A digital human gait generation method and device, electronic equipment and storage medium

Abstract

This application discloses a method, apparatus, electronic device, and storage medium for generating gait of a digital human. The method includes: acquiring three-dimensional terrain data of a virtual scene; constructing a terrain traction field of the virtual scene using the three-dimensional terrain data; wherein the terrain traction field is a data set representing the traction intensity of each voxel of the three-dimensional terrain on the feet of the virtual digital human and the suggested landing posture; predicting the gait of a target virtual digital human based on the terrain traction field of the virtual scene to generate an optimal landing point sequence; determining the current target landing point from the optimal landing point sequence; adjusting the gait data of the target virtual digital human landing at the current target landing point based on the data at the current target landing point in the terrain traction field of the virtual scene, and rigidly locking the foot when it touches the ground to obtain the current gait of the target virtual digital human.

Description

Technical Field

[0001] This application relates to the field of virtual digital human processing technology, and in particular to a digital human gait generation method and apparatus, electronic device, and storage medium. Background Technology

[0002] In human-computer interaction scenarios of virtual reality (VR), there are common scenarios where virtual digital humans (such as user avatars, virtual guides, and companion characters) walk, and even walk in complex terrain, such as in large-scale mountain climbing and hiking scenarios.

[0003] The current method for generating the gait effect of virtual digital humans during walking mainly involves driving the skeletal position animation based on the animation data pre-made by the animator to realize the movement of the virtual digital human, and then performing post-processing, that is, adjusting the foot posture according to the ground conditions, in an attempt to make the soles of the feet conform to the terrain.

[0004] However, this method involves moving based on pre-made data and post-processing, and only adjusting the foot posture. Therefore, it is easy for the foot model to become mismatched with the ground, resulting in relative slippage, i.e., the digital virtual human sliding while walking. This is especially serious on irregular terrains such as mountains, potholes, and slopes, which directly destroys the immersiveness and realism of the VR experience. Summary of the Invention

[0005] In view of the shortcomings of the prior art, this application provides a digital human gait generation method and apparatus, electronic device and storage medium to solve the problem of slippage when walking in the prior art.

[0006] To achieve the above objectives, this application provides the following technical solution:

[0007] The first aspect of this application provides a method for generating digital human gait, including:

[0008] Acquire 3D terrain data of the virtual scene;

[0009] The terrain traction field of the virtual scene is constructed using the three-dimensional terrain data of the virtual scene; wherein, the terrain traction field is a data set representing the traction strength of each voxel of the three-dimensional terrain on the feet of the virtual digital human and the suggested footing posture.

[0010] Based on the terrain traction field of the virtual scene, the gait of the target virtual digital human is predicted, and the optimal landing point sequence is generated.

[0011] The current target landing point is determined from the optimal landing point sequence;

[0012] Based on the data of the current target landing point in the terrain traction field of the virtual scene, the gait data of the target virtual digital human landing at the current target landing point is adjusted, and rigid locking is performed when the foot touches the ground to obtain the current gait of the target virtual digital human.

[0013] Optionally, in the above-described digital human gait generation method, the step of constructing the terrain traction field of the virtual scene using the three-dimensional terrain data of the virtual scene includes:

[0014] The three-dimensional terrain in the virtual scene is discretized into multiple voxels;

[0015] Based on the three-dimensional terrain data of the virtual scene, the average normal, slope and roughness of each voxel are analyzed;

[0016] The traction coefficient of each voxel is calculated using the slope, roughness, and material affinity factor of each voxel; where the material affinity factor represents the affinity of the ground material to the foot.

[0017] The suggested gait parameters corresponding to each voxel are determined using the average normal, slope, roughness, and material type of each voxel.

[0018] Using the traction coefficients and suggested gait parameters of each voxel, a terrain traction field representing the traction data of the virtual scene is constructed.

[0019] Optionally, in the above-described digital human gait generation method, determining the suggested gait parameters corresponding to each voxel using the average normal, slope, roughness, and material affinity factor of each voxel includes:

[0020] Gaussian smoothing filter is applied to the average normal of each voxel to obtain the suggested landing normal for each voxel.

[0021] Based on the slope, roughness, traction coefficient and material type of each voxel, parameter mapping is performed to obtain the suggested stride, suggested foot lift height and suggested ground contact time percentage for each voxel.

[0022] The suggested foot landing normal, suggested stride, suggested foot lift height, and suggested ground contact time percentage for each voxel are used to form the suggested gait parameters for each voxel.

[0023] Optionally, in the above-described digital human gait generation method, the step of predicting the gait of the target virtual digital human based on the terrain traction field of the virtual scene and generating an optimal foot placement sequence includes:

[0024] Analyze the forward distance of the target virtual digital human based on its current motion state;

[0025] Determine the current search center location of the target virtual digital human;

[0026] Starting from the current search center position, along the current direction of movement, collect the centers of multiple voxels within the fan-shaped area of ​​the look-ahead distance as candidate landing points.

[0027] For each candidate landing point, a comprehensive evaluation is performed based on the traction coefficient at the candidate landing point, its distance from the current search center position, and its distance from the previous predicted landing point position to obtain a comprehensive score for each candidate landing point.

[0028] Based on the comprehensive score of each candidate landing point and the cost of transferring to other candidate landing points, the optimal landing point sequence is analyzed.

[0029] Optionally, in the above-described digital human gait generation method, determining the current search center position of the target virtual digital human includes:

[0030] If the target virtual digital human is not a follower, then the current supporting foot position of the target virtual digital human is determined as the current search center position of the target virtual digital human;

[0031] If the target virtual digital person is a follower, then the current search center position of the target virtual digital person is obtained by linearly interpolating the current supporting foot position of the target virtual digital person and the current predicted landing point position of the leader using the follower strength coefficient.

[0032] Optionally, in the above-described digital human gait generation method, adjusting the gait data of the target virtual digital human at the current target landing point based on the data at the current target landing point in the terrain traction field of the virtual scene includes:

[0033] Adjust the ankle joint angle vector of the target virtual digital human as it moves to the current target landing point;

[0034] The rotation parameters of the suggested foot normal in the suggested gait parameters of the target virtual digital human's foot rotation to the current target foot position are analyzed;

[0035] Based on the traction coefficient at the current target landing point and the suggested foot lift height in the suggested gait parameters, adjust the current foot lift height of the target virtual digital human.

