Game strategy performance prediction model construction method and device, equipment and medium

Abstract

The invention discloses a game strategy performance prediction model construction method and device, equipment and a medium, which are applied to the technical field of game strategy evaluation, and the method comprises the steps: determining capability vectors of a game strategy corresponding to different capability dimensions based on game strategy coding and a strategy capability learnable parameter matrix; based on the test scene code and the difficulty learnable parameter matrix, the discrimination learnable parameter matrix and the random coefficient learnable parameter matrix, determining difficulty, discrimination and a random coefficient; constructing a performance prediction model for predicting the performance of the game strategy based on the capability vector, the difficulty, the distinction degree and the random coefficient corresponding to the game strategy and the capability vector of the test scene; the capability vector corresponding to the test scene is determined by the capability dimension corresponding to the test scene. According to the invention, the performance prediction model constructed based on the multi-dimensional capability can quantify the multi-dimensional capability of the game strategy and the key attribute of the test scene at the same time, thereby achieving the precise prediction of the performance of the game strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of game strategy evaluation, and in particular to a method, device, equipment and medium for constructing a game strategy performance prediction model. Background Art

[0002] The current development of multi-agent game strategies faces a lack of an evaluation system. Due to the lack of unified standards for environment settings, task complexity, and evaluation metrics, existing research results struggle to form objective and consistent evaluations for performance comparison and validation, limiting the generalization and system optimization of related algorithms.

[0003] It can be seen that how to improve the accuracy of predicting the performance of game strategies is a technical problem that those skilled in the art urgently need to solve. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for constructing a game strategy performance prediction model, which solves the technical problem of low prediction accuracy of game strategy performance in the prior art.

[0005] To solve the above technical problems, the present invention provides a method for constructing a game strategy performance prediction model, comprising:

[0006] Performing feature coding processing on the game strategy and the test scenario to obtain the game strategy code and the test scenario code;

[0007] Determining the ability vectors of the game strategy corresponding to different ability dimensions based on the game strategy encoding and the strategy ability learnable parameter matrix;

[0008] Determine the difficulty, discrimination, and random coefficient based on the test scenario code and the difficulty learnable parameter matrix, the discrimination learnable parameter matrix, and the random coefficient learnable parameter matrix; the difficulty is a parameter that measures the strength of the test scenario's requirements for the game strategy; the discrimination is a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; and the random coefficient represents the probability of success in the test scenario without the guidance of the game strategy.

[0009] A performance prediction model for predicting the performance of a game strategy is constructed based on the ability vector corresponding to the game strategy, the difficulty, the discrimination, the random coefficient, and the ability vector of the test scenario; the ability vector corresponding to the test scenario is determined by the ability dimension corresponding to the test scenario.

[0010] Optionally, after constructing a performance prediction model for predicting the performance of a game strategy based on the capability vector, the difficulty, the discrimination, the random coefficient, and the capability vector of a test scenario corresponding to the game strategy, the method further includes:

[0011] Running the game strategy a preset number of times in each test scenario to obtain a predicted number of real performance data;

[0012] Determining comprehensive real performance data of the gaming strategy in each test scenario based on the preset number of real performance data;

[0013] Determine predicted performance data based on the performance prediction model, construct a loss function based on the predicted performance data and comprehensive real performance data corresponding to the predicted performance data, optimize the performance prediction model by minimizing the loss function, and obtain an optimized performance prediction model.

[0014] Optionally, before constructing a performance prediction model for predicting the performance of a game strategy based on the ability vector, the difficulty, the discrimination, the random coefficient, and the ability vector of a test scenario corresponding to the game strategy, the method further includes:

[0015] Divide the environment of the game task into test scenarios according to the set target capability dimension to obtain a target test scenario set, and construct a scenario-capability mapping matrix based on the target test scenario set and the target capability dimension;

[0016] Determining a priori capability vector corresponding to each of the test scenarios based on the scenario-capability mapping matrix;

[0017] Extracting features of the test scenario using multimodal data, and fusing the extracted multimodal features to obtain a multimodal vector of the scenario;

[0018] Obtaining a modified capability vector based on a scenario-learnable parameter matrix and a multimodal vector of the scenario;

[0019] The modified capability vector and the priori capability vector are fused to obtain the capability vector of the test scenario.

[0020] Optionally, the environment of the gaming task is divided into test scenarios according to a set target capability dimension to obtain a target test scenario set, and a scenario-capability mapping matrix is constructed based on the target test scenario set and the target capability dimension, including:

[0021] Adding disturbance factors to each target test scenario set to obtain test scenario variants;

[0022] The scenario-capability mapping matrix is constructed based on the test scenario variants and the target capability dimensions.

[0023] Optionally, the multimodal data includes image features in the test scenario and text features in the test scenario.

[0024] Optionally, after constructing a performance prediction model for predicting the performance of a game strategy based on the capability vector, the difficulty, the discrimination, the random coefficient, and the capability vector of a test scenario corresponding to the game strategy, the method further includes:

[0025] Determine all the game strategies corresponding to the current game task, determine the comprehensive ability score based on the ability dimension corresponding to each game strategy, and use the game strategy with the highest comprehensive ability score as the anchor game strategy;

[0026] Determine the comprehensive real performance data of the anchor game strategy under the new test environment, and fine-tune the performance prediction model based on the comprehensive real performance data corresponding to the anchor game strategy to obtain a fine-tuned performance prediction model, so that the fine-tuned performance prediction model can achieve cross-scenario performance prediction.

[0027] An embodiment of the present invention further provides a method for predicting gaming strategy performance, comprising:

[0028] Determine a target test scenario corresponding to the game strategy to be tested, and encode the game strategy to be tested and the target test scenario based on the feature code to obtain a code of the game strategy to be tested and a code of the target test scenario;

[0029] The game strategy code to be tested and the target test scenario code are used as inputs of a performance prediction model to obtain the target game strategy performance of the game strategy to be tested under the target test scenario; wherein, the performance prediction model is a model obtained based on the above-mentioned game strategy performance prediction model construction method.

