A method and system for adjusting motion data in a virtual battlefield environment

Abstract

The application provides a kind of adjustment method and system of motion data in virtual battlefield environment, the method comprises: obtaining the current motion data of combat robot in virtual battlefield environment;With the trained residual network, the matching degree score between the current motion data and the standard motion data corresponding to virtual battlefield environment is determined;In the case where the matching degree score is less than the preset threshold, the motion deviation vector between the current motion data and the standard motion data is determined by using the difference measurement formula with the introduction of correction factor;Wherein, the correction factor includes: the environment class correction factor of virtual battlefield environment and the equipment constraint class correction factor of combat robot;Motion deviation vector is input into multi-agent near end strategy optimizer, and optimization correction instruction that satisfies combat effectiveness maximization is generated;According to optimization correction instruction, the current motion data is adjusted.The application can improve the motion accuracy of combat robot in virtual battlefield environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of virtual reality, in particular to a method and system for adjusting motion data in a virtual battlefield environment. BACKGROUND

[0002] In a virtual battlefield environment, the motion data of participating combat robots, such as tactical actions and movement paths, need to be evaluated in real time in order to provide optimization instructions to improve the action accuracy of combat robots. However, the algorithms involved in real-time evaluation and instruction optimization in the virtual battlefield training environment in the prior art usually do not have enough adaptability and robustness, resulting in the phenomenon that the adjusted motion data of combat robots cannot meet the requirements of virtual combat. SUMMARY

[0003] To solve the problems in the prior art, the embodiments of the present application provide a method and system for adjusting motion data in a virtual battlefield environment, aiming to improve the motion accuracy of combat robots in a virtual battlefield environment.

[0004] The purpose of the present application is achieved by using the following technical solutions:

[0005] In one aspect, the embodiments of the present application provide a method for adjusting motion data in a virtual battlefield environment, the method comprising:

[0006] obtaining current motion data of a combat robot in a virtual battlefield environment;

[0007] using a trained residual network to determine a matching degree score between the current motion data and standard motion data corresponding to the virtual battlefield environment;

[0008] in the case where the matching degree score is less than a preset threshold, using a difference measurement formula introducing a correction factor to determine a motion deviation vector between the current motion data and the standard motion data; wherein the correction factor includes an environment type correction factor of the virtual battlefield environment and an equipment constraint type correction factor of the combat robot;

[0009] inputting the motion deviation vector into a multi-agent proximal policy optimizer to generate an optimization correction instruction that satisfies the maximization of combat effectiveness;

[0010] adjusting the current motion data according to the optimization correction instruction.

[0011] Optionally, the current motion data includes movement path data and tactical action data, and the obtaining of the current motion data of the combat robot in the virtual battlefield environment comprises:

[0012] Adopting a sensor node posture acquisition module deployed at each joint of the combat robot, joint action data of the combat robot is acquired;

[0013] Adopting a multimedia module deployed in the virtual battlefield environment, multimedia data including the combat robot is acquired;

[0014] The multimedia data is analyzed to obtain the movement path data, and the joint action data is analyzed to obtain the tactical action data.

[0015] Optionally, the trained residual network is used to determine a matching degree score between the current motion data and standard motion data corresponding to the virtual battlefield environment, comprising:

[0016] Determine the current feature data corresponding to the current motion data, and the standard feature data corresponding to the standard motion data;

[0017] Using time series difference method, the current feature data and the standard feature data are time series synchronized to obtain processed current feature data and processed standard feature data;

[0018] The processed current feature data and the processed standard feature data are input into the trained residual network for feature processing to obtain the matching degree score.

[0019] Optionally, the training process of the trained residual network comprises:

[0020] Obtain a set of sample data pairs labeled with similarity scores; wherein the set of sample data pairs includes a plurality of negative motion data pairs with similarity scores less than a similarity threshold, and a plurality of positive motion data pairs with similarity scores greater than or equal to the similarity threshold;

[0021] Based on the set of sample data pairs, an initial residual network is iteratively trained to obtain the trained residual network.

[0022] Optionally, in the case that the matching degree score is less than a preset threshold, a difference measurement formula introducing a correction factor is used to determine a motion deviation vector between the current motion data and the standard motion data, comprising:

[0023] In the case that the matching degree score is less than a preset threshold, a space-time domain joint feature algorithm is used to extract feature information from the current motion data to obtain a current feature vector, and to extract feature information from the standard motion data to obtain a standard feature vector;

[0024] The current feature vector and the standard feature vector are substituted into the difference measurement formula of the introduced correction factor to obtain the motion deviation vector.

[0025] Optionally, in the case that the matching degree score is less than the preset threshold, a space-time domain joint feature algorithm is adopted to extract feature information from the current motion data to obtain a current feature vector and to extract feature information from the standard motion data to obtain a standard feature vector, including:

[0026] In the case that the matching degree score is less than the preset threshold, space-time features and time domain features are extracted from the current motion data to obtain current space-time features and current time domain features, and space-time features and time domain features are extracted from the standard motion data to obtain standard space-time features and standard time domain features.

[0027] The current space-time features and the current time domain features are weighted and fused to obtain the current feature vector, and the standard space-time features and the standard time domain features are weighted and fused to obtain the standard feature vector.

