JP2005071285A - Collision detection method that change detail degree according to interaction in space and virtual space formation device using its method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the collision detection accuracy and detection speed and to make effective use of calculation resources in order to detect the collision of objects in detail, for a collision detection method that changes details degree according to the interaction in space and a virtual space generation device using its method. <P>SOLUTION: A user 2 is involved interactively in a three-dimensional virtual space 1 generated by a computer. The condition of objects including collision is presented to the user 2, and the user 2 virtually realizes and recognizes the presentation. As for objects or partial space in which the user 2 pays attention to with high interest degree like partial space 3 or an object 5 and 6, collision is detected more precisely and result is presented not to have contradiction. The details degree of collision detection is dynamically changed and collision is detected based on a collision detection evaluation function for every item the user 2 takes interest in. A partial space 4 does not have high interest degree, so detailed calculation is skipped. Calculation resources are allocated especially at points necessary, and the accuracy and speed of collision detection against objects drawing attention can be improved. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a collision detection method for a collision phenomenon of an object in a virtual space and a virtual space generation apparatus using the method.

  Conventionally, attempts to interactively use a virtual three-dimensional space (virtual space) generated by a computer have been made in a wide range of fields. To use a three-dimensional space interactively (interactively) is to enter a life-size space with a sense of self (projection) and handle the three-dimensional space interactively. Interacting with a three-dimensional space means that people (participants, users) who enter the space interact with the environment in the space in real time, that is, with the environmental objects that make up the environment. Acting with some kind of interrelationship includes touching an object (environmental object) in the space, moving it to another location, or simply observing it.

  If there is a flow, approach, contact, deformation, etc. in addition to the movement of such an object, collision / interference between the objects may occur, so a technology for detecting the collision / interference in the virtual space becomes an essential technology. . As an application field of the collision / interference detection technique, there is, for example, the one shown in FIG. In technical fields such as visualization 11, artificial reality (virtual reality) 12, entertainment 13, design 14, robot engineering 15, computer graphics 16, etc., virtual objects that collide and interfere are moved or interactively performed. Object handling and movement are handled. Based on information acquired by the user through the five senses, it is possible to recognize non-objects that are not present as actual objects, and to experience, train, learn, study, perform actual work, and change moods.

  The necessity of collision / interference detection will be further described. 18A and 18B show the collision phenomenon in the virtual space. In order to ensure the autonomy (consistency) of a virtual space where human interaction occurs, a calculation (simulation) that makes the components in the space autonomously is necessary. It is necessary to consider the physical laws that govern the real world on the earth, such as the non-permeability between objects, gravity, and friction. For example, in the virtual space 20 shown in FIG. 18A, a virtual object, for example, a rugby ball 22 and a user's hand 25 exist, and each moves, so that the virtual after a certain time shown in FIG. Assume that the rugby ball 22 and the hand 25 collide in the space 30. In the real world, when the rugby ball 22 collides and hits the hand 25, they repel each other due to the elasticity of the ball. However, if the collision is not detected in the virtual space and the subsequent space display is inappropriate, the arrow 26 As shown in FIG. 4, the hand 25 is inserted into the rugby ball 22 and further recognized as slipping through the rugby ball 22, and the virtual space becomes impaired in autonomy. As described above, the collision / interference detection between objects is an indispensable technique for the object manipulation simulation based on the physical laws.

  A general collision detection process flow in a real-time virtual space system with user interaction will be described. FIG. 19 shows the processing flow. First, when the system is activated, initial settings such as data reading and initial display for environment setting are performed (S300). When the user moves in the virtual space generated by the computer or operates an object in the virtual space, an interaction between the user and the space / object occurs (S301). In the next collision detection step, it is checked whether or not each object has collided, and if it is determined that a collision has occurred, it is checked where and how the collision has occurred (S302). . Next, response calculation according to the collision detection result is performed (S303). In this response calculation, physical properties and occurrence phenomena such as reflection, elasticity, reaction force, sound, and destruction at the collision site are considered based on the physical law. The response calculation result obtained in this way is presented to the user by a predetermined presentation function (S304). The user reacts (reacts) to the presented response and performs a new movement or operation (S305). In a real-time virtual space system with interaction, such a series of processes is continuously performed in real time.