[0030] An embodiment of the present invention further provides a device for constructing a game strategy performance prediction model, comprising:

[0031] The encoding module is used to perform feature encoding processing on the game strategy and the test scenario to obtain the game strategy code and the test scenario code;

[0032] A capability vector determination module, configured to determine capability vectors corresponding to different capability dimensions of a game strategy based on the game strategy code and the strategy capability learnable parameter matrix;

[0033] a parameter determination module for determining difficulty, discrimination, and a random coefficient based on the test scenario code and a difficulty learnable parameter matrix, a discrimination learnable parameter matrix, and a random coefficient learnable parameter matrix; the difficulty being a parameter that measures the strength of the test scenario's requirements for the game strategy; the discrimination being a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; and the random coefficient representing the probability of success in the test scenario without the guidance of a game strategy;

[0034] A performance prediction model construction module is used to construct a performance prediction model for predicting the performance of a game strategy based on the ability vector corresponding to the game strategy, the difficulty, the discrimination, the random coefficient and the ability vector of the test scenario; the ability vector corresponding to the test scenario is determined by the ability dimension corresponding to the test scenario.

[0035] An embodiment of the present invention further provides an electronic device, including:

[0036] memory for storing computer programs;

[0037] A processor is configured to execute the computer program to implement the steps of the above method.

[0038] An embodiment of the present invention further provides a medium (ie, a computer-readable storage medium) on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0039] An embodiment of the present invention provides a computer program product, including a computer program / instructions, which implements the steps of the above method when executed by a processor.

[0040] As can be seen, the present invention obtains game strategy encoding and test scenario encoding by performing feature encoding processing on game strategies and test scenarios; determines the ability vector of the game strategy corresponding to different ability dimensions based on the game strategy encoding and the strategy ability learnable parameter matrix; determines the difficulty, discrimination and random coefficient based on the test scenario encoding and the difficulty learnable parameter matrix, the discrimination learnable parameter matrix and the random coefficient learnable parameter matrix; difficulty is a parameter that measures the strength of the test scenario's requirements for the game strategy's ability; discrimination is a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; the random coefficient represents the probability of success in the test scenario without the guidance of the game strategy; constructs a performance prediction model for predicting the performance of the game strategy based on the ability vector, difficulty, discrimination and random coefficient corresponding to the game strategy and the ability vector of the test scenario; the ability vector corresponding to the test scenario is determined by the ability dimension corresponding to the test scenario. Compared with the current macro-level evaluation of game strategies, the performance prediction model constructed by considering multi-dimensional capabilities can simultaneously quantitatively evaluate the multi-dimensional capabilities of the game strategy and the key attributes of the test scenario, realizing a detailed evaluation of the game strategy and providing effective guidance for the improvement of the game strategy.

[0041] In addition, the present invention also provides a device, equipment and medium for constructing a game strategy performance prediction model, which also has the above-mentioned beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0043] Figure 1 A flowchart of a method for constructing a game strategy performance prediction model provided by an embodiment of the present invention;

[0044] Figure 2 A flowchart illustrating a method for constructing a game strategy performance prediction model according to an embodiment of the present invention;

[0045] Figure 3 A flowchart illustrating a method for predicting gaming strategy performance according to an embodiment of the present invention;

[0046] Figure 4 A schematic diagram of the structure of a device for constructing a game strategy performance prediction model provided by an embodiment of the present invention;

[0047] Figure 5 A schematic diagram of the structure of a gaming strategy performance prediction device provided by an embodiment of the present invention;

[0048] Figure 6 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0050] With the rapid development of deep learning technology, reinforcement learning (RL) has demonstrated remarkable performance in decision-making problems such as gaming, Go, and robot control. Building on this success, multi-agent reinforcement learning (MARL) has emerged as a research hotspot to address more complex collaborative and competitive scenarios. Based on game theory, this field studies the interaction mechanisms and policy optimization of multiple agents in a shared environment, and has been widely applied in key areas such as intelligent transportation systems and energy scheduling. However, the current development of multi-agent game strategies still faces the challenge of a lack of an evaluation system. Due to the lack of unified standards for environment settings, task complexity, and evaluation metrics, existing research results struggle to form objective and consistent evaluations for performance comparison and validation, hindering the generalization and system optimization of related algorithms.

[0051] To address these issues, researchers have gradually explored various evaluation methods. Common approaches include comparative analysis of algorithms based on quantitative metrics such as win rate and average return value. Some studies have introduced meta-game evaluation frameworks to measure the generalization capabilities of multi-agent reinforcement learning algorithms. In addition, some evaluation benchmark platforms provide standard environments and test sets, facilitating the comparison of algorithms under uniform conditions. However, overall, existing methods still rely on macro-performance indicators, making it difficult to reveal the multi-dimensional key capabilities of multi-agent game strategies at a fine-grained level. This results in a lack of interpretability in evaluation results and makes it difficult to provide effective guidance for subsequent algorithm improvements.

[0052] The purpose of this invention is to provide a multi-agent game strategy evaluation method based on cognitive diagnosis, which can simultaneously quantitatively evaluate the multidimensional capabilities of the game strategy and the key attributes of the test scenario, and further realize the prediction of the generalization performance of the game algorithm in the new environment.

[0053] Please refer to Figure 1 , Figure 1 A flowchart of a method for constructing a game strategy performance prediction model provided by an embodiment of the present invention. The method may include:

[0054] S101, perform feature coding processing on the game strategy and the test scenario to obtain the game strategy code and the test scenario code.