[0028] Optionally, the environment class correction factor includes a temperature correction factor, a humidity correction factor, a terrain correction factor and an electromagnetic coupling correction factor, and the equipment constraint class correction factor includes a mechanical dynamics correction factor and a bionic motion correction factor; the difference measurement formula of the introduced correction factor satisfies the following formula:

[0029]

[0030] Wherein, AV is the motion deviation vector, k1 is the temperature correction factor, k2 is the humidity correction factor, k3 is the terrain correction factor, k4 is the electromagnetic environment coupling correction factor, k5 is the mechanical dynamics correction factor, k6 is the bionic motion correction factor, V1 is the current feature vector, and V2 is the standard feature vector.

[0031] Optionally, the optimization correction instruction includes a path optimization instruction and an action optimization instruction; the motion deviation vector is input into a multi-agent proximal policy optimizer to generate an optimization correction instruction that satisfies maximum combat effectiveness, including:

[0032] State normalization processing is performed on the motion deviation vector to obtain a to-be-input vector.

[0033] The to-be-input vector and a current global parameter of the virtual battlefield environment are input into the multi-agent proximal policy optimizer to generate the path optimization instruction and the action optimization instruction that satisfy the maximum combat effectiveness.

[0034] The adjusting the current motion data according to the optimization correction instruction comprises:

[0035] adjusting movement path data in the current motion data according to the path optimization instruction to obtain adjusted path data, and adjusting tactical action data in the current motion data according to the action optimization instruction to obtain adjusted action data.

[0036] Optionally, after the adjusting movement path data in the current motion data according to the path optimization instruction to obtain adjusted path data, and adjusting tactical action data in the current motion data according to the action optimization instruction to obtain adjusted action data, the method further comprises:

[0037] obtaining position information of the combat robot in the virtual battlefield environment;

[0038] adjusting the combativeness data of the combat robot in the virtual battlefield environment based on the position information, the adjusted path data and the adjusted action data.

[0039] In another aspect, an embodiment of the present application provides a system for adjusting motion data in a virtual battlefield environment, the system comprising:

[0040] a obtaining module configured to obtain current motion data of a combat robot in a virtual battlefield environment;

[0041] a determining module configured to determine a matching degree score between the current motion data and standard motion data corresponding to the virtual battlefield environment by using a trained residual network;

[0042] The determining module is further configured to determine a motion deviation vector between the current motion data and the standard motion data by using a difference measurement formula with a correction factor introduced, if the matching degree score is less than a preset threshold; wherein the correction factor comprises an environment type correction factor of the virtual battlefield environment and an equipment constraint type correction factor of the combat robot.

[0043] a generating module configured to input the motion deviation vector into a multi-agent proximal policy optimizer to generate an optimization correction instruction that satisfies combat effectiveness maximization;

[0044] an adjusting module configured to adjust the current motion data according to the optimization correction instruction.

[0045] Optionally, the current motion data comprises: movement path data and tactical action data, the acquisition module is specifically configured to acquire joint action data of the combat robot by using a sensor node posture acquisition module deployed at each joint node of the combat robot; acquire multimedia data comprising the combat robot by using a multimedia module deployed in the virtual battlefield environment; analyze the multimedia data to obtain the movement path data, and analyze the joint action data to obtain the tactical action data.

[0046] Optionally, the determination module is specifically configured to determine current feature data corresponding to the current motion data, and standard feature data corresponding to the standard motion data; perform time sequence synchronization on the current feature data and the standard feature data by using a time sequence difference method to obtain processed current feature data and processed standard feature data; input the processed current feature data and the processed standard feature data into the trained residual network for feature processing to obtain the matching degree score.

[0047] Optionally, the system further comprises a training module configured to acquire a sample data pair set labeled with a similarity score; wherein the sample data pair set comprises: a plurality of negative motion data pairs with a similarity score less than a similarity threshold, and a plurality of positive motion data pairs with a similarity score greater than or equal to the similarity threshold; based on the sample data pair set, iteratively train an initial residual network to obtain the trained residual network.

[0048] Optionally, the determination module is specifically configured to, in a case where the matching degree score is less than a preset threshold, extract feature information from the current motion data to obtain a current feature vector, and extract feature information from the standard motion data to obtain a standard feature vector by using a spatio-temporal domain joint feature algorithm; substitute the current feature vector and the standard feature vector into a difference measurement formula with the introduced correction factor to obtain the motion deviation vector.

[0049] Optionally, the determination module is specifically configured to, in a case where the matching degree score is less than a preset threshold, extract spatio-temporal features and time domain features from the current motion data to obtain current spatio-temporal features and current time domain features, and extract spatio-temporal features and time domain features from the standard motion data to obtain standard spatio-temporal features and standard time domain features; perform weighted fusion on the current spatio-temporal features and the current time domain features to obtain the current feature vector, and perform weighted fusion on the standard spatio-temporal features and the standard time domain features to obtain the standard feature vector.

[0050] Optionally, the environment class correction factor comprises a temperature correction factor, a humidity correction factor, a terrain correction factor and an electromagnetic coupling correction factor, and the equipment constraint class correction factor comprises a mechanical dynamics correction factor and a bionic motion correction factor; the difference measure formula of the introduced correction factor satisfies the following formula:

[0051]

[0052] wherein, AV is the motion deviation vector, k1 is the temperature correction factor, k2 is the humidity correction factor, k3 is the terrain correction factor, k4 is the electromagnetic environment coupling correction factor, k5 is the mechanical dynamics correction factor, k6 is the bionic motion correction factor, V1 is the current feature vector, and V2 is the standard feature vector.