[0036] Optionally, the above-described digital human gait generation method further includes:

[0037] If the target virtual digital human is in a walking queue and there is a leading role, then the gait phase of the target virtual digital human is set to the value of the gait phase of the leading role before a fixed time delay.

[0038] A second aspect of this application provides a digital human gait generation device, comprising:

[0039] The data acquisition unit is used to acquire three-dimensional terrain data of the virtual scene;

[0040] The traction field construction unit is used to construct the terrain traction field of the virtual scene using the three-dimensional terrain data of the virtual scene; wherein, the terrain traction field is a data set representing the traction strength of each voxel of the three-dimensional terrain on the feet of the virtual digital human and the suggested footing posture.

[0041] The landing point prediction unit is used to predict the gait of the target virtual digital human based on the terrain traction field of the virtual scene and generate the optimal landing point sequence.

[0042] A landing point determination unit is used to determine the current target landing point from the optimal landing point sequence;

[0043] The attitude control unit is used to adjust the gait data of the target virtual digital human when landing at the current target landing point based on the data of the current target landing point in the terrain traction field of the virtual scene, and to rigidly lock the foot when it touches the ground, so as to obtain the current gait of the target virtual digital human.

[0044] Optionally, in the above-described digital human gait generation device, the traction field construction unit includes:

[0045] A discretization unit is used to discretize the three-dimensional terrain in the virtual scene into multiple voxels;

[0046] The terrain parameter analysis unit is used to analyze the average normal, slope and roughness of each voxel based on the three-dimensional terrain data of the virtual scene.

[0047] The traction coefficient calculation unit is used to calculate the traction coefficient of each voxel using the slope, roughness, and material affinity factor of each voxel; wherein, the material affinity factor represents the affinity of the ground material to the foot.

[0048] The gait parameter analysis unit is used to determine the suggested gait parameters corresponding to each voxel by utilizing the average normal, slope, roughness, and material type of each voxel.

[0049] The terrain traction field construction unit is used to construct a terrain traction field of the virtual scene representing the traction data of continuous voxels by utilizing the traction coefficients and suggested gait parameters of each voxel.

[0050] Optionally, in the above-described digital human gait generation device, the gait parameter analysis unit includes:

[0051] A smoothing unit is used to perform Gaussian smoothing filtering on the average normal of each voxel to obtain the suggested landing normal corresponding to each voxel.

[0052] The mapping unit is used to perform parameter mapping based on the slope, roughness, traction coefficient and material type of each voxel to obtain the suggested stride, suggested foot lift height and suggested ground contact time percentage for each voxel.

[0053] The combination unit is used to combine the suggested foot landing normal, suggested stride, suggested foot lift height and suggested ground contact time percentage corresponding to each voxel to form the suggested gait parameters corresponding to each voxel.

[0054] Optionally, in the above-described digital human gait generation device, the foot placement prediction unit includes:

[0055] The distance analysis unit is used to analyze the forward distance of the target virtual digital human based on the current motion state of the target virtual digital human.

[0056] A center position determination unit is used to determine the current search center position of the target virtual digital human;

[0057] The candidate data acquisition unit is used to acquire the centers of multiple voxels as candidate landing points within a fan-shaped area of ​​the look-ahead distance, starting from the current search center position and moving along the current direction of motion.

[0058] An evaluation unit is used to comprehensively evaluate each candidate landing point based on the traction coefficient at the candidate landing point, its distance from the current search center position, and its distance from the previous predicted landing point position, to obtain a comprehensive score for each candidate landing point.

[0059] The filtering unit is used to analyze the optimal landing point sequence based on the comprehensive score of each candidate landing point and the cost of transferring to other candidate landing points.

[0060] Optionally, in the above-described digital human gait generation device, the center position determination unit includes:

[0061] The first position unit is used to determine the current support foot position of the target virtual digital human as the current search center position of the target virtual digital human when the target virtual digital human is not a follower role;

[0062] The second position unit is used to obtain the current search center position of the target virtual digital person by using the following intensity coefficient to linearly differ between the current supporting foot position of the target virtual digital person and the current predicted landing point position of the leader role when the target virtual digital person is a following role.

[0063] Optionally, in the above-described digital human gait generation device, when the posture control unit executes the data at the current target landing point in the terrain traction field based on the virtual scene, and adjusts the gait data of the target virtual digital human landing at the current target landing point, it is used to:

[0064] Adjust the ankle joint angle vector of the target virtual digital human as it moves to the current target landing point;

[0065] The rotation parameters of the suggested foot normal in the suggested gait parameters of the target virtual digital human's foot rotation to the current target foot position are analyzed;

[0066] Based on the traction coefficient at the current target landing point and the suggested foot lift height in the suggested gait parameters, adjust the current foot lift height of the target virtual digital human.

[0067] Optionally, the above-described digital human gait generation device further includes:

[0068] If the target virtual digital human is in a walking queue and there is a leading role, then the gait phase of the target virtual digital human is set to the value of the gait phase of the leading role before a fixed time delay.

[0069] A third aspect of this application provides an electronic device, comprising:

[0070] Memory and processor;

[0071] The memory is used to store programs;

[0072] The processor is used to execute the program, which, when executed, is specifically used to implement the digital human gait generation method as described in any of the above.

[0073] A fourth aspect of this application provides a computer storage medium for storing a computer program, which, when executed by a processor, is used to implement the digital human gait generation method as described in any of the preceding claims.