[0055] The executor of this embodiment is an electronic device. The electronic device can specifically be a server, a portable terminal or other forms. The game strategy in this embodiment is generally multiple. When there are multiple, the game tasks corresponding to each game strategy are consistent. This embodiment does not limit the specific game strategy. For example, the game strategy can be a basic encirclement strategy, a role division strategy, a disturbance response strategy, an attack strategy, etc. This embodiment does not limit the specific test scenario. For example, the test scenario can be a synergy test scenario, an offensive ability test scenario, a dribbling ability test scenario, a defensive ability test scenario, etc. In this embodiment, one game task corresponds to multiple game strategies, and each game strategy can correspond to multiple ability dimensions. The direction of each ability test is different, and the corresponding test scenario can be determined based on the ability dimension. This embodiment does not limit the specific feature encoding process. For example, this embodiment can perform a one-hot encoding process, or this embodiment can also be a binary encoding, as long as the representation of the two parameters can be unified.

[0056] S102, determining the capability vectors of the game strategy corresponding to different capability dimensions based on the game strategy encoding and the strategy capability learnable parameter matrix.

[0057] The strategy capability learnable parameter matrix in this embodiment is a learnable parameter matrix in the model, which may have an initial value. As the model is continuously updated, the learnable parameter matrix is updated. For example, the strategy capability learnable parameter matrix in this embodiment may be , It is a parameter matrix used to learn strategic capabilities. At this time, the capability vector of the game strategy is , where d represents the capability dimension corresponding to the game strategy, represents the set of real numbers. . , is the encoding representation of the game strategy. t indicates that the number of game strategies is t. The different capability dimensions of this embodiment refer to the capabilities corresponding to the game strategy.

[0058] S103, determine the difficulty, discrimination and random coefficient based on the test scenario encoding and the difficulty learnable parameter matrix, the discrimination learnable parameter matrix and the random coefficient learnable parameter matrix; the difficulty is a parameter that measures the strength of the test scenario's requirements for the game strategy; the discrimination is a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; the random coefficient represents the probability of success in the test scenario without the guidance of the game strategy.

[0059] The difficulty learning parameter matrix, discrimination learning parameter matrix and random coefficient learning parameter matrix in this embodiment are custom designed according to the requirements. The difficulty corresponding to the scenario is , the discrimination is And the random coefficient is The difficulty of the scenario is used to measure the strength of the test scenario's requirements for game strategy ability, and the difficulty of the scenario is characterized according to the game strategy; the discrimination reflects the effectiveness of the scenario in distinguishing different strategy levels, and the random coefficient represents the probability of success in the scenario without strategy guidance. The above process can be formally expressed as follows:

[0060] ;in 、 are the one-hot encodings of the game strategy and test scenario, respectively. 、 、 They represent the difficulty learnable parameter matrix, the discrimination learnable parameter matrix and the random coefficient learnable parameter matrix respectively. represents the hyperbolic tangent function.

[0061] S104, constructing a performance prediction model for predicting the performance of the game strategy based on the ability vector, difficulty, discrimination and random coefficient corresponding to the game strategy and the ability vector of the test scenario; the ability vector corresponding to the test scenario is determined by the ability dimension corresponding to the test scenario.

[0062] In this embodiment, the ability vector corresponding to the test situation is determined by the ability dimension corresponding to the test scenario, which means that the current target ability dimension includes d abilities, and the ability dimension corresponding to the current test scenario is part or all of the d abilities. This embodiment does not limit the specific method of constructing a performance prediction model for predicting the performance of a game strategy based on the ability vector, difficulty, discrimination, and random coefficient corresponding to the game strategy and the ability vector of the test scenario. For example, this embodiment can construct a performance prediction model for the performance of a game strategy based on the IRT (Item Response Theory) formula (which can be a success rate, time step, etc.). It can be formally expressed as: ;in represents element-wise multiplication, represents a multilayer perceptron with two hidden layers with ReLU activation function and BN (batch normalization), and an output layer with only ReLU activation function. q represents the ability vector of the test scenario.

[0063] It should be further explained that, based on any of the above embodiments, in order to improve the accuracy of determining the capability vector of the test scenario, before constructing the performance prediction model for predicting the performance of the game strategy based on the capability vector, difficulty, discrimination, and random coefficient corresponding to the game strategy and the capability vector of the test scenario, the following steps may be further included:

[0064] S1, divide the test scenarios of the game task environment according to the set target capability dimension to obtain the target test scenario set, and construct the scenario-capability mapping matrix based on the target test scenario set and the target capability dimension.

[0065] This embodiment sets the target capability dimension based on the actual situation of the current game task and prior knowledge, and the target capability dimension includes capabilities of different capability dimensions. In this embodiment, each capability corresponds to a test scenario. According to the capability dimension to be evaluated, the environment of the game task is divided into scenarios, and a scenario set for testing the performance of the game strategy is constructed. In this embodiment of the present invention, for a specific game task, , according to the evaluation The task environment is decomposed into different ability dimensions, and several basic test scenarios are constructed to independently examine each ability.

[0066] S2, determine the prior ability vector corresponding to each test scenario based on the scenario-ability mapping matrix.

[0067] This embodiment is based on the scenario-capability mapping matrix Get the test scenario The prior ability vector , formally expressed as follows: , T represents transpose, Indicates test scenario All the corresponding prior ability vectors. It is understandable that the prior ability vectors corresponding to different test scenarios will be different. and scenarios Take the input as an example.

[0068] S3, using multimodal data to extract features of the test scenario, and fusing the extracted multimodal features to obtain a multimodal vector of the scenario.