  As described above, in order to perform an object manipulation simulation in a real-time virtual space system with interaction, it is necessary to specify the location on the object where the collision has occurred in more detail in consideration of physical laws. The calculation amount for the identification increases with the calculation accuracy. Furthermore, since an inconsistency with respect to a temporal change is also required under an interactive situation, a collision detection speed (calculation speed) that can ensure the real-time property of the phenomenon (real-time property) is required. Thus, in a real-time virtual space system with interaction, both collision detection accuracy and collision detection speed are required, and these are in a trade-off relationship. In addition, the amount of calculation increases as the object and space become more complex.

  A conventional method for detecting a collision between objects using a hierarchical representation of the object shape will be described. FIG. 20A shows a three-dimensional space where an object exists, and FIG. 20B shows a state where the space is divided into eight. FIGS. 21 (a) and 21 (b) show an octree method for examining collision hierarchically. As a method for efficiently detecting a collision, a hierarchical expression method of an object based on space division using a circumscribed cuboid or a circumscribed sphere is known. In addition, an LOD (level of detail) method that hierarchically represents an object according to the position and state of the object itself is known. What is described here is a method based on space division by the former octree.

  In FIG. 20A, for example, the objects B and D collide in the three-dimensional space R. In order to specify the collision site, the work of dividing the three-dimensional space R into eight parts including the object is repeated. FIG. 20B shows an eight-division state in the first hierarchy j = 1 (the object illustration is omitted). In the second hierarchy j = 2, each of the eight partial spaces is further divided into eight as necessary. As described above, an octree method can be used as a method of detecting a specific subspace by repeating the division into eight spaces.

  An explanation will be given by applying the octree method to the collision detection of the objects F and G in a certain space R1. Both objects F and G are represented by octrees shown in FIGS. 21 (a) and 21 (b), respectively. Each partial space in each layer is expressed by being divided into a space outside the object (white circle in the figure), a space inside the object (black circle in the figure), and a space including a part of the object (square with hatching in the figure). In addition, eight subspaces in each hierarchy are identified by an argument i.

  Comparing the octrees of the objects F and G, the subspace of (j, i) = (1, 5), (2, 3), (3, 6), (4, 3) It can be seen that there is a partial coexistence state where there is no partial space where both objects coexist in addition to these partial spaces in layers j = 1 to 4. In the hierarchy j = 5, both objects coexist in (j, i) = (5, 5). Further, in (j, i) = (5, 7), both objects F and G pass through the space. Satisfies completely. Thus, the collision of the objects F and G is detected at least in the partial space represented by (j, i) = (5, 7). The amount of computation in collision detection using an octree is only proportional to the number of branch points (number of nodes) actually followed and the number of objects. Collisions can be detected with a smaller amount of computation than when checking.

In addition, as a collision detection method for accurately examining a collision between planes using the above-described octree concept, an object that is roughly interfering is detected using a rectangular parallelepiped of a three-dimensional object in the first step, In the second step, a plane intersecting the overlapping area of the containing rectangular parallelepiped is extracted, and in the third step, the possibility of collision of the extracted plane is roughly checked by space division by each plane octree, and the fourth step There is known a method for examining a collision between surfaces for each combination of surfaces on which a surface octree interferes (see, for example, Patent Document 1).
JP-A-9-81776

  However, the above-described collision detection method using space division or octree as shown in FIG. 20 or FIG. 21 is particularly efficient for a complicated object because the calculation amount does not depend on the complexity of the object. However, when an additional shape data structure other than polyhedral representation is used together, there is a problem that it becomes difficult to cope with a deformed object and the amount of calculation increases.

  Moreover, in the collision detection method as shown in Patent Document 1 described above, there is a problem that the efficiency of collision detection greatly depends on the shape of the target object. For example, when an object with many protrusions is included in a rectangular parallelepiped, there are many space parts without an object between the object-containing rectangular parallelepiped and the included object, and even if there is an overlap between two object-containing rectangular parallelepipeds, the overlapping space There are frequent occurrences of no object, and the efficiency of detecting an interfering object deteriorates.

  The present invention solves the above-mentioned problems, and effectively uses computational resources for detecting in detail a collision of an object in a space under a user's interactive activity status in a virtual three-dimensional space. Using the collision detection method and the method for changing the degree of detail according to the interaction in the space, which can improve the collision detection accuracy and collision detection speed, and support the user's interactive activities An object is to provide a virtual space generation device.