[0053] Optionally, the optimization correction instruction comprises a path optimization instruction and an action optimization instruction; the generation module is specifically configured to perform state normalization processing on the motion deviation vector to obtain a to-be-input vector; input the to-be-input vector and the current global parameter of the virtual battlefield environment into the multi-agent proximal policy optimizer to perform calculation, and generate the path optimization instruction and the action optimization instruction that satisfy the combat effectiveness maximization; and the adjustment module is specifically configured to adjust the movement path data in the current motion data according to the path optimization instruction to obtain adjusted path data, and adjust the tactical action data in the current motion data according to the action optimization instruction to obtain adjusted action data.

[0054] Optionally, the acquisition module is further configured to acquire position information of the combat robot in the virtual battlefield environment; and the adjustment module is further configured to adjust the adversarial data of the combat robot in the virtual battlefield environment based on the position information, the adjusted path data and the adjusted action data.

[0055] In still another aspect, an embodiment of the present application further provides an electronic device, comprising at least one processor and a memory; the memory and the processor are connected through a bus; the memory is used for storing one or more programs; when the one or more programs are executed by the at least one processor, a method for adjusting motion data in a virtual battlefield environment as described in any of the above aspects is implemented.

[0056] Correspondingly, an embodiment of the present application further provides a readable storage medium, which has an execution program stored thereon, and the execution program is executed to implement the method for adjusting motion data in a virtual battlefield environment as described in any of the above aspects.

[0057] Compared with the prior art, the present application has the following beneficial effects:

[0058] The method and system for adjusting motion data in a virtual battlefield environment provided by the embodiments of the present application, in the execution process of the method, first, the current motion data of the combat robot in the virtual battlefield environment is obtained; then, the matching degree score between the current motion data and the standard motion data corresponding to the virtual battlefield environment is determined by using the trained residual network; in the case that the matching degree score is less than a preset threshold, the motion deviation vector between the current motion data and the standard motion data is determined by using the difference measurement formula with the introduction of the correction factor; wherein the correction factor includes: the environment type correction factor of the virtual battlefield environment and the equipment constraint type correction factor of the combat robot; finally, the motion deviation vector is input into the multi-agent proximal policy optimizer to generate an optimized correction instruction that satisfies the maximization of combat effectiveness, and the current motion data is adjusted according to the optimized correction instruction. In this way, the matching degree score is determined according to the residual network with residual connection and jump structure, and in the case that the matching degree score is less than the preset threshold, the motion deviation vector between the current motion data and the standard motion data is determined by using the difference measurement formula with the introduction of the environment type correction factor and the equipment constraint type correction factor, so as to improve the accuracy and efficiency of the obtained matching degree score and motion deviation vector, and on this basis, with the help of the multi-agent proximal policy optimizer, the optimized correction instruction that can satisfy the maximum combat effectiveness is generated to adjust the current running data, which can further improve the motion accuracy of the combat robot in the virtual battlefield environment.

[0059] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, but not limiting the technical solutions provided by the embodiments of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0061] Figure 1 A flowchart of a method for adjusting motion data in a virtual battlefield environment provided by the embodiments of the present application;

[0062] Figure 2 A composition schematic diagram of a system for adjusting motion data in a virtual battlefield environment provided by the embodiments of the present application;

[0063] Figure 3 A composition schematic diagram of an electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0064] Other advantages and embodiments of the application will be more readily appreciated from the following description, taken in conjunction with the accompanying drawings, and preferred embodiments of the application. The description and drawings are not intended to limit the scope of the application, but merely to describe some of the embodiments of the application. The description and drawings should be considered in conjunction with the claims.

[0065] In the following description, reference is made to the terms "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" can be the same subset or different subsets as each other and can be combined with each other as long as there is no conflict.

[0066] In the following description, the terms "first", "second", "third" are only used to distinguish similar objects, and do not represent a specific order of the objects. It is understood that "first", "second", "third" can be interchanged in a specific order or sequence as long as it is allowed, so that the embodiments of the application described herein can be implemented in an order other than that illustrated or described herein.

[0067] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiments of the application belong. The terms used herein are only for the purpose of describing the embodiments of the application and are not intended to limit the embodiments of the application.

[0068] Embodiment 1:

[0069] Referring to Figure 1 Fig. 1 is a flow diagram of a method for adjusting motion data in a virtual battlefield environment according to an embodiment of the application, and Figure 1 The following description is made in conjunction with

[0070] Step 101: Obtain current motion data of the combat robot in the virtual battlefield environment.

[0071] In some embodiments of the application, the combat robot is an automated intelligent equipment applied in the military field, which can further be a military robot, i.e., an automatic machine with certain human-like functions used in the military field.

[0072] It should be noted that the combat robot provided by the embodiments of the application can perform reconnaissance, attack, logistics and other tasks through autonomous programs.

[0073] In some embodiments of the present application, the virtual battlefield environment can be used to refer to a highly simulated battlefield space constructed using virtual reality (VR), augmented reality (AR), and digital twin technologies, which simulates a real combat environment through multi-dimensional perception interaction, and serves military training, combat simulation, and decision support.

[0074] In some embodiments of the present application, the current motion data of the combat robot in the virtual battlefield environment includes, but is not limited to, current movement path data and current tactical action data.

[0075] In some embodiments of the present application, when the current motion data includes movement path data and tactical action data, the above step 101 can be implemented by the following steps 1011 to 1013: Figure 1

[0076] Step 1011, using a sensor node posture acquisition module deployed at each joint node of the combat robot, to obtain joint action data of the combat robot.

[0077] Step 1012, using a multimedia module deployed in the virtual battlefield environment, to obtain multimedia data including the combat robot.