[0069] This embodiment uses multimodal data to extract features from the test scenario and generate a multimodal vector of the scenario. This embodiment does not limit the specific multimodal data. For example, the multimodal data may include a combination of images, text, video, voice, etc. (for example, the multimodal data includes image features in the test scenario and text features in the test scenario). The scenario is feature modeled through multimodal embedding guided by a large model. For example, CNN (convolutional neural network) is used to extract image features in the scenario (such as the terrain of the roundup task, the chess game of the Go task, etc.) to generate an image embedding vector. At the same time, the multimodal large model is used to encode the situation description text of the scene and obtain the corresponding text embedding vector Etc. In order to fuse multimodal features, the features of different modalities are first projected into a unified common embedding space, and then a token sequence containing the embeddings of each modality is constructed and input into the Transformer encoder for feature fusion. Finally, a unified representation vector of the scene is obtained by performing an average pooling operation on the Transformer output. Token usually refers to the basic unit in the input text sequence, which can be a single word, a subword, or an encoded representation of an image in visual tasks. The Transformer encoder is the core part of the deep learning model for processing sequence data, especially playing an important role in natural language processing tasks. The above process can be formally expressed as follows:

[0070] ;in, is a linear mapping, Represents position encoding, i represents the column vector, and “.” represents all parameters in this column vector.

[0071] S4, obtains the modified capability vector based on the scenario learnable parameter matrix and the scenario multimodal vector.

[0072] This embodiment takes into account that although the correspondence between test scenarios and capabilities has been established based on expert knowledge in the previous steps, in actual applications, there may still be a small amount of deviation in the ability classification. In addition, a single test scenario may involve multiple ability dimensions at the same time. Therefore, it is necessary to further learn the ability vector of the test scenario. , by introducing the scenario-learnable parameter matrix Get the corrected capability vector , formally expressed as follows: .

[0073] S5, fusing the modified capability vector and the prior capability vector to obtain the capability vector of the test scenario.

[0074] In this embodiment of the present invention, the two capability vectors are further fused and normalized through a Softmax layer to obtain the capability vector of the final test scenario, which is formally expressed as follows: .

[0075] It should be further explained that, based on any of the above embodiments, in order to improve the accuracy of constructing the scenario-capability mapping matrix, the above-mentioned test scenario division of the gaming task environment according to the set target capability dimension is performed to obtain a target test scenario set, and the scenario-capability mapping matrix is constructed based on the target test scenario set and the target capability dimension, which may include: adding a disturbance factor to each target test scenario set to obtain a test scenario variant; and constructing a scenario-capability mapping matrix based on the test scenario variant and the target capability dimension. This embodiment can further introduce a disturbance factor (such as position disturbance in the roundup task) to obtain a total of Several similar test scenario variations Among them, the a priori association between each test scenario and capability dimension is represented by the scenario-ability mapping matrix Represents that the matrix elements Indicates the situation Involving A capability dimension.

[0076] It should be further explained that, in order to improve the accuracy of the performance prediction model, after constructing the performance prediction model for predicting the performance of the game strategy based on the ability vector, difficulty, discrimination, and random coefficient corresponding to the game strategy and the ability vector of the test scenario, it may also include:

[0077] Step 1: Run the game strategy a preset number of times in each test scenario to obtain a predicted number of real performance data.

[0078] This embodiment can optimize the performance prediction model based on the loss function after the performance prediction model is initially constructed, thereby improving the accuracy of the performance prediction model prediction.

[0079] Step 2: Determine the comprehensive real performance data of the gaming strategy in each test scenario based on a preset number of real performance data.

[0080] This embodiment does not limit the specific method of determining the comprehensive real performance data. For example, this embodiment can use a single success or failure as the performance test result, in which case the loss function in step 3 should use the cross entropy function; or A trained multi-agent game strategy , run independently in each test scenario The average performance of each game strategy in this test scenario is counted as the comprehensive real performance data.

[0081] Step 3: Determine the predicted performance data based on the performance prediction model, construct a loss function based on the predicted performance data and the comprehensive real performance data corresponding to the predicted performance data, optimize the performance prediction model by minimizing the loss function, and obtain an optimized performance prediction model.

[0082] In the embodiment of the present invention, in order to use the optimization algorithm to reduce the prediction error of the performance prediction model, the mean square error between the prediction result (the result of performance prediction based on the performance prediction model) and the actual result (the comprehensive actual performance data) is constructed as the loss function, which is formally expressed as follows: ; By minimizing this loss function, continuous optimization and updating of strategic capabilities and situational attributes can be achieved.

[0083] It should be further explained that, based on any of the above embodiments, in order to further improve the accuracy of performance prediction, after constructing a performance prediction model for predicting the performance of a game strategy based on the ability vector, difficulty, discrimination, and random coefficient corresponding to the game strategy and the ability vector of the test scenario, the following may also be included:

[0084] Determine all the game strategies corresponding to the current game task, determine the comprehensive ability score based on the ability dimension corresponding to each game strategy, and use the game strategy with the highest comprehensive ability score as the anchor game strategy;

[0085] Determine the comprehensive real performance data of the anchor game strategy in the new test environment, and fine-tune the performance prediction model based on the comprehensive real performance data corresponding to the anchor game strategy to obtain a fine-tuned performance prediction model, so that the fine-tuned performance prediction model can achieve cross-scenario performance prediction.

[0086] In the embodiment of the present invention, based on the multi-dimensional capability vector of the game strategy, in order to measure each capability dimension and select a representative anchor game strategy , it is necessary to calculate the ability score of each game strategy , where d is each capability, The capabilities of different capability dimensions corresponding to the game strategy can be compressed to the (0, 1) interval by the Sigmoid function, where the anchor game strategy is defined as the strategy with the highest score, i.e. , Represents continuous multiplication, and the entire formula represents geometric mean. Based on the performance prediction model or the optimized performance prediction model, the game strategy is used Based on the real performance data in the new test environment, the prediction head of the model is fine-tuned to adapt to the new environment distribution. Adjusted to: ,in and Finally, the fine-tuned model can be used to predict the cross-environment generalization performance of other game strategies in the target environment.