[0074] This application provides a method for generating gait of a digital human, which acquires 3D terrain data of a virtual scene. Then, it constructs a terrain traction field using this 3D terrain data. The terrain traction field is a data set representing the traction strength of each voxel of the 3D terrain on the feet of the virtual digital human and the suggested landing posture. This transforms the terrain features of the virtual scene into a scalar field of continuous traction data, quantifying the adhesion strength to the feet to reflect the stability of landing at each voxel position, facilitating subsequent analysis of the most stable landing position. Furthermore, it analyzes relevant parameters of the landing suggestions, allowing for more suitable landing parameters and avoiding slippage. When generating gait for the target virtual digital human, gait prediction is performed based on the terrain traction field of the virtual scene, generating an optimal landing point sequence. This pre-predicts the optimal landing point sequence to pre-determine the landing position least likely to slip, rather than performing post-processing after landing. Therefore, the current target landing point is determined from the optimal landing point sequence. Finally, based on the data of the current target landing point in the terrain traction field of the virtual scene, the gait data of the target virtual digital human when landing at the current target landing point is adjusted, and rigid locking is performed when the foot touches the ground to obtain the current gait of the target virtual digital human. Thus, when landing, the gait is adaptively adjusted based on the traction intensity of the foot and the landing suggestion, so that the landing posture can be most suitable for the current landing point. Furthermore, rigid locking is performed when the foot touches the ground to avoid changes in foot posture, thereby effectively ensuring that no slippage is detected in the entire gait process from landing to lifting the foot. Attached Figure Description

[0075] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0076] Figure 1 A flowchart illustrating a digital human gait generation method provided in this application embodiment;

[0077] Figure 2 A flowchart illustrating a method for constructing a terrain traction field in a virtual scene, as provided in this application embodiment;

[0078] Figure 3 A flowchart illustrating a method for determining suggested gait parameters corresponding to a voxel, provided in an embodiment of this application;

[0079] Figure 4 A flowchart illustrating a method for generating an optimal landing point sequence as provided in an embodiment of this application;

[0080] Figure 5 A flowchart illustrating a method for real-time adjustment of foot posture and foot lift height, provided in an embodiment of this application;

[0081] Figure 6 A schematic diagram of the architecture of a digital human gait generation device provided in this application embodiment;

[0082] Figure 7 This is a schematic diagram of the architecture of an electronic device provided in an embodiment of this application. Detailed Implementation

[0083] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0084] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0085] This application provides a method for generating digital human gait, such as... Figure 1 As shown, it includes the following steps:

[0086] S101. Obtain the three-dimensional terrain data of the virtual scene.

[0087] To analyze the terrain of a virtual scene and predict footholds in advance, allowing for adaptive adjustment of the virtual android's foot model based on the terrain, and generating a gait that precisely matches the foot model to the terrain to prevent slippage, the first step is to acquire 3D terrain data of the virtual scene. Optionally, this 3D terrain data may include geometric and physical property data of the terrain, such as its size and material.

[0088] S102. Construct the terrain traction field of the virtual scene using the three-dimensional terrain data of the virtual scene.

[0089] The terrain traction field is a data set representing the traction strength of each voxel of the 3D terrain on the feet of the virtual digital human and the suggested footing posture. In other words, the terrain traction field includes the traction data of each voxel. A voxel is short for volume pixel, which is the smallest unit in 3D spatial segmentation. Optionally, the traction strength and the suggested footing method can be represented by two parameters respectively. Therefore, the traction data of a voxel can specifically include a traction coefficient representing the magnitude of the traction strength on the feet of the virtual digital human and suggested gait parameters for footing.

[0090] It should be noted that in the embodiments of this application, the concept of traction coefficient is proposed to quantitatively describe the traction strength of each voxel of the terrain that makes up the virtual scene on the feet of the virtual digital human, that is, the "adhesion strength", which can reflect the stability of landing on each voxel, that is, the ease or difficulty of slipping.

[0091] The suggested gait parameters for landing are those recommended when landing on a voxel. Specifically, these can be the gait parameters that are theoretically compatible with that voxel when landing. Therefore, using these as suggested gait parameters for landing on that voxel can more effectively prevent slippage. Specific parameters may include, but are not limited to, suggested foot lift height, stride length, and ground contact time percentage.

[0092] Therefore, in this embodiment of the application, the three-dimensional terrain of the virtual scene is divided into multiple voxels, and the terrain data of each voxel is analyzed to determine the traction data of each voxel, thereby constructing the terrain traction field of the virtual scene, and thus converting the terrain features of the virtual scene into a scalar field of continuous traction data.

[0093] Optionally, in another embodiment of this application, one specific implementation of step S102 is as follows: Figure 2 As shown, it includes:

[0094] S201. Discretize the three-dimensional terrain in the virtual scene into multiple voxels.

[0095] Specifically, the 3D terrain in the virtual scene is discretized into a regular voxel mesh. Each voxel in the voxel mesh is the smallest cubic unit in 3D space. Each voxel V(i, j, k) corresponds to a tiny cubic region in space, which stores the terrain feature data of that region.

[0096] S202. Based on the 3D terrain data of the virtual scene, analyze the average normal, slope and roughness of each voxel.

[0097] It should be noted that the traction data is determined by comprehensively considering the influence of the voxel's normal, slope, and roughness on the "adhesion" of the foot. Therefore, it is necessary to analyze the average normal, slope, and roughness of each voxel before the 3D terrain data of the virtual scene.

[0098] The normal reflects the orientation of the voxel region. To avoid gait tremors caused by abrupt changes in the normal of a single tiny triangular facet, the average value of the normals of all triangular faces within the voxel is used to represent the orientation of the voxel region.

[0099] Optionally, since a simple arithmetic mean ignores the influence of area size, in this embodiment of the application, the normals of each triangular face within the voxel are averaged according to the area weight to calculate the average normal vector. Thus, by using area weighting, large-area triangular faces (the dominant surface) can be given higher weights, more realistically reflecting the dominant topographic orientation of the area.

[0100] Therefore, the average normal of the voxel in this embodiment is calculated as follows:

[0101] .

[0102] Where N: the average normal of the voxel, which is a three-dimensional unit vector representing the average orientation of the surface in the voxel region. f: the triangular facets contained within the currently calculated voxel V. : The area of ​​the triangular facet f. : The unit normal vector of the triangular facet f, which is also a three-dimensional unit vector. V: The voxel being calculated.

[0103] The angle between the normal and the vertical direction is the slope angle of the ground surface relative to the horizontal plane. To more intuitively represent the slope and facilitate subsequent practical applications, the inverse cosine function is used to directly convert the vector relationship into an intuitive angle value to represent the voxel slope. Therefore, the specific method for calculating the voxel slope is as follows:

[0104] .