[0078] Step 1013, analyzing the multimedia data to obtain the movement path data, and analyzing the joint action data to obtain the tactical action data.

[0079] In some embodiments of the present application, a sensor node posture acquisition module is deployed at each joint node (including but not limited to: revolute joint, prismatic joint, fixed joint, composite joint, and spherical joint, etc.) of the combat robot to obtain joint action data of the combat robot; at the same time, a multimedia module (including: image / video acquisition device, audio acquisition device, etc.) is deployed in the virtual battlefield environment to obtain multimedia data including the combat robot. Here, the acquisition frequency of the joint action data and the multimedia data of the combat robot can be real-time, or the corresponding acquisition frequency can be determined according to actual needs.

[0080] In some embodiments of the present application, the data obtained by analyzing the multimedia data can be used as the movement path data of the combat robot, and the data obtained by analyzing the joint action data can be used as the tactical action data of the combat robot; or, the multimedia data and the joint action data can be fused together for analysis, and the movement path data and the tactical action data of the combat robot can be obtained synchronously.

[0081] ​Step 102, using the trained residual network, determining a matching degree score between the current motion data and standard motion data corresponding to the virtual battlefield environment.

[0082] In some embodiments of the present application, the trained residual network is a convolutional neural network which has been trained and whose corresponding output loss meets a preset condition.

[0083] In some embodiments of the present application, the trained residual network can be obtained by training an initial residual network through specific training samples. The initial residual network is characterized by easy optimization and the ability to improve accuracy by increasing depth. In addition, the residual blocks inside the residual network use skip connections to alleviate the gradient vanishing problem caused by increasing the depth of the deep neural network.

[0084] In some embodiments of the present application, the training process of the trained residual network can be achieved by the following process:

[0085] First, a sample data pair set labeled with similarity scores is obtained.

[0086] The sample data pair set includes a plurality of negative motion data pairs whose similarity scores are less than a similarity threshold, and a plurality of positive motion data pairs whose similarity scores are greater than or equal to the similarity threshold.

[0087] In some embodiments of the present application, the plurality of negative motion data pairs and the plurality of positive motion data pairs can both be historical motion data of a plurality of combat robots in a virtual battlefield environment.

[0088] It should be noted that the similarity threshold can be a pre-set value, such as 0.7, etc. The present application does not make any limitation on this.

[0089] In some embodiments of the present application, the number of the plurality of negative motion data pairs and the plurality of positive motion data pairs included in the sample data pair set can be determined according to actual needs.

[0090] Then, based on the sample data pair set, the initial residual network is iteratively trained to obtain the trained residual network.

[0091] In some embodiments of the present application, the sample data pair set is used as a training sample to iteratively train the initial residual network until the loss of the predicted similarity or the predicted matching degree score output by the trained residual network meets the convergence condition.

[0092] In some embodiments of the present application, the standard motion data corresponding to the virtual battlefield environment can be pre-stored motion data, i.e., standard motion data corresponding to each frame generated according to the relevant standards. Here, the standard motion data can also include standard movement path data and standard tactical action data.

[0093] In some embodiments of the present application, the above step 102 can be implemented by the following steps 1021 to 1023: Figure 1

[0094] Step 1021, determining current feature data corresponding to the current motion data and standard feature data corresponding to the standard motion data.

[0095] Step 1022, using a time series difference method to time sequence synchronize the current feature data and the standard feature data to obtain processed current feature data and processed standard feature data.

[0096] Step 1023, inputting the processed current feature data and the processed standard feature data into the trained residual network for feature processing to obtain the matching degree score.

[0097] In some embodiments of the present application, first, the current motion data and the standard motion data are respectively subjected to feature extraction, thereby obtaining corresponding current feature data and standard feature data; then, the time series difference method can be used to time sequence synchronize the current feature data and the standard feature data, thereby inputting the time sequence synchronized current feature data and standard feature data into the trained residual network for feature processing and analysis to obtain the matching degree score between the current motion data and the standard motion data.

[0098] The time series difference method can be used to adjust and compensate missing data or reconstruct non-equidistant observation sequences, etc.

[0099] In some embodiments of the present application, the matching degree score has a value range of [0, 1], and the matching degree score can be exemplified by 0.7, etc., which is not limited by the present application.

[0100] Step 103, in the case that the matching degree score is less than a preset threshold, using a difference measurement formula introducing a correction factor to determine a motion deviation vector between the current motion data and the standard motion data.

[0101] The correction factor includes an environment type correction factor of the virtual battlefield environment and an equipment constraint type correction factor of the combat robot.

[0102] ​In some embodiments of the present application, the preset threshold value can be determined according to actual needs, such as 0.8 or 0.9, and the present application does not make any limitation on this. Meanwhile, the correction factor at least includes two types, i.e. an environment type correction factor of the virtual battlefield environment and an equipment constraint type correction factor of the combat robot. The environment type correction factor of the virtual battlefield environment can further include a temperature correction factor, a humidity correction factor, a terrain correction factor and an electromagnetic coupling correction factor, and correspondingly, the equipment constraint type correction factor includes a mechanical dynamics correction factor and a bionic motion correction factor.

[0103] In some embodiments of the present application, the step 103 can be realized by the following step 1031 and step 1032: Figure 1

[0104] The step 1031, in the case that the matching degree score is less than the preset threshold value, a spatio-temporal domain joint feature algorithm is adopted to extract feature information from the current motion data to obtain a current feature vector and to extract feature information from the standard motion data to obtain a standard feature vector.

[0105] In some embodiments of the present application, the spatio-temporal domain joint feature algorithm is an algorithm combining space domain and time domain information.