[0087] A method for constructing a game strategy performance prediction model provided by an embodiment of the present invention may include: S101, performing feature encoding processing on the game strategy and the test scenario to obtain the game strategy code and the test scenario code; S102, determining the ability vector of the game strategy corresponding to different ability dimensions based on the game strategy code and the strategy ability learnable parameter matrix; S103, determining the difficulty, discrimination and random coefficient based on the test scenario code and the difficulty learnable parameter matrix, the discrimination learnable parameter matrix and the random coefficient learnable parameter matrix; the difficulty is a parameter that measures the intensity of the requirements of the test scenario on the ability of the game strategy; the discrimination is a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; the random coefficient represents the probability of success in the test scenario without the guidance of the game strategy; S104, constructing a performance prediction model for predicting the performance of the game strategy based on the ability vector, difficulty, discrimination and random coefficient corresponding to the game strategy and the ability vector of the test scenario; the ability vector corresponding to the test scenario is determined by the ability dimension corresponding to the test scenario. The beneficial effects of the present invention are: compared with the current macro-level evaluation of game strategies, this application is based on a performance prediction model constructed based on multi-dimensional capabilities, which can simultaneously quantitatively evaluate the multi-dimensional capabilities of game strategies and key attributes of test scenarios, thereby realizing a detailed evaluation of game strategies, thereby providing effective guidance for the improvement of game strategies, and can improve the effect of users selecting game strategies for gaming.

[0088] In order to make the present invention easier to understand, please refer to Figure 2 , Figure 2 A flowchart illustrating a method for constructing a game strategy performance prediction model provided by an embodiment of the present invention may specifically include:

[0089] S201, determining the target capability dimension to be evaluated according to the characteristics of the game task, dividing the environment of the game task into scenarios according to the target capability dimension, and obtaining a target test scenario set.

[0090] This embodiment divides the environment into scenarios based on the specific characteristics of the task and expert prior knowledge, creating several representative test scenarios. For example, if the game task is a roundup, the corresponding ability dimensions include search ability, coordination ability, and finishing ability (for example, the search ability test scenario is different initial positions, the coordination ability scenario is the enemy's coordination in the encirclement, and the finishing ability scenario is the division of labor to quickly complete the roundup). The corresponding game can be considered a football match, and the abilities included include coordination, offense, dribbling, defense, and interception.

[0091] S202 , constructing a scenario-capability mapping matrix based on the target test scenario set and the target capability dimension, and determining a priori capability vector corresponding to each test scenario based on the scenario-capability mapping matrix.

[0092] The target set of test scenarios in this embodiment incorporates perturbations (e.g., positional perturbations in a roundup mission) into each test scenario. This embodiment can introduce random perturbations, including initial states and agent constraints, to construct test scenarios for multi-agent game strategies. By running the game strategies in these test scenarios, performance data can be collected under different test scenarios.

[0093] S203 , extracting features of the test scenario using the multimodal data, and fusing the extracted multimodal features to obtain a multimodal vector of the scenario.

[0094] This embodiment introduces multimodal data such as images and text, extracts features of test scenarios from different dimensions, and completes feature fusion in a common embedding space to generate a unified representation vector for each test scenario.

[0095] S204 , obtaining a modified capability vector based on the scenario-learnable parameter matrix and the scenario's multimodal vector, and fusing the modified capability vector with the prior capability vector to obtain a capability vector for the test scenario.

[0096] S205 , performing feature coding processing on the game strategy and the test scenario to obtain the game strategy code and the test scenario code, and determining the ability vectors of the game strategy corresponding to different ability dimensions based on the game strategy code and the strategy ability learnable parameter matrix.

[0097] In this embodiment, there may be multiple game strategies and test scenarios corresponding to each game strategy.

[0098] S206 , determining the difficulty, discrimination and random coefficient based on the test scenario code and the difficulty learnable parameter matrix, the discrimination learnable parameter matrix and the random coefficient learnable parameter matrix.

[0099] This paper designs a multi-dimensional capability evaluation framework for gaming strategies. Based on the performance test results of gaming strategies, the performance data is fitted, and the performance of gaming strategies in multiple dimensions is quantified by updating the framework parameters. At the same time, key attributes such as the difficulty and discrimination of the test scenarios are evaluated.

[0100] S207 , constructing a performance prediction model for predicting the performance of the game strategy based on the ability vector, difficulty, discrimination, and random coefficient corresponding to the game strategy and the ability vector of the test scenario.

[0101] S208 , running the game strategy n times in each test scenario to obtain n real performance data, and determining comprehensive real performance data of the game strategy in each test scenario based on the n real performance data.

[0102] S209, determining predicted performance data based on the performance prediction model, constructing a loss function based on the predicted performance data and the comprehensive real performance data, and optimizing the performance prediction model by minimizing the loss function to obtain an optimized performance prediction model.

[0103] S210, determining all game strategies corresponding to the current game task, determining a comprehensive ability score based on the ability dimension corresponding to each game strategy, and using the game strategy with the highest comprehensive ability score as the anchor game strategy.

[0104] S211, determining the comprehensive real performance data of the anchor game strategy in the new test environment, and fine-tuning the prediction head of the performance prediction model based on the comprehensive real performance data corresponding to the anchor game strategy to obtain a fine-tuned performance prediction model.

[0105] The present invention can use a small number of test data from game strategies in a new test environment to fine-tune the framework and predict the generalization ability of game strategies in unknown or changing environments. The application scenario for performance prediction based on the fine-tuned performance prediction model in this embodiment can be a real-time strategy game based on multiple agents, thereby determining the game strategy with the best performance from the game strategies corresponding to the application scenario based on the performance prediction model. The main purpose of the present invention is to form a unified strategy capability assessment under several settings, predict the performance of game strategies across environments, and improve the accuracy of performance prediction.

[0106] As can be seen from the technical solution provided by the present invention, expert prior knowledge is used to divide the game task environment into scenarios, and a perturbation mechanism is used to generate an initial environment for testing the performance of game strategies. Game strategies are tested in multiple scenarios and corresponding performance results are obtained. A game strategy evaluation model is then used to assess the multi-dimensional capabilities of the game strategy and key environmental attributes. Finally, the model is fine-tuned using limited test data on the game strategy in new environments, thereby achieving efficient cross-environmental performance prediction of multi-agent game strategies.