  In order to achieve the above object, the invention of claim 1 detects a collision between a moving object and another object in a virtual space, which is generated by a computer and can be interactively engaged with the user, in order to present it to the user in the virtual space. A collision detection method, an interest level acquisition step of acquiring a user's interest level information with respect to a partial space of a virtual space and / or an object in the virtual space, and an object's interest level according to the user's interest level information acquired in the step It is a collision detection method provided with the detail level determination step which changes and determines the detail level and the detail level of collision detection dynamically.

  According to a second aspect of the present invention, in the collision detection method according to the first aspect, an evaluation function generation step of generating a collision detection evaluation function as a function of spatial coordinates for hierarchically associating the interest level information and the level of detail of collision detection. And a collision detection step of detecting a collision based on the level of detail determined in the level of detail determination step, and each subspace using the collision detection evaluation function generated in the level of detail determination step Is a collision detection method in which the detail level of an object and the detail level of collision detection are dynamically changed and determined, and the detection operation with an accuracy exceeding the detail level is discontinued in the collision detection step and no further detection is performed. .

  According to a third aspect of the present invention, in the collision detection method according to the second aspect, each object in the virtual space has a plurality of expression formats having different levels of detail, and the collision detection evaluation function is used in the level of detail determination step. The object representation format in each subspace is dynamically selected based on the above, and the collision detection is performed using the selected representation format in the collision detection step.

  The invention according to claim 4 is the collision detection method according to claim 2 or claim 3, wherein in the evaluation function generation step, the user's pointing, the distance from the pointer position, or the distance from the user's viewpoint, Or the distance from the user or an important object in the space, the angle from the user's visual field center (gaze direction), the angle from the user's pointing or pointing direction, or the movement of the object across the user's work area or visual field A collision detection evaluation function value is generated by increasing or decreasing the collision detection evaluation function value according to the speed.

  A fifth aspect of the present invention is a virtual space generation device using the collision detection method according to any one of the first to fourth aspects.

  According to the first aspect of the invention, the interest level information of the user with respect to the space or the object is acquired in the interest level acquisition step, and the detail level of the object and the detail level of the collision detection according to the interest level information in the detail level determination step. Since it is determined by dynamically changing and the calculation resources for detailed collision detection can be concentrated on the necessary places, both the collision detection accuracy and the collision detection speed for the object of interest can be improved. .

  According to the second aspect of the invention, in the evaluation function generation step, the collision detection evaluation function for hierarchically associating the user interest level information and the collision detection detail level is generated as a function of spatial coordinates, and the detail level determination is performed. In step, collision detection evaluation function is used to dynamically change the object detail and collision detection detail in each subspace, detect the collision in the collision detection step, and the degree of detail not based on interest level Since the detection work with the accuracy exceeding the limit is discontinued and no further detection is performed, calculation resources for detailed collision detection can be assigned more focused to the necessary locations, and the collision detection accuracy and collision detection for the object of interest Both speeds can be improved.

  Thus, the following effects are acquired by performing collision detection by dynamically changing the level of detail of collision detection according to the degree of interest (interaction) indicated by the user. When working using a relatively large 3D space, even when multiple users share the space, the attention of the user (participant) is evenly applied to the entire space. Instead, there is a space that is the focus of work and a space that is not. In this case, since accurate simulation is required for the object in the subspace that is the focus of the work, it is necessary to specify the location on the object where the collision occurred and the orientation of the surface of the object in detail, For objects in other spaces, a highly accurate simulation is not required, and it is not necessary to detect a collision in detail.

  In addition, the partial space that is the focus of the work and the partial space that the user is highly interested in vary depending on the progress of the work and various interactions. Therefore, depending on the user's interest in each subspace resulting from various user interactions in the three-dimensional space and the importance of each space changing in relation to the work contents, By changing the level of detail of collision detection, both the collision detection accuracy and the collision detection speed can be improved. Therefore, according to such a collision detection method, it is possible to secure both the autonomy (consistency) of the virtual space and the real time property (real time property) of the phenomenon in the virtual space. There is an effect that it is possible to realize interactivity, and thus to support user's interactive activities in the virtual space.

  According to the third aspect of the present invention, each object in the virtual space has a plurality of expression formats having different degrees of detail. Therefore, the collision detection may be performed using any one of the expression formats. Since the expression format is dynamically selected based on the evaluation function, the object collision detection in the virtual space and the display of the object and object collision state to the user (display) are made faster and more realistic. It can be carried out.

  A plurality of expression formats having different degrees of detail may be created in advance offline or in real time at the time of execution. For example, a hierarchical expression of an object shape such as Octree can be used as an expression with different levels of detail for each object. In addition, a level of detail (LOD) method that is used to reduce the load of normal drawing can be used in combination.