[0105] Where N is the average normal of the voxel. Z is the unit vector perpendicular to the coordinate system, i.e., Z = (0, 0, 1).

[0106] The roughness of a voxel reflects the degree of unevenness of its surface. Too little roughness (like a smooth ice surface) makes it easy to slip, while too much roughness (like a rocky surface) makes it unstable. Therefore, through normalization, the roughness and traction coefficient show a relationship of first increasing and then decreasing. Thus, the ratio of the standard deviation of the height of all vertices within the voxel to the voxel's side length is taken as the voxel's roughness. Therefore, the method for calculating the roughness of a voxel is as follows:

[0107] .

[0108] in, is the standard deviation of the height of all vertices within the voxel. Let be the voxel side length.

[0109] S203. Calculate the traction coefficient of each voxel using the slope, roughness, and material affinity factor of each voxel.

[0110] The material affinity factor is a dimensionless parameter that represents the affinity of the ground material to the feet. Therefore, it reflects the "comfort" of different ground materials in human perception, that is, the degree of material affinity to the feet. A larger material affinity factor results in greater stability when the foot lands on the voxel. Therefore, the material affinity factor is also considered in the embodiments of this application.

[0111] Specifically, the calculation method for the voxel traction coefficient is as follows:

[0112] .

[0113] in, The static friction coefficient of the surface material can be obtained from a pre-configured physical material library. is the roughness adjustment coefficient, which can be specifically calibrated experimentally. M (material) is the material affinity factor. Therefore, it can be seen that in the embodiments of this application, the calculation model of the voxel's traction coefficient is not a purely physical friction law, but a result of a heuristic comprehensive evaluation that integrates geometry, physics, biomechanics, and psychophysics, thus more accurately reflecting the "adhesion strength" of the voxel to the foot.

[0114] S204. Using the average normal, slope, roughness, and material type of each voxel, determine the suggested gait parameters corresponding to the voxel.

[0115] Given the mean normal, slope, roughness, and material type of the voxel, the most stable footing can be determined, thus establishing recommended gait parameters corresponding to the mean normal, slope, roughness, and material type of the voxel.

[0116] Optionally, in another embodiment of this application, one specific implementation of step S204 is as follows: Figure 3 As shown, it includes:

[0117] S301. Perform Gaussian smoothing filtering on the average normal of each voxel to obtain the suggested landing normal for each voxel.

[0118] To avoid drastic foot angle fluctuations caused by abrupt changes in the normals of adjacent voxels, the average normal of each voxel is not directly used as the suggested foot landing normal. Instead, the average normal of each voxel is Gaussian smoothed and filtered before being used as the suggested foot landing normal. Therefore, the suggested foot landing normal is calculated as follows:

[0119] .

[0120] in, This is the suggested landing normal at position p, which can specifically be the suggested landing normal of the voxel at position p. This is the average normal of the voxel at position p, specifically the average normal of the voxel at position p. This is a Gaussian smoothing filter function.

[0121] S302. Based on the slope, roughness, traction coefficient and material type of each voxel, perform parameter mapping to obtain the suggested stride, suggested foot lift height and suggested ground contact time percentage for each voxel.

[0122] Specifically, suggested gait parameters can be mapped based on the voxel parameters that influence the suggested gait parameters. Optionally, suggested stride length and suggested ground contact time percentage can be mapped based on the voxel's slope, material type, and traction coefficient. Suggested foot lift height can be mapped based on the voxel's slope, roughness, traction coefficient, and material type.

[0123] Therefore, the recommended mapping method for stride length, recommended foot lift height, and recommended ground contact time percentage is as follows:

[0124] ;

[0125] ;

[0126] .

[0127] in, Suggested stride length; Recommended leg lift height; Recommended percentage of time to ground.

[0128] S303. Combine the suggested foot landing normal, suggested stride, suggested foot lift height, and suggested ground contact time percentage for each voxel to form the suggested gait parameters for each voxel.

[0129] S205. Using the traction coefficients and suggested gait parameters of each voxel, construct a terrain traction field for a virtual scene that represents the traction data of continuous voxels.

[0130] It should be noted that since only the traction data of individual voxels has been obtained, in order to obtain traction data of voxels at any location, and not just discrete voxel traction data, a terrain traction field representing the traction data of continuous voxels is constructed using the traction coefficients and suggested gait parameters of each voxel. Optionally, the entire terrain traction field can be regarded as a continuous mapping function from three-dimensional spatial location to multi-dimensional parameters, and data of any continuous point can be queried through the three-line linear difference. Therefore, the mapping function of the terrain traction field is:

[0131] .

[0132] in, Let p be the terrain traction field mapping function; p = (x, y, z), which is any continuous position in three-dimensional space.

[0133] Optionally, the specific query method can be as follows: for any position p, locate the voxel to which it belongs, and perform trilinear interpolation on the traction data of the surrounding 8 voxels to obtain the traction data of the voxel to which the position belongs.

[0134] Optionally, the terrain traction field of the virtual scene can be shared globally, not limited to the user's avatar. All virtual avatars query the data of a uniformly pre-constructed terrain traction field. Specifically, it can be represented as:

[0135] .

[0136] in, (p) represents the globally shared terrain traction field; This refers to the terrain traction field constructed through S102.

[0137] S103. Based on the terrain traction field of the virtual scene, the gait of the target virtual digital human is predicted, and the optimal landing point sequence is generated.

[0138] The optimal landing point sequence must include at least one landing point. The target virtual digital person is a specified virtual digital person. Specifically, it can be any virtual digital person in the virtual scene.

[0139] It should be noted that the traction coefficients in the terrain traction field of the virtual scene quantify the "adhesion strength" of each voxel to the foot, reflecting the stability when landing at various positions. They also include suggested gait parameters for landing. Therefore, based on the terrain traction field of the virtual scene, the optimal landing point is searched within the virtual scene, i.e., the optimal landing point within the predicted range of possible landings is obtained, thus yielding the landing point least likely to slip. For example, the sequence of landing points with the largest predicted sum of traction coefficients is taken as the optimal landing point sequence.