[0106] Here, the current motion data and the standard motion data can be represented by video or image data, so that the spatio-temporal domain joint feature algorithm can be used to extract more abundant feature information therefrom.

[0107] In some embodiments of the present application, the step 1031 can be realized by the following process:

[0108] Firstly, in the case that the matching degree score is less than the preset threshold value, spatio-temporal features and time domain features are extracted from the current motion data to obtain current spatio-temporal features and current time domain features, and spatio-temporal features and time domain features are extracted from the standard motion data to obtain standard spatio-temporal features and standard time domain features.

[0109] Then, the current spatio-temporal features and the current time domain features are weightedly fused to obtain the current feature vector, and the standard spatio-temporal features and the standard time domain features are weightedly fused to obtain the standard feature vector.

[0110] ​In some embodiments of the present application, a spatio-temporal domain joint feature algorithm is adopted to extract features from the current motion data and the standard motion data respectively to obtain current spatio-temporal features and current time domain features, standard spatio-temporal features and standard time domain features, and the extracted feature data is weighted and fused according to a set weight ratio to obtain a current feature vector of the current motion data and a standard feature vector of the standard motion data. The set weight ratio is related to the specific application of the virtual battlefield environment (such as a team coordination scene, a battle scene, etc.).

[0111] In step 1032, the current feature vector and the standard feature vector are substituted into the difference measurement formula with the introduced correction factor to obtain the motion deviation vector.

[0112] In some embodiments of the present application, the environment type correction factor includes a temperature correction factor, a humidity correction factor, a terrain correction factor, and an electromagnetic coupling correction factor, and the equipment constraint type correction factor includes a mechanical dynamics correction factor and a bionic motion correction factor. The difference measurement formula with the introduced correction factor can be as shown in the following formula (1):

[0113]

[0114] Wherein, ΔV is the motion deviation vector, k1 is the temperature correction factor, k2 is the humidity correction factor, k3 is the terrain correction factor, k4 is the electromagnetic environment coupling correction factor, k5 is the mechanical dynamics correction factor, k6 is the bionic motion correction factor, V1 is the current feature vector, and V2 is the standard feature vector.

[0115] In step 104, the motion deviation vector is input into a multi-agent proximal policy optimizer to generate an optimized correction instruction that satisfies the maximum combat effectiveness.

[0116] In some embodiments of the present application, the network structure of the multi-agent proximal policy optimizer (MAPPO) includes a policy network (Actor network) and a value network (Critic network). In the embodiments of the present application, the reward function of the multi-agent proximal policy optimizer needs to balance the tactical target and the physical constraint, that is, the motion deviation vector is input into the multi-agent proximal policy optimizer using the target reward function balancing the tactical target and the physical constraint as the incentive function to process, so as to generate an optimized correction instruction that satisfies the maximum combat effectiveness.

[0117] Here, the related description of the multi-agent proximal policy optimizer is similar to that in the prior art, and will not be repeated here.

[0118] It should be noted that the multi-agent proximal policy optimizer is an efficient algorithm in multi-agent reinforcement learning (MARL), which is based on the extension of proximal policy optimization (PPO) and is suitable for cooperative or competitive multi-agent scenarios. The core goal is to maximize the global combat effectiveness of the multi-agent system while ensuring the stability of policy update.

[0119] Step 105, adjusting the current motion data according to the optimization correction instruction.

[0120] According to the above description, the current motion data includes mobile path data and tactical action data, and the corresponding standard motion data also includes standard mobile path data and standard tactical action data. Here, the motion deviation vector (motion deviation vector between the current motion data and the standard motion data) can also be parsed into two vectors, i.e., a path deviation vector and an action deviation vector, and then corresponding path optimization instructions corresponding to the path deviation vector and action optimization instructions corresponding to the action deviation vector can be generated. That is, the above step 104 can be realized by the following steps A1 and step A2:

[0121] Step A1, performing state normalization processing on the motion deviation vector to obtain a to-be-input vector.

[0122] In some embodiments of the present application, the motion deviation vector is normalized to ensure that the to-be-input vector of different dimensions can be obtained in the same dimension.

[0123] Here, the motion deviation vector includes but is not limited to position offset, heading angle deviation, linear velocity error, and angular velocity error, etc.

[0124] Step A2, inputting the to-be-input vector and the current global parameter of the virtual battlefield environment into the multi-agent proximal policy optimizer for calculation to generate the path optimization instruction and the action optimization instruction that satisfy the maximum combat effectiveness.

[0125] In some embodiments of the present application, the current global parameter of the virtual battlefield environment includes but is not limited to friendly position information, friendly combat data, and local threat distribution in the virtual battlefield environment, etc.

[0126] Correspondingly, the above step 105 can be realized by the following step B:

[0127] Step B, adjusting the moving path data in the current motion data according to the path optimization instruction to obtain adjusted path data, and adjusting the tactical action data in the current motion data according to the action optimization instruction to obtain adjusted action data.

[0128] In some embodiments of the present application, after the current action data is decomposed into two parts: moving path and tactical action, the two parts can be adjusted respectively, so that the accuracy of the adjusted motion data (including: adjusted path data and adjusted action data) can be improved.

[0129] In some embodiments of the present application, after step B is performed, steps C1 and C2 can be further performed.

[0130] Step C1, obtaining position information of the combat robot in the virtual battlefield environment.

[0131] Step C2, adjusting the adversarial data of the combat robot in the virtual battlefield environment based on the position information, the adjusted path data and the adjusted action data.