  According to the fourth aspect of the present invention, the user's pointing position is based on the normal relationship that the higher the user's interest in the work target, the more focused the object is, and vice versa. , The position of the pointer, the user's viewpoint, the distance from the user or important object position in the space, the center of the user's visual field (gaze direction), the pointing, the direction from the pointing direction, the object crossing the work area or the visual field Since the collision detection evaluation function is generated by increasing / decreasing according to the moving speed, the level of detail of collision detection can be dynamically changed according to the degree of interest of the user.

  In other words, the closer to the object the user is aware of, the more interested the object is, and the more the user is away from the user's field of view, the more the object is interested. Consistent with the usual relationship of increased interest and increased interest in objects that move forward than objects that are stationary, among objects in front of which the user is conscious, Interest level information can be incorporated into the collision detection evaluation function. As a result, the higher the degree of interest, the more detailed collision detection can be performed, and both the collision detection accuracy and the collision detection speed can be improved as described above.

  According to the invention described in claim 5, in the virtual space generated by the computer, the level of detail of collision detection based on the collision detection evaluation function generated according to the degree of interest of the user with respect to the subspace or the object in the space. To detect collisions and assign detailed computational resources for detailed collision detection where necessary, improving both collision detection accuracy and collision detection speed in the space of interest or for the object Can be realized. As a result, the user can create self-projection and real-time in a virtual space that is generated by such a virtual space generation device and is compatible with the autonomy (consistency) of the virtual space and the real-time nature of the phenomenon in the virtual space. The virtual space can be used interactively in a state where the interactivity is realized.

  Hereinafter, a collision detection method for changing the level of detail according to an interaction in a space according to an embodiment of the present invention and a virtual space generation apparatus using the method will be described with reference to the drawings. FIG. 1 shows the concept of the relationship between the degree of interest and the level of detail of collision detection in a virtual three-dimensional space. The three-dimensional virtual space 1 is generated by an interactive real-time virtual space system using a computer, and (virtual) objects 5 to 8 in the space are recognized by the user 2. Further, the user 2 can interactively operate (work) at least one of the objects 5 to 8, or can influence any motion or presence state. Furthermore, the virtual space system generating the virtual space 1 can be understood by the input I / F unit (described later) that the user 2 is interested in any of the objects 5 to 8. Yes. In such a virtual space 1 in which the user 2 can be involved, the movement of the object and the change of the state are presented to the user 2, and the user 2 virtually recognizes and perceives the presentation as a real thing. In order to make the recognition / perception more realistic, a collision between a moving object and another object is detected in more detail, and the result is presented to the user 2.

  Further, in order to enhance the recognition / perception effect by the user 2 on the phenomenon occurring in the virtual space 1, the user is paying attention and interested in the virtual space 1, such as the partial space 3 and the specific objects 5 and 6, for example. Collisions are detected in more detail with emphasis on subspaces and objects that are high. In this way, for the item that the user 2 is interested in, a collision detection evaluation function is determined for each item in advance or hierarchically for each collision detection (described later), and based on this collision detection evaluation function, a collision is determined. Collisions are detected by dynamically changing the level of detail of detection.

  As a result, it is determined that the user's interest is not high in the subspace 4 behind or below the user 2, and detailed state change tracking is omitted for the objects 7 and 8 in the subspace 4, for example. Therefore, calculation resources for detailed collision detection can be assigned to a phenomenon in a necessary place, for example, the partial space 3, and both the collision detection accuracy and the collision detection speed for the object of interest are improved. it can. Therefore, according to such a collision detection method, it is possible to secure both the autonomy (consistency) of the virtual space 1 and the real-time property of the phenomenon in the virtual space 1 (real-time property). Real-time interactivity can be realized, and the user 2 can be more interactive in the virtual space 1.

  Next, the above-described calculation resource allocation will be described. FIG. 2 shows segmentation of a subspace of a virtual three-dimensional space determined by the degree of interest and the level of detail of collision detection. The interest level NC, the collision detection detail level ND, and the entire space VR shown in FIGS. 2A to 2D are all standardized to take a value from the minimum value 0 to the maximum value 1 for convenience of explanation. It is. In the example shown here, the degree of interest NC is hierarchized into four levels by the hierarchized numerical values c1, c2, and c3, and the collision detection detail ND is