual with the highest fitness in the final population; S38, the optimized game difficulty D * Applied to the actual game, the game difficulty is adjusted in real time to adapt to the player's current ability and performance. The game difficulty after real-time adjustment is set to D real : D real (t+1)=α·D real (t)+(1-α)·D * ; Among them, α is the adjustment smoothing coefficient, which ranges from 0 to 1 and is used to control the smoothness of real-time adjustment.

5. The method for adaptively adjusting the difficulty of online entertainment games based on evolutionary algorithms according to claim 4, characterized in that: The S4 specifically includes: S41, monitoring the player's game performance in real time through an adaptive feedback control system, wherein the game performance includes the player's score S, game time T, operation speed V, and number of erroneous operations E; S42, the monitored player performance data is associated with the current game difficulty to construct a real-time feedback matrix F, the matrix element F ij Represents the real-time feedback value of the jth data in the i-th time period: S43. Calculate the comprehensive performance index I of the real-time feedback matrix F f : in, and are the average values ​​of player score, game time, operation speed and number of wrong operations respectively, α, β, γ, δ are weight coefficients, satisfying α+β+γ+δ=1; S44. Comprehensive Performance Index I f Perform dynamic adjustments, set the adjustment coefficient to K, and adjust the game difficulty D based on the real-time feedback from players: D new =D current +K·(I target -I f ); Among them, D current is the current game difficulty, D new The game difficulty is adjusted. target is the target comprehensive performance index; S45. Introduce an integral control mechanism into the real-time feedback control system, and set the integral time as T i : Among them, ΔD is the integral control adjustment amount, and t is the time; S46, combining proportional control and integral control to generate the final adjusted game difficulty D final , and the final adjusted game difficulty D final Applied to the game: D final =D new +ΔD。 6. The method for adaptively adjusting the difficulty of online entertainment games based on evolutionary algorithms according to claim 5, characterized in that: The S5 specifically includes: S51. Predicting the player's ability and behavior pattern using real-time data analysis and machine learning technology, wherein the player data includes score S, game time T, operation speed V, and number of erroneous operations E; S52. Input the player data into the prediction model. The prediction model uses the long short-term memory (LSTM) neural network to construct the input sequence X and output sequence Y of the prediction model: Y=[S n+1 T n+1 V n+1 E n+1 ]; S53. Train the LSTM model and use the time series cross-validation method to evaluate the model performance. The loss function is defined as the mean square error (MSE): Among them, Y i is the true value, is the predicted value; S54. Use the trained LSTM model to predict the player's future performance and generate a prediction value in Including future scores Game duration Operation speed and number of misoperations S55. According to the prediction results, the game difficulty is adjusted in advance, and the adjustment coefficient is set to K′: D pred =D current +K′·(I target -I pred ); Among them, D current is the current game difficulty, D pred To predict the adjusted game difficulty, I target is the target comprehensive performance index, I pred According to the predicted value Calculated Composite Performance Index; S56. In the actual game process, the real-time feedback adjustment and prediction adjustment are combined to generate the final game difficulty D final : D final =β·D real +(1-β)·D pred ; Among them, D real The game difficulty is adjusted according to real-time feedback, and β is the weight coefficient, ranging from 0 to 1, which is used to balance the real-time adjustment and predictive adjustment.

7. The method for adaptively adjusting the difficulty of online entertainment games based on evolutionary algorithms according to claim 6, characterized in that: The S6 specifically includes: S61. In the selection, crossover and mutation process of the evolutionary algorithm, multiple candidate game difficulties are processed in parallel using multi-threading technology. Assume that the number of threads for parallel processing is M, and the set of candidate game difficulties processed by each thread is {D i1 ,D i2 ,...,D ik }: Among them, P t is the t-th generation population; S62, in the fitness evaluation process, GPU acceleration technology is used to parallel calculate the fitness function, the fitness function F(D i )for: F(D i )=w1·I t (D i )+w2·G(D i )+w3·S(D i ); Among them, D i is the difficulty of the i-th candidate game, w1, w2, w3 are weight coefficients, satisfying w1+w2+w3=1, I t (D i ) is the player's comprehensive performance index, G(D i ) is the game fluency score, S(D i ) to rate player satisfaction; S63. In the crossover operation and mutation operation, the distributed computing architecture is used for parallel processing. Assume that the number of distributed computing nodes is N, and the set of crossover and mutation operations processed by each node is {C i1 ,C i2 ,...,C ik } and {M i1 ,M i2 ,...,M ik }: Among them, P t+1 is the t+1 generation population, Crossover(D i1 ,D i2 ,C ij ) is the jth crossover operation in the i-th node, Mutate(D ij ,M ij ) is the jth mutation operation in the i-th node; S64, introduce a dynamic load balancing mechanism to dynamically adjust task allocation according to the real-time load of each computing node. Suppose the node load is L i , the total load is L total : L i,new =L i +ΔL i ; in, Balanced distribution of computing resources is achieved by dynamically adjusting the load of each node; S65. At each stage of the evolution process, optimize the allocation strategy of computing resources by monitoring computing performance indicators: Among them, T compute is the total computation time, is the utilization of the i-th GPU, is the utilization rate of the i-th CPU thread, by optimizing P eff Improve overall computing efficiency.

Citation Information

Cited By

  • Mobile game dynamic balancing method and system based on behavior analysis and performance self-adaption

    CN121003818A

  • Method and system for dynamic balance of mobile games based on behavior analysis and performance adaptation

    CN121003818B