[0132] In some embodiments of the present application, the position information of the combat robot in the virtual battlefield environment can be obtained by real-time monitoring by sensors. Then, the adversarial data of the combat robot in the virtual battlefield environment can be further adjusted based on the position information, the adjusted path data and the adjusted action data, so as to realize the simulation of the tactical pressure of the real battlefield.

[0133] In this way, after receiving the optimization correction instruction, the combat robot will adjust the action and moving path of the combat robot in the virtual battlefield environment according to the optimization correction instruction. Correspondingly, the virtual battlefield environment will also update the dynamic changes of the virtual battlefield environment in real time according to the adjusted action and moving path of the combat robot, including but not limited to: terrain, obstacles and enemy positions, etc. At the same time, it can also be judged whether the relative position between the combat robot in the virtual battlefield environment and its corresponding adversarial data (such as: the opponent combat robot) changes. If the positional relationship between the two parties changes, the environment simulation device further adjusts the terrain and enemy action in the virtual battlefield to simulate the tactical pressure of the real battlefield.

[0134] The method for adjusting motion data in a virtual battlefield environment provided by the embodiment of the present application, in the execution process of the method, first, the current motion data of a combat robot in a virtual battlefield environment is acquired; then, a trained residual network is used to determine a matching degree score between the current motion data and the standard motion data corresponding to the virtual battlefield environment; in the case that the matching degree score is less than a preset threshold, a motion deviation vector between the current motion data and the standard motion data is determined by using a difference measurement formula with a correction factor introduced; wherein the correction factor includes an environment type correction factor of the virtual battlefield environment and an equipment constraint type correction factor of the combat robot; finally, the motion deviation vector is input into a multi-agent proximal policy optimizer to generate an optimized correction instruction meeting the maximization of combat effectiveness, and the current motion data is adjusted according to the optimized correction instruction. In this way, the matching degree score is determined according to the residual network with residual connection and jump structure, and in the case that the matching degree score is less than the preset threshold, the motion deviation vector between the current motion data and the standard motion data is determined by using the difference measurement formula with the environment type correction factor and the equipment constraint type correction factor introduced, so as to improve the accuracy and efficiency of the obtained matching degree score and motion deviation vector, and on this basis, the multi-agent proximal policy optimizer is used to generate the optimized correction instruction meeting the maximization of combat effectiveness to adjust the current running data, which can further improve the motion accuracy of the combat robot in the virtual battlefield environment.

[0135] Embodiment 2

[0136] Based on the same inventive concept, the embodiment of the present application also provides a system for adjusting motion data in a virtual battlefield environment, as shown in Figure 2 The system 200 includes:

[0137] The acquisition module 201 is configured to acquire the current motion data of the combat robot in the virtual battlefield environment.

[0138] The determination module 202 is configured to use the trained residual network to determine the matching degree score between the current motion data and the standard motion data corresponding to the virtual battlefield environment.

[0139] The determination module 202 is further configured to, in the case that the matching degree score is less than a preset threshold, determine the motion deviation vector between the current motion data and the standard motion data by using the difference measurement formula with the correction factor introduced; wherein the correction factor includes the environment type correction factor of the virtual battlefield environment and the equipment constraint type correction factor of the combat robot.

[0140] The generation module 203 is configured to input the motion deviation vector into the multi-agent proximal policy optimizer to generate the optimized correction instruction meeting the maximization of combat effectiveness.

[0141] The adjusting module 204 is configured to adjust the current motion data according to the optimization correction instruction.

[0142] Optionally, the current motion data comprises movement path data and tactical action data, and the acquisition module 201 is specifically configured to acquire joint action data of the combat robot by using a sensor node posture acquisition module arranged at each joint node of the combat robot, acquire multimedia data of the combat robot by using a multimedia module arranged in the virtual battlefield environment, analyze the multimedia data to obtain the movement path data, and analyze the joint action data to obtain the tactical action data.

[0143] Optionally, the determining module 202 is specifically configured to determine current feature data corresponding to the current motion data and standard feature data corresponding to the standard motion data, perform time sequence synchronization on the current feature data and the standard feature data by using a time sequence difference method to obtain processed current feature data and processed standard feature data, and input the processed current feature data and the processed standard feature data into the trained residual network for feature processing to obtain the matching degree score.

[0144] Optionally, the system further comprises a training module configured to acquire a sample data pair set labeled with a similarity score, wherein the sample data pair set comprises a plurality of negative motion data pairs with a similarity score less than a similarity threshold value and a plurality of positive motion data pairs with a similarity score greater than or equal to the similarity threshold value, and iteratively train an initial residual network based on the sample data pair set to obtain the trained residual network.

[0145] Optionally, the determining module 202 is specifically configured to, in a case where the matching degree score is less than a preset threshold value, extract feature information from the current motion data to obtain a current feature vector and extract feature information from the standard motion data to obtain a standard feature vector by using a space-time domain joint feature algorithm, and calculate the motion deviation vector by substituting the current feature vector and the standard feature vector into a difference measurement formula to which a correction factor is introduced.

[0146] Optionally, the determining module 202 is specifically configured to, in a case where the matching degree score is less than a preset threshold, perform space-time feature and time domain feature extraction on the current motion data to obtain current space-time features and current time domain features, and perform space-time feature and time domain feature extraction on the standard motion data to obtain standard space-time features and standard time domain features; perform weighted fusion on the current space-time features and the current time domain features to obtain the current feature vector, and perform weighted fusion on the standard space-time features and the standard time domain features to obtain the standard feature vector.