[0107] In order to make the present invention easier to understand, please refer to Figure 3 , Figure 3 The flowchart of a method for predicting the performance of a gaming strategy provided by an embodiment of the present invention may specifically include:

[0108] S301 , determining a target test scenario corresponding to a game strategy to be tested, and encoding the game strategy to be tested and the target test scenario based on the feature coding to obtain a code of the game strategy to be tested and a code of the target test scenario.

[0109] This embodiment is to predict the performance of the game strategy to be tested in the target test scenario. This embodiment does not limit the specific feature encoding, and the feature encoding method is consistent with the encoding method used when constructing the performance prediction model.

[0110] S302, using the game strategy code to be tested and the target test scenario code as inputs of a performance prediction model to obtain the target game strategy performance of the game strategy to be tested in the target test scenario; wherein the performance prediction model is a model obtained based on the above-mentioned game strategy performance prediction model construction method.

[0111] The embodiments of the present invention can predict the performance of the game strategy to be tested in the target test scenario, so that it is possible to determine whether to adjust the game strategy based on the performance. When adjusting, the various capabilities corresponding to the game strategy to be tested can be determined, thereby accurately improving the game strategy.

[0112] The following is an introduction to the game strategy performance prediction model construction device provided by an embodiment of the present invention. The game strategy performance prediction model construction device described below and the game strategy performance prediction model construction method described above can be referenced to each other.

[0113] Please refer to Figure 4 , Figure 4 A schematic diagram of a device for constructing a game strategy performance prediction model provided by an embodiment of the present invention may include:

[0114] The encoding module 100 is used to perform feature encoding processing on the game strategy and the test scenario to obtain the game strategy code and the test scenario code;

[0115] A capability vector determination module 200 is configured to determine capability vectors corresponding to different capability dimensions of a game strategy based on the game strategy code and the strategy capability learnable parameter matrix;

[0116] Parameter determination module 300, configured to determine difficulty, discrimination, and random coefficient based on the test scenario code and a difficulty learnable parameter matrix, a discrimination learnable parameter matrix, and a random coefficient learnable parameter matrix; difficulty being a parameter that measures the strength of the test scenario's requirements for the game strategy; discrimination being a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; and random coefficient representing the probability of success in the test scenario without the guidance of a game strategy;

[0117] The performance prediction model construction module 400 is used to construct a performance prediction model for predicting the performance of the game strategy based on the ability vector corresponding to the game strategy, the difficulty, the discrimination, the random coefficient and the ability vector of the test scenario; the ability vector corresponding to the test scenario is determined by the ability dimension corresponding to the test scenario.

[0118] Furthermore, based on any of the above embodiments, the gaming strategy performance prediction model construction device may further include:

[0119] A real performance data determination module is used to run the game strategy in each test scenario for a preset number of times to obtain a predicted number of real performance data;

[0120] a comprehensive real performance data determination module, configured to determine the comprehensive real performance data of the gaming strategy in each test scenario based on the preset number of real performance data;

[0121] An optimization model module is used to determine predicted performance data based on the performance prediction model, construct a loss function based on the predicted performance data and the comprehensive real performance data corresponding to the predicted performance data, and optimize the performance prediction model by minimizing the loss function to obtain an optimized performance prediction model.

[0122] Furthermore, based on any of the above embodiments, the gaming strategy performance prediction model construction device may further include:

[0123] A scenario-capability mapping matrix construction module is used to divide the gaming task environment into test scenarios according to a set target capability dimension, obtain a target test scenario set, and construct a scenario-capability mapping matrix based on the target test scenario set and the target capability dimension;

[0124] A priori capability vector determination module, configured to determine the priori capability vector corresponding to each of the test scenarios based on the scenario-capability mapping matrix;

[0125] A scenario multimodal vector determination module is used to extract features of the test scenario using multimodal data and fuse the extracted multimodal features to obtain a scenario multimodal vector;

[0126] a modified capability vector determination module, configured to obtain a modified capability vector based on a scenario-learnable parameter matrix and a multimodal vector of the scenario;

[0127] The capability vector determination module of the test scenario is used to fuse the modified capability vector and the priori capability vector to obtain the capability vector of the test scenario.

[0128] Furthermore, based on any of the above embodiments, the scenario-capability mapping matrix construction module may include:

[0129] a test scenario variant determining unit, configured to add a disturbance factor to each target test scenario set to obtain a test scenario variant;

[0130] A scenario-capability mapping matrix construction unit is used to construct the scenario-capability mapping matrix based on the test scenario variants and the target capability dimension.

[0131] Further, based on any of the above embodiments, the multimodal data includes image features in the test scenario and text features in the test scenario.

[0132] Furthermore, based on any of the above embodiments, the gaming strategy performance prediction model construction device may further include:

[0133] Anchor game strategy determination module, used to determine all game strategies corresponding to the current game task, determine the comprehensive ability score based on the ability dimension corresponding to each game strategy, and use the game strategy with the highest comprehensive ability score as the anchor game strategy;

[0134] The fine-tuning model module is used to determine the comprehensive real performance data of the anchor game strategy under the new test environment, and fine-tune the performance prediction model based on the comprehensive real performance data corresponding to the anchor game strategy to obtain a fine-tuned performance prediction model, so that the fine-tuned performance prediction model can achieve cross-scenario performance prediction.

[0135] It should be noted that the order of the modules and units in the above-mentioned game strategy performance prediction model construction device can be changed without affecting the logic.