[0140] Since the chosen landing point can affect the next landing point, such as due to stride constraints and direction change constraints, gait planning usually cannot plan for one step. Therefore, it usually directly predicts the sequence of optimal landing points for the next multiple steps.

[0141] Optionally, in another embodiment of this application, one specific implementation of step S103 is as follows: Figure 4 As shown, it includes:

[0142] S401. Analyze the forward distance of the target virtual digital human based on the current motion state of the target virtual digital human.

[0143] Among them, the forward range is the distance within which the road ahead can be observed and predicted.

[0144] In order to determine the forward walking range and predict the optimal landing point, the forward distance of the target virtual digital human is first analyzed based on the current movement state of the target virtual digital human, so as to obtain the search range of the current search center.

[0145] Since the planning focuses on the location of landing points across multiple steps, rather than the timing of those landings, the look-ahead distance is independent of speed. Ultimately, the look-ahead distance can be determined based on the target's virtual numbers and the number of steps to be predicted, with a safety margin further considered. Therefore, the look-ahead distance can be calculated as follows:

[0146] .

[0147] in, The current movement speed of the target virtual digital human; The number of steps to predict can be set as needed, such as the default value of 3. The target is a virtual digital human with the natural stride on flat ground; This is for safety margin.

[0148] S402. Determine the current search center location of the target virtual digital human.

[0149] After determining the look-ahead distance, the center position for the search along that look-ahead distance can be determined. Optionally, the current supporting foot position of the target virtual digital human is typically used as the current search center position.

[0150] Optionally, in another embodiment of this application, one specific implementation of step S402 includes:

[0151] If the target virtual digital human is not a follower, then the current supporting foot position of the target virtual digital human will be determined as the current search center position of the target virtual digital human.

[0152] If the target virtual digital person is a follower, the current search center position of the target virtual digital person is obtained by linearly interpolating the current supporting foot position of the target virtual digital person and the current predicted landing point position of the leader using the follower strength coefficient.

[0153] It should be noted that in some virtual scenarios, multiple virtual digital humans may follow a shape. In this case, the follower can step on places already stepped on by the leader (user or guide), because these places have been verified as safe to stand on and can be followed well. For example, in a hiking scenario. Therefore, in this embodiment, when the target virtual digital human is a follower, the position of its own foothold and the position of the leader's foothold are taken into account.

[0154] In order to smoothly transition from the current position to the determined search center position, in this embodiment of the application, the current search center position of the target virtual digital human is obtained by linearly interpolating the current supporting foot position of the target virtual digital human and the current predicted landing point position of the leader role using a following intensity coefficient. Specifically, the calculation method is as follows:

[0155] .

[0156] in, To follow the current supporting foot position of the character itself, that is, the current supporting foot position of the target virtual digital human; The predicted foot placement for the leader's current gait; To follow the intensity coefficient, the value range is [0,1], for example, it can be set to 0.7.

[0157] S403. Starting from the current search center position, collect the centers of multiple voxels as candidate landing points within a fan-shaped area along the current direction of movement.

[0158] Since the character is moving forward, starting from the current search center position, multiple voxels are collected as candidate landing points within a fan-shaped area along the current direction of movement. Specifically, all voxels in the fan-shaped area can be collected as candidate landing points.

[0159] S404. For each candidate landing point, a comprehensive evaluation is performed based on the traction coefficient at that candidate landing point, its distance from the current search center position, and its distance from the previous predicted landing point position, to obtain a comprehensive score for each candidate landing point.

[0160] Among them, the traction coefficient at the candidate landing point is the traction coefficient of the voxel at the candidate landing point.

[0161] It should be noted that the traction coefficient at the candidate landing point affects the stability of the landing position, making it a primary decision-making factor. Since the distance between the candidate landing point and the current search center is too great, requiring a significant traverse to land, which is inconvenient for walking, the distance between candidate landing points is also considered. The greater the distance between candidate landing points, the lower the overall score. Furthermore, to ensure that each step's landing point is as close as possible to the previously predicted landing point, effectively avoiding gait decision jitter, this embodiment also considers the distance between the candidate landing point and the previously predicted landing point, i.e., the previously selected optimal landing point. The closer the candidate landing point is to the previously predicted landing point, the lower the overall score.

[0162] Optionally, the candidate landing point can be weighted by the traction coefficient, its distance from the current search center, and its distance from the previous predicted landing point to obtain a comprehensive score for the candidate landing point.

[0163] Optionally, in another embodiment of this application, a specific method for calculating the comprehensive score of candidate landing points may be:

[0164] .

[0165] in, The traction coefficient at the candidate landing point p; From candidate landing point p to the current search center position The horizontal distance, and Let the angle between the direction of the line segment connecting these two positions and the direction of movement be denoted as ; Maximum acceptable stride length; This is the predicted landing point location; , , , These are the weights corresponding to each item. The traction coefficient is given the highest weight and is considered as the primary factor.

[0166] It is a sequence consistency factor, which takes into account the distance to the previous predicted landing point.

[0167] S405. Based on the comprehensive score of each candidate landing point and the cost of transferring to other candidate landing points, the optimal landing point sequence is analyzed.

[0168] Since the ease or difficulty of traversing between two adjacent candidate landing points also affects the stability of the landing, this embodiment quantifies the difficulty of each candidate landing point by calculating the cost of each candidate landing point relative to other candidate landing points, thus determining the ease or difficulty of traversing between adjacent candidate landing points. The greater the difficulty of traversing between two adjacent candidate landing points, the greater the cost. Specifically, the optimal landing point sequence can be selected based on the difference between the comprehensive score of each candidate landing point and its corresponding cost.

[0169] Optionally, the cost between two candidate landing points can be calculated based on the stride length and the angle of gait direction change between the two candidate landing points. Optionally, in this embodiment, the cost between two candidate landing points can be calculated as follows:

[0170] .

[0171] in, This refers to the gait length between two steps, i.e., the stride length between two candidate foot placements; The recommended stride length for candidate landing points is specifically obtained from the suggested gait parameters of the voxels at that location. The angle of directional change for the two gaits; and These are the weighting coefficients, and their sum is 1.

[0172] Optionally, in another embodiment of this application, the optimal landing point sequence is obtained by setting an optimal gait sequence objective function. The optimal gait sequence objective function is specifically:

[0173] .