[0147] Optionally, the environment class correction factor includes a temperature correction factor, a humidity correction factor, a terrain correction factor, and an electromagnetic coupling correction factor, and the equipment constraint class correction factor includes a mechanical dynamics correction factor and a bionic motion correction factor; the difference measurement formula in which the correction factor is introduced satisfies the following formula:

[0148]

[0149] wherein, AV is the motion deviation vector, k1 is the temperature correction factor, k2 is the humidity correction factor, k3 is the terrain correction factor, k4 is the electromagnetic environment coupling correction factor, k5 is the mechanical dynamics correction factor, k6 is the bionic motion correction factor, V1 is the current feature vector, and V2 is the standard feature vector.

[0150] Optionally, the optimization correction instruction includes a path optimization instruction and an action optimization instruction; the generating module 203 is specifically configured to perform state normalization processing on the motion deviation vector to obtain an input vector; input the input vector and the current global parameter of the virtual battlefield environment into the multi-agent proximal policy optimizer for calculation to generate the path optimization instruction and the action optimization instruction that satisfy the combat effectiveness maximization; and the adjusting module is specifically configured to adjust the movement path data in the current motion data according to the path optimization instruction to obtain adjusted path data, and adjust the tactical action data in the current motion data according to the action optimization instruction to obtain adjusted action data.

[0151] Optionally, the acquisition module 201 is further configured to acquire position information of the combat robot in the virtual battlefield environment; and the adjusting module 204 is further configured to adjust the adversarial data of the combat robot in the virtual battlefield environment based on the position information, the adjusted path data, and the adjusted action data.

[0152] It should be noted that the description of the system is similar to the description of the above-mentioned method embodiments, and has similar beneficial effects as the method embodiments. For technical details not disclosed in the system embodiments of the present application, please refer to the description of the method embodiments of the present application for understanding.

[0153] Embodiment 3

[0154] As Figure 3 shown, the present application also provides an electronic device 300, which can be a computer device, a single-chip microcomputer device, a smart mobile device, etc. The electronic device 300 in the present embodiment can include a processor 310, a memory 320, a transceiver component 330, etc. The memory 320, the processor 310 and the transceiver component 330 are connected through a bus 340; the memory 320 can be used to store an execution program, and the exemplary execution program can include instructions; the processor 310 is used to execute the instructions stored in the memory. The memory 320 can also be used to store data, which can be called and / or modified when the instructions are executed.

[0155] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), ready programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in the storage medium to implement a corresponding method flow or a corresponding function, to implement the steps of the virtual battlefield environment motion data adjustment method involved in the above-mentioned embodiments.

[0156] Embodiment 4

[0157] Based on the same inventive concept, the application further provides a readable storage medium, specifically, an electronic device readable storage medium (Memory). The electronic device readable storage medium is a memory device in the electronic device, and is used for storing programs and data. It can be understood that the storage medium herein can include a built-in storage medium in the electronic device, and of course can also include an extended storage medium supported by the electronic device. The storage medium provides a storage space, and the storage space stores an operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more execution programs (including program codes). It should be noted that the storage medium herein can be a high-speed RAM memory or a non-volatile memory such as at least one disk memory. The processor loads and executes one or more instructions stored in the storage medium to realize the steps of the method for adjusting motion data in a virtual battlefield environment in the above-mentioned embodiments.

[0158] Those skilled in the art will appreciate that embodiments of the application can be supplied as a method, a system, or a computer program product. Therefore, the application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage media, etc.) containing computer-usable program code.

[0159] The application is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce an apparatus that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that implements the functions specified in the flowcharts and / or block diagrams.

[0160] These computer program instructions can also be stored in a computer-readable memory that can direct the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction apparatus, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1the function specified in the one or more blocks.

[0161] These computer program instructions can also be loaded into computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable data processing devices to generate computer-implemented processes, so that the instructions executed on the computer or other programmable data processing devices provide processes for implementing the flows Figure 1 the flows or the flows and / or blocks Figure 1 the function specified in the one or more blocks.

[0162] The above merely illustrates the embodiments of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of the claims of the present application.

Claims

1. A method for adjusting motion data in a virtual battlefield environment, characterized in that, The method includes: Acquire current movement data of combat robots in a virtual battlefield environment; Using a trained residual network, the matching score between the current motion data and the standard motion data corresponding to the virtual battlefield environment is determined. If the matching score is less than a preset threshold, a difference measurement formula with a correction factor is used to determine the motion deviation vector between the current motion data and the standard motion data; wherein, the correction factor includes: an environmental correction factor of the virtual battlefield environment and an equipment constraint correction factor of the combat robot; The motion deviation vector is input into the multi-agent proximal policy optimizer to generate optimization correction instructions that maximize combat effectiveness; The current motion data is adjusted according to the optimization and correction instructions. When the matching score is less than a preset threshold, the motion deviation vector between the current motion data and the standard motion data is determined using a difference metric formula that incorporates a correction factor, including: When the matching score is less than a preset threshold, a spatiotemporal joint feature algorithm is used to extract feature information from the current motion data to obtain the current feature vector and extract feature information from the standard motion data to obtain the standard feature vector. The motion deviation vector is obtained by substituting the current feature vector and the standard feature vector into the difference measurement formula that introduces the correction factor. The environmental correction factors include: temperature correction factor, humidity correction factor, terrain correction factor, and electromagnetic coupling correction factor; the equipment constraint correction factors include: mechanical dynamics correction factor and biomimetic motion correction factor; the difference measurement formula for the introduced correction factors satisfies the following formula: Wherein, ΔV is the motion deviation vector, k1 is the temperature correction factor, k2 is the humidity correction factor, k3 is the terrain correction factor, k4 is the electromagnetic coupling correction factor, k5 is the mechanical dynamics correction factor, k6 is the biomimetic motion correction factor, V1 is the current feature vector, and V2 is the standard feature vector.