[0136] A device for constructing a game strategy performance prediction model provided by an embodiment of the present invention may include: an encoding module 100, used to perform feature encoding processing on a game strategy and a test scenario to obtain a game strategy code and a test scenario code; a capability vector determination module 200, used to determine the capability vector of the game strategy corresponding to different capability dimensions based on the game strategy code and a strategy capability learnable parameter matrix; a parameter determination module 300, used to determine the difficulty, discrimination and random coefficient based on the test scenario code and a difficulty learnable parameter matrix, a discrimination learnable parameter matrix and a random coefficient learnable parameter matrix; the difficulty is a parameter that measures the intensity of the test scenario's requirements for the game strategy's capability; the discrimination is a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; the random coefficient represents the probability of success in the test scenario without the guidance of the game strategy; a performance prediction model construction module 400, used to construct a performance prediction model for predicting the performance of the game strategy based on the capability vector, the difficulty, the discrimination and the random coefficient corresponding to the game strategy and the capability vector of the test scenario; the capability vector corresponding to the test scenario is determined by the capability dimension corresponding to the test scenario. The beneficial effects of the present invention are: compared with the current macro-level evaluation of game strategies, this application is based on a performance prediction model constructed based on multi-dimensional capabilities, which can simultaneously quantitatively evaluate the multi-dimensional capabilities of game strategies and key attributes of test scenarios, thereby realizing a detailed evaluation of game strategies, thereby providing effective guidance for the improvement of game strategies, and can improve the effect of users selecting game strategies for gaming.

[0137] The following is an introduction to the game strategy performance prediction device provided by an embodiment of the present invention. The game strategy performance prediction device described below and the game strategy performance prediction method described above can be referenced to each other.

[0138] Please refer to Figure 5 , Figure 5 A schematic diagram of the structure of a gaming strategy performance prediction device provided by an embodiment of the present invention may include:

[0139] It should be noted that the order of the modules and units in the above-mentioned game strategy performance prediction device can be changed without affecting the logic.

[0140] The game strategy performance prediction device provided by the embodiment of the present invention may include: a coding input module 500, which is used to determine the target test scenario corresponding to the game strategy to be tested, and encode the game strategy to be tested and the target test scenario based on the feature coding to obtain the game strategy to be tested code and the target test scenario code; a performance prediction module 600: using the game strategy to be tested code and the target test scenario code as inputs to a performance prediction model to obtain the target game strategy performance of the game strategy to be tested under the target test scenario; wherein the performance prediction model is a model obtained based on the above-mentioned game strategy performance prediction model construction method. The embodiment of the present invention can predict the performance of the game strategy to be tested in the target test scenario, so that it can be determined whether to adjust the game strategy based on the performance. When adjusting, the various capabilities corresponding to the game strategy to be tested can be determined, thereby accurately improving the game strategy.

[0141] An electronic device provided by an embodiment of the present invention is introduced below. The electronic device described below and the game strategy performance prediction model construction method and / or game strategy performance prediction method described above can be referenced to each other.

[0142] Please refer to Figure 6 , Figure 6 A schematic structural diagram of an electronic device provided in an embodiment of the present invention may include:

[0143] Memory 10, for storing computer programs;

[0144] The processor 20 is used to execute a computer program to implement the above-mentioned game strategy performance prediction model construction method and / or game strategy performance prediction method.

[0145] The memory 10 , the processor 20 , and the communication interface 30 all communicate with each other via a communication bus 40 .

[0146] In the embodiment of the present invention, the memory 10 is used to store one or more programs. The program may include program code, and the program code includes computer operation instructions. In the embodiment of the present invention, the memory 10 may store programs for implementing the following functions:

[0147] Performing feature coding processing on the game strategy and the test scenario to obtain the game strategy code and the test scenario code;

[0148] Determine the ability vectors of game strategies corresponding to different ability dimensions based on game strategy encoding and strategy ability learnable parameter matrix;

[0149] Based on the test scenario encoding and the difficulty learnable parameter matrix, the discrimination learnable parameter matrix, and the random coefficient learnable parameter matrix, the difficulty, discrimination, and random coefficient are determined; the difficulty is a parameter that measures the strength of the test scenario's requirements for the game strategy; the discrimination is a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; and the random coefficient represents the probability of success in the test scenario without the guidance of the game strategy.

[0150] A performance prediction model for predicting the performance of a game strategy is constructed based on the ability vector, difficulty, discrimination and random coefficient corresponding to the game strategy and the ability vector of the test scenario; the ability vector corresponding to the test scenario is determined by the ability dimension corresponding to the test scenario.

[0151] and / or:

[0152] Determine the target test scenario corresponding to the game strategy to be tested, and encode the game strategy to be tested and the target test scenario based on the feature coding to obtain the game strategy code to be tested and the target test scenario code.

[0153] The game strategy code to be tested and the target test scenario code are used as inputs of the performance prediction model to obtain the target game strategy performance of the game strategy to be tested in the target test scenario; wherein, the performance prediction model is a model obtained based on the above-mentioned game strategy performance prediction model construction method.

[0154] In one possible implementation, the memory 10 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and applications required for at least one function, etc.; the data storage area may store data created during use.

[0155] In addition, the memory 10 may include a read-only memory and a random access memory, and provides instructions and data to the processor. A portion of the memory may also include NVRAM. The memory stores an operating system and operating instructions, executable modules or data structures, or a subset or an extended set thereof. The operating instructions may include various operating instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and processing hardware-based tasks.

[0156] The processor 20 may be a central processing unit (CPU), an application-specific integrated circuit, a digital signal processor, a field programmable gate array, or other programmable logic device. The processor 20 may be a microprocessor or any conventional processor. The processor 20 may call a program stored in the memory 10 .

[0157] The communication interface 30 may be an interface of a communication module, and is used to connect to other devices or systems.

[0158] Of course, it needs to be explained that Figure 6 The structure shown does not constitute a limitation on the electronic device in the embodiment of the present invention. In actual applications, the electronic device may include Figure 6 More or fewer components than shown, or combinations of certain components.

[0159] The computer-readable storage medium provided in an embodiment of the present invention is introduced below. The computer-readable storage medium described below and the game strategy performance prediction model construction method and / or game strategy performance prediction method described above can be referenced to each other.

[0160] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the above-mentioned game strategy performance prediction model construction method and / or game strategy performance prediction method.

[0161] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., which can store program codes.