[0174] in, This is the Nth optimal landing location. and These are the candidate landing points for the kth and k+1th points.

[0175] S104. Determine the current target landing point from the optimal landing point sequence.

[0176] Specifically, based on the current location of the target virtual digital human, the current landing point is determined from the optimal landing point sequence, that is, the current target landing point is determined.

[0177] S105. Based on the data of the current target landing point in the terrain traction field of the virtual scene, adjust the gait data of the target virtual digital man landing at the current target landing point, and rigidly lock the foot when it touches the ground to obtain the current gait of the target virtual digital man.

[0178] Since the terrain traction field of the virtual scene includes data on the traction intensity of the feet and the most recommended landing posture, in order to ensure that the gait at the landing point adapts well to the current actual situation, that is, to adapt to the current target landing point and effectively prevent slippage, the gait data of the target virtual digital person when landing at the current target landing point is adjusted based on the data in the terrain traction field of the virtual scene. Specifically, this can involve adjusting the foot posture, that is, the posture of the foot bones, and modulating the lifting height of the target virtual digital person's foot, that is, adaptively adjusting the recommended lifting height.

[0179] Finally, since the gait data determined at this point is already optimal, generating a gait based on it—that is, a walking gait based on it—can effectively prevent slippage. Therefore, to avoid deviation during walking, a rigid lock needs to be applied when the foot touches the ground until it leaves the ground. This generates a gait from landing at the current target foot position until the foot leaves the ground, ensuring that the entire gait will not slip. Therefore, ground contact direction locking is the last physical line of defense against slippage.

[0180] Optionally, rigid ground contact can be achieved through ground contact position constraints. Specifically, the ground contact position constraints are as follows:

[0181] .

[0182] in, The coordinates of the ankle at a certain moment. The moment the foot touches the ground; The moment the feet leave the ground.

[0183] Optionally, in another embodiment of this application, step S105 involves a method for real-time adjustment of foot posture and foot lift height, such as... Figure 5 As shown, it includes:

[0184] S501, Adjust the ankle joint angle vector of the target virtual digital human as it moves to the current target landing point.

[0185] Having determined the current target location Subsequently, this application uses inverse kinematics (IK) to move the ankle bones to the current target landing point while maintaining the length constraints and joint angle constraints of the foot skeletal chain. Specifically, the joint angle vector of the ankle in the current frame is adjusted according to the current target landing point to obtain the latest joint angle vector, that is, the ankle joint angle vector moved to the current target landing point.

[0186] Optionally, the method of adjusting the ankle bones through inverse kinematics iteration can be specifically as follows:

[0187] .

[0188] Where q is the joint angle vector of the ankle in the current frame; This represents the position of the current frame's endpoint. It is the pseudo-inverse of the Jacobian matrix.

[0189] S502. Analyze the rotation parameters of the suggested foot normal in the suggested gait parameters of the target virtual digital human's foot rotation to the current target foot position.

[0190] After determining the ankle position, it is necessary to rotate the angle of the foot bones to ensure that the lower plane conforms to the terrain surface. In this embodiment, the foot bones are rotated according to the suggested landing normal in the suggested gait parameters. The rotation parameters of the suggested landing normal in the suggested gait parameters to the current target landing point are analyzed, and the foot bones of the target virtual digital human are rotated according to these rotation parameters. The rotation parameters can be a rotation quaternion, allowing the foot bones of the target virtual digital human to be rotated according to this rotation quaternion.

[0191] The specific method for solving the rotational quaternion of the foot bones can be as follows:

[0192] .

[0193] in, The unit vector of the foot bone's current local coordinate system Y-axis (upward direction); The suggested landing normal for the current target landing point.

[0194] S503. Based on the traction coefficient at the current target landing point and the suggested foot lift height in the suggested gait parameters, adjust the current foot lift height of the target virtual digital human.

[0195] It should be noted that in traditional technologies, the height of the swing leg is a fixed value. However, this application innovatively uses a traction coefficient as the modulation input for the leg lift height. A lower traction coefficient at the landing point indicates a more slippery, steeper, and more uncomfortable landing point. Therefore, the character lifts their leg higher to simulate the cautious gait of humans on dangerous terrain—a "high leg lift, light step"—effectively preventing slippage. The suggested leg lift height in the suggested gait parameters is the theoretically optimal landing height at the current target landing point. Therefore, by comprehensively considering the traction coefficient at the current target landing point and the suggested leg lift height in the suggested gait parameters, the leg lift height of the target virtual digital human is adjusted to obtain the current leg lift height, allowing the character to walk to the current target landing point according to the current leg lift height.

[0196] Optionally, the specific implementation method may be: adjusting the current foot height difference using the traction coefficient at the current target landing point to obtain the height adjustment amount, and using the height adjustment amount to modulate the reference foot height to obtain the current foot height of the target virtual digital person.

[0197] The current foot lift height difference is the difference between the baseline foot lift height and the suggested foot lift height in the suggested gait parameters at the current target foot landing point.

[0198] Therefore, the method for modulating the height of the raised foot, that is, the method for calculating the current height of the raised foot of the target virtual digital human, is as follows:

[0199] .

[0200] in, The baseline lift height for the original animation data; The suggested foot-lifting height; T is the traction coefficient at the current target foot landing point.

[0201] Optionally, in another embodiment of this application, it may further include:

[0202] If the target virtual digital human is in a walking queue and there is a front-row role, then the gait phase of the target virtual digital human is set to the value of the gait phase of the front-row role before a fixed time delay.

[0203] It should be noted that when multiple characters are closely following each other on a narrow path, i.e., walking in a walking queue, if the gait phases of each virtual digital human are completely random, unnatural phenomena such as overlapping steps and tripping may occur. Therefore, in this embodiment, phase synchronization is used to maintain a fixed delay between the gait of the character behind and the character in front. The specific gait phase tracking control is expressed as follows:

[0204] .

[0205] in, The gait phase of the virtual digital human character at time t, with a value range of [0,1), where 0 indicates that the right foot is touching the ground. Let be the gait phase of the leading virtual digital human character at time (t-τ). It is a fixed phase delay.