2. The method according to claim 1, characterized in that, The current motion data includes: movement path data and tactical action data. Acquiring the current motion data of the combat robot in the virtual battlefield environment includes: The joint motion data of the combat robot are acquired by using sensor node attitude acquisition modules deployed at each joint of the combat robot. The multimedia module deployed in the virtual battlefield environment is used to acquire multimedia data including that of the combat robot; The multimedia data is analyzed to obtain the movement path data, and the joint motion data is analyzed to obtain the tactical motion data.

3. The method according to claim 1, characterized in that, The step of using a trained residual network to determine the matching score between the current motion data and the standard motion data corresponding to the virtual battlefield environment includes: Determine the current feature data corresponding to the current motion data, and the standard feature data corresponding to the standard motion data; The time series interpolation method is used to synchronize the current feature data and the standard feature data in time series, so as to obtain the processed current feature data and the processed standard feature data. The processed current feature data and the processed standard feature data are input into the trained residual network for feature processing to obtain the matching score.

4. The method according to claim 1 or 3, characterized in that, The training process of the trained residual network includes: Obtain a set of sample data pairs labeled with similarity scores; wherein the set of sample data pairs includes: multiple negative motion data pairs with similarity scores less than a similarity threshold, and multiple positive motion data pairs with similarity scores greater than or equal to the similarity threshold; Based on the sample data set, the initial residual network is iteratively trained to obtain the trained residual network.

5. The method according to claim 1, characterized in that, When the matching score is less than a preset threshold, a spatiotemporal joint feature algorithm is used to extract feature information from the current motion data to obtain a current feature vector, and to extract feature information from the standard motion data to obtain a standard feature vector, including: If the matching score is less than a preset threshold, spatiotemporal and temporal features are extracted from the current motion data to obtain current spatiotemporal and temporal features, and spatiotemporal and temporal features are extracted from the standard motion data to obtain standard spatiotemporal and temporal features. The current spatiotemporal features and the current time-domain features are weighted and fused to obtain the current feature vector, and the standard spatiotemporal features and the standard time-domain features are weighted and fused to obtain the standard feature vector.

6. The method according to claim 1, characterized in that, The optimization correction instructions include: path optimization instructions and action optimization instructions; the step of inputting the motion deviation vector into the multi-agent proximal policy optimizer to generate optimization correction instructions that maximize combat effectiveness includes: The motion deviation vector is normalized to obtain the input vector. The input vector and the current global parameters of the virtual battlefield environment are input to the multi-agent proximal policy optimizer for calculation, generating the path optimization instruction and the action optimization instruction that maximize combat effectiveness; The step of adjusting the current motion data according to the optimization correction instruction includes: According to the path optimization instruction, the movement path data in the current motion data is adjusted to obtain the adjusted path data, and according to the action optimization instruction, the tactical action data in the current motion data is adjusted to obtain the adjusted action data.

7. The method according to claim 6, characterized in that, After adjusting the movement path data in the current motion data according to the path optimization instruction to obtain adjusted path data, and adjusting the tactical movement data in the current motion data according to the action optimization instruction to obtain adjusted action data, the method further includes: Obtain the location information of the combat robot in the virtual battlefield environment; Based on the location information, the adjusted path data, and the adjusted motion data, the adversarial data of the combat robot in the virtual battlefield environment is adjusted.

8. A system for adjusting motion data in a virtual battlefield environment, characterized in that, The system includes: The acquisition module is used to acquire the current movement data of the combat robot in the virtual battlefield environment; The determination module is used to use a trained residual network to determine the matching score between the current motion data and the standard motion data corresponding to the virtual battlefield environment. The determination module is further configured to, when the matching score is less than a preset threshold, determine the motion deviation vector between the current motion data and the standard motion data using a difference measurement formula that introduces a correction factor; wherein, the correction factor includes: an environmental correction factor of the virtual battlefield environment and an equipment constraint correction factor of the combat robot; The generation module is used to input the motion deviation vector into the multi-agent proximal policy optimizer to generate optimization correction instructions that maximize combat effectiveness; The adjustment module is used to adjust the current motion data according to the optimization correction instruction; The determining module is specifically used to, when the matching score is less than a preset threshold, employ a spatiotemporal joint feature algorithm to extract feature information from the current motion data to obtain a current feature vector, and extract feature information from the standard motion data to obtain a standard feature vector; then, substitute the current feature vector and the standard feature vector into the difference measurement formula with the introduced correction factor to calculate the motion deviation vector. The environmental correction factors include: temperature correction factor, humidity correction factor, terrain correction factor, and electromagnetic coupling correction factor; the equipment constraint correction factors include: mechanical dynamics correction factor and biomimetic motion correction factor; the difference measurement formula for the introduced correction factors satisfies the following formula: Wherein, ΔV is the motion deviation vector, k1 is the temperature correction factor, k2 is the humidity correction factor, k3 is the terrain correction factor, k4 is the electromagnetic coupling correction factor, k5 is the mechanical dynamics correction factor, k6 is the biomimetic motion correction factor, V1 is the current feature vector, and V2 is the standard feature vector.

Images