[0162] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Reference can be made to the descriptions of the identical or similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions of the methods.

[0163] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0164] Finally, it should be noted that, in this document, relationships such as first and second, etc., are used solely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0165] The above is a detailed introduction to the method, device, equipment and medium for constructing a game strategy performance prediction model provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for constructing a game strategy performance prediction model, characterized in that: include: Performing feature coding processing on the game strategy and the test scenario to obtain the game strategy code and the test scenario code; Determining the ability vectors of the game strategy corresponding to different ability dimensions based on the game strategy encoding and the strategy ability learnable parameter matrix; Determine the difficulty, discrimination, and random coefficient based on the test scenario code and the difficulty learnable parameter matrix, the discrimination learnable parameter matrix, and the random coefficient learnable parameter matrix; the difficulty is a parameter that measures the strength of the test scenario's requirements for the game strategy; the discrimination is a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; and the random coefficient represents the probability of success in the test scenario without the guidance of the game strategy. A performance prediction model for predicting the performance of a game strategy is constructed based on the ability vector corresponding to the game strategy, the difficulty, the discrimination, the random coefficient, and the ability vector of the test scenario; the ability vector corresponding to the test scenario is determined by the ability dimension corresponding to the test scenario.

2. The method for constructing a game strategy performance prediction model according to claim 1, wherein: After constructing a performance prediction model for predicting the performance of the game strategy based on the capability vector, the difficulty, the discrimination, the random coefficient, and the capability vector of the test scenario corresponding to the game strategy, the method further includes: Running the game strategy a preset number of times in each test scenario to obtain a predicted number of real performance data; Determining comprehensive real performance data of the gaming strategy in each test scenario based on the preset number of real performance data; Determine predicted performance data based on the performance prediction model, construct a loss function based on the predicted performance data and comprehensive real performance data corresponding to the predicted performance data, optimize the performance prediction model by minimizing the loss function, and obtain an optimized performance prediction model.

3. The method for constructing a game strategy performance prediction model according to claim 1, wherein: Before building a performance prediction model for predicting the performance of a game strategy based on the capability vector, the difficulty, the discrimination, the random coefficient, and the capability vector of a test scenario corresponding to the game strategy, the method further includes: Divide the environment of the game task into test scenarios according to the set target capability dimension to obtain a target test scenario set, and construct a scenario-capability mapping matrix based on the target test scenario set and the target capability dimension; Determining a priori capability vector corresponding to each of the test scenarios based on the scenario-capability mapping matrix; Extracting features of the test scenario using multimodal data, and fusing the extracted multimodal features to obtain a multimodal vector of the scenario; Obtaining a modified capability vector based on a scenario-learnable parameter matrix and a multimodal vector of the scenario; The modified capability vector and the priori capability vector are fused to obtain the capability vector of the test scenario.

4. The method for constructing a game strategy performance prediction model according to claim 3, wherein: The environment of the gaming task is divided into test scenarios according to the set target capability dimension to obtain a target test scenario set, and a scenario-capability mapping matrix is constructed based on the target test scenario set and the target capability dimension, including: Adding disturbance factors to each target test scenario set to obtain test scenario variants; The scenario-capability mapping matrix is constructed based on the test scenario variants and the target capability dimensions.

5. The method for constructing a game strategy performance prediction model according to claim 3, wherein: The multimodal data includes image features in a test scenario and text features in the test scenario.

6. The method for constructing a game strategy performance prediction model according to any one of claims 1 to 5, characterized in that: After constructing a performance prediction model for predicting the performance of the game strategy based on the capability vector, the difficulty, the discrimination, the random coefficient, and the capability vector of the test scenario corresponding to the game strategy, the method further includes: Determine all the game strategies corresponding to the current game task, determine the comprehensive ability score based on the ability dimension corresponding to each game strategy, and use the game strategy with the highest comprehensive ability score as the anchor game strategy; Determine the comprehensive real performance data of the anchor game strategy under the new test environment, and fine-tune the performance prediction model based on the comprehensive real performance data corresponding to the anchor game strategy to obtain a fine-tuned performance prediction model, so that the fine-tuned performance prediction model can achieve cross-scenario performance prediction.

7. A method for predicting the performance of a game strategy, characterized in that: include: Determine a target test scenario corresponding to the game strategy to be tested, and encode the game strategy to be tested and the target test scenario based on the feature code to obtain a code of the game strategy to be tested and a code of the target test scenario; The game strategy code to be tested and the target test scenario code are used as inputs of a performance prediction model to obtain the target game strategy performance of the game strategy to be tested under the target test scenario; wherein, the performance prediction model is a model obtained based on the game strategy performance prediction model construction method described in any one of claims 1 to 6.

8. A device for constructing a game strategy performance prediction model, characterized in that: include: The encoding module is used to perform feature encoding processing on the game strategy and the test scenario to obtain the game strategy code and the test scenario code; A capability vector determination module, configured to determine capability vectors corresponding to different capability dimensions of a game strategy based on the game strategy code and the strategy capability learnable parameter matrix; a parameter determination module for determining difficulty, discrimination, and a random coefficient based on the test scenario code and a difficulty learnable parameter matrix, a discrimination learnable parameter matrix, and a random coefficient learnable parameter matrix; the difficulty being a parameter that measures the strength of the test scenario's requirements for the game strategy; the discrimination being a parameter that reflects the effectiveness of the test scenario in distinguishing different game strategies; and the random coefficient representing the probability of success in the test scenario without the guidance of a game strategy; A performance prediction model construction module is used to construct a performance prediction model for predicting the performance of a game strategy based on the ability vector corresponding to the game strategy, the difficulty, the discrimination, the random coefficient and the ability vector of the test scenario; the ability vector corresponding to the test scenario is determined by the ability dimension corresponding to the test scenario.

9. An electronic device, characterized in that: include: memory for storing computer programs; A processor, configured to execute the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A medium, characterized in that The medium stores a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

Images