[0206] Therefore, if there is a front-row character in front of the target virtual digital human, that is, the target virtual digital human is a back-row character, then the gait phase of the target virtual digital human is set to the value of the gait phase of the front-row character before a fixed time delay.

[0207] This application provides a method for generating gait of a digital human, which acquires 3D terrain data of a virtual scene. Then, it constructs a terrain traction field using this 3D terrain data. The terrain traction field is a data set representing the traction strength of each voxel of the 3D terrain on the feet of the virtual digital human and the suggested landing posture. This transforms the terrain features of the virtual scene into a scalar field of continuous traction data, quantifying the adhesion strength to the feet to reflect the stability of landing at each voxel position, facilitating subsequent analysis of the most stable landing position. Furthermore, it analyzes relevant parameters of the landing suggestions, allowing for more suitable landing parameters and avoiding slippage. When generating gait for the target virtual digital human, gait prediction is performed based on the terrain traction field of the virtual scene, generating an optimal landing point sequence. This pre-predicts the optimal landing point sequence to pre-determine the landing position least likely to slip, rather than performing post-processing after landing. Therefore, the current target landing point is determined from the optimal landing point sequence. Finally, based on the data of the current target landing point in the terrain traction field of the virtual scene, the gait data of the target virtual digital man is adjusted when landing at the current target landing point, and rigid locking is performed when the foot touches the ground to obtain the current gait of the target virtual digital man. Thus, when landing, the gait is adaptively adjusted based on the traction intensity of the foot and the landing suggestion, so that the landing posture can be most suitable for the current landing point. Furthermore, rigid locking is performed when the foot touches the ground to avoid changes in foot posture, thereby effectively ensuring that no slippage is detected in the entire gait process from landing to lifting the foot.

[0208] Another embodiment of this application provides a digital human gait generation device, such as... Figure 6 As shown, it includes:

[0209] The data acquisition unit 601 is used to acquire three-dimensional terrain data of the virtual scene.

[0210] The traction field construction unit 602 is used to construct the terrain traction field of the virtual scene using the three-dimensional terrain data of the virtual scene. The terrain traction field is a data set representing the traction strength of each voxel of the three-dimensional terrain on the feet of the virtual digital human and the suggested footing posture.

[0211] The landing point prediction unit 603 is used to predict the gait of the target virtual digital human based on the terrain traction field of the virtual scene and generate the optimal landing point sequence.

[0212] The landing point determination unit 604 is used to determine the current target landing point from the optimal landing point sequence.

[0213] The attitude control unit 605 is used to adjust the gait data of the target virtual digital human at the current target landing point based on the data of the current target landing point in the terrain traction field of the virtual scene, and to rigidly lock the foot when it touches the ground to obtain the current gait of the target virtual digital human.

[0214] Optionally, in another embodiment of the digital human gait generation device provided in this application, the traction field construction unit includes:

[0215] Discretization unit, used to discretize the 3D terrain in a virtual scene into multiple voxels.

[0216] The terrain parameter analysis unit is used to analyze the average normal, slope, and roughness of each voxel based on 3D terrain data from a virtual scene.

[0217] The traction coefficient calculation unit is used to calculate the traction coefficient of each voxel using the slope, roughness, and material affinity factor of each voxel. The material affinity factor represents the affinity of the ground material for the feet.

[0218] The gait parameter analysis unit is used to determine the suggested gait parameters for each voxel using the average normal, slope, roughness, and material type of each voxel.

[0219] The terrain traction field construction unit is used to construct a terrain traction field of a virtual scene that represents the traction data of continuous voxels by utilizing the traction coefficients and suggested gait parameters of each voxel.

[0220] Optionally, in another embodiment of the digital human gait generation device provided in this application, the gait parameter analysis unit includes:

[0221] The smoothing unit is used to perform Gaussian smoothing filtering on the average normal of each voxel to obtain the suggested landing normal for each voxel.

[0222] The mapping unit is used to map parameters based on the slope, roughness, traction coefficient and material type of each voxel to obtain the suggested stride, suggested lift height and suggested ground contact time percentage for each voxel.

[0223] The combination unit is used to combine the suggested foot landing normal, suggested stride, suggested foot lift height and suggested ground contact time percentage for each voxel to form the suggested gait parameters for each voxel.

[0224] Optionally, in another embodiment of the digital human gait generation device provided in this application, the foot placement prediction unit includes:

[0225] The distance analysis unit is used to analyze the forward distance of the target virtual digital human based on its current motion state.

[0226] The center position determination unit is used to determine the current search center position of the target virtual digital human.

[0227] The candidate data acquisition unit is used to collect the centers of multiple voxels as candidate landing points within a fan-shaped area of ​​the look-ahead distance, starting from the current search center position and moving along the current direction of motion.

[0228] The evaluation unit is used to comprehensively evaluate each candidate landing point based on its traction coefficient, its distance from the current search center, and its distance from the previous predicted landing point, and obtain a comprehensive score for each candidate landing point.

[0229] The filtering unit is used to analyze the optimal landing point sequence based on the comprehensive score of each candidate landing point and the cost of transferring to other candidate landing points.

[0230] Optionally, in another embodiment of the digital human gait generation device provided in this application, the center position determination unit includes:

[0231] The first position unit is used to determine the current support foot position of the target virtual digital human as the current search center position of the target virtual digital human when the target virtual digital human is not a follower role.

[0232] The second position unit is used to obtain the current search center position of the target virtual digital person by linearly subtracting the current supporting foot position of the target virtual digital person from the current predicted landing point position of the leader role using the following intensity coefficient when the target virtual digital person is a following role.

[0233] Optionally, in another embodiment of the digital human gait generation device provided in this application, when the posture control unit executes data at the current target landing point in the terrain traction field based on the virtual scene, and adjusts the gait data of the target virtual digital human landing at the current target landing point, it is used for:

[0234] Adjust the ankle joint angle vector of the target virtual digital human as it moves to the current target landing point.

[0235] The rotation parameters of the suggested foot normal in the suggested gait parameters of the target virtual digital human's foot rotation to the current target foot position are analyzed.

[0236] Based on the traction coefficient at the current target landing point and the suggested foot lift height in the suggested gait parameters, adjust the current foot lift height of the target virtual digital human.

[0237] Optionally, in another embodiment of the digital human gait generation device provided in this application, the device further includes:

[0238] If the target virtual digital human is in a walking queue and there is a front-row role, then the gait phase of the target virtual digital human is set to the value of the gait phase of the front-row role before a fixed time delay.

[0239] It should be noted that the specific working process of each unit provided in the above embodiments of this application can be referred to the implementation process of the corresponding steps in the above method embodiments, and will not be repeated here.

[0240] Another embodiment of this application provides an electronic device, such as... Figure 7 As shown, it includes:

[0241] Memory 701 and processor 702.

[0242] The memory 701 is used to store the program.

[0243] The processor 702 is used to execute the program stored in the memory 701. When the program is executed, it is specifically used to implement the digital human gait generation method provided in any of the above embodiments.

[0244] Another embodiment of this application provides a computer storage medium for storing a computer program, which, when executed by a processor, is used to implement the digital human gait generation method provided in any of the above embodiments.

[0245] Computer storage media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0246] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0247] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating digital human gait, characterized in that, include: Acquire 3D terrain data of the virtual scene; The terrain traction field of the virtual scene is constructed using the three-dimensional terrain data of the virtual scene; wherein, the terrain traction field is a data set representing the traction strength of each voxel of the three-dimensional terrain on the feet of the virtual digital human and the suggested footing posture. Based on the terrain traction field of the virtual scene, the gait of the target virtual digital human is predicted, and the optimal landing point sequence is generated. The current target landing point is determined from the optimal landing point sequence; Based on the data of the current target landing point in the terrain traction field of the virtual scene, the gait data of the target virtual digital human landing at the current target landing point is adjusted, and rigid locking is performed when the foot touches the ground to obtain the current gait of the target virtual digital human.

2. The method according to claim 1, characterized in that, The process of constructing the terrain traction field of the virtual scene using the three-dimensional terrain data of the virtual scene includes: The three-dimensional terrain in the virtual scene is discretized into multiple voxels; Based on the three-dimensional terrain data of the virtual scene, the average normal, slope and roughness of each voxel are analyzed; The traction coefficient of each voxel is calculated using the slope, roughness, and material affinity factor of each voxel; where the material affinity factor represents the affinity of the ground material to the foot. The suggested gait parameters corresponding to each voxel are determined using the average normal, slope, roughness, and material type of each voxel. Using the traction coefficients and suggested gait parameters of each voxel, a terrain traction field representing the traction data of the virtual scene is constructed.

3. The method according to claim 2, characterized in that, The process of determining the suggested gait parameters corresponding to each voxel using the average normal, slope, roughness, and material affinity factor of each voxel includes: Gaussian smoothing filter is applied to the average normal of each voxel to obtain the suggested landing normal for each voxel. Based on the slope, roughness, traction coefficient and material type of each voxel, parameter mapping is performed to obtain the suggested stride, suggested foot lift height and suggested ground contact time percentage for each voxel. The suggested foot landing normal, suggested stride, suggested foot lift height, and suggested ground contact time percentage for each voxel are used to form the suggested gait parameters for each voxel.

4. The method according to claim 2, characterized in that, The method of predicting the gait of the target virtual digital human based on the terrain traction field of the virtual scene and generating an optimal landing point sequence includes: Analyze the forward distance of the target virtual digital human based on its current motion state; Determine the current search center location of the target virtual digital human; Starting from the current search center position, along the current direction of movement, collect the centers of multiple voxels within the fan-shaped area of ​​the look-ahead distance as candidate landing points. For each candidate landing point, a comprehensive evaluation is performed based on the traction coefficient at the candidate landing point, its distance from the current search center position, and its distance from the previous predicted landing point position to obtain a comprehensive score for each candidate landing point. Based on the comprehensive score of each candidate landing point and the cost of transferring to other candidate landing points, the optimal landing point sequence is analyzed.

5. The method according to claim 1, characterized in that, Determining the current search center location of the target virtual digital human includes: If the target virtual digital human is not a follower, then the current supporting foot position of the target virtual digital human is determined as the current search center position of the target virtual digital human; If the target virtual digital person is a follower, then the current search center position of the target virtual digital person is obtained by linearly interpolating the current supporting foot position of the target virtual digital person and the current predicted landing point position of the leader using the follower strength coefficient.

6. The method according to claim 2, characterized in that, The data at the current target landing point in the terrain traction field based on the virtual scene, adjusting the gait data of the target virtual digital human landing at the current target landing point, includes: Adjust the ankle joint angle vector of the target virtual digital human as it moves to the current target landing point; The rotation parameters of the suggested foot normal in the suggested gait parameters of the target virtual digital human's foot rotation to the current target foot position are analyzed; Based on the traction coefficient at the current target landing point and the suggested foot lift height in the suggested gait parameters, adjust the current foot lift height of the target virtual digital human.

7. The method according to claim 1, characterized in that, Also includes: If the target virtual digital human is in a walking queue and there is a leading role, then the gait phase of the target virtual digital human is set to the value of the gait phase of the leading role before a fixed time delay.

8. A digital human gait generation device, characterized in that, include: The data acquisition unit is used to acquire 3D terrain data of the virtual scene; The traction field construction unit is used to construct the terrain traction field of the virtual scene using the three-dimensional terrain data of the virtual scene; wherein, the terrain traction field is a data set representing the traction strength of each voxel of the three-dimensional terrain on the feet of the virtual digital human and the suggested footing posture. The landing point prediction unit is used to predict the gait of the target virtual digital human based on the terrain traction field of the virtual scene and generate the optimal landing point sequence. A landing point determination unit is used to determine the current target landing point from the optimal landing point sequence; The attitude control unit is used to adjust the gait data of the target virtual digital human when landing at the current target landing point based on the data of the current target landing point in the terrain traction field of the virtual scene, and to rigidly lock the foot when it touches the ground, so as to obtain the current gait of the target virtual digital human.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store programs; The processor is used to execute the program, which, when executed, is specifically used to implement the digital human gait generation method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, Used to store a computer program, which, when executed by a processor, is used to implement the digital human gait generation method as described in any one of claims 1 to 7.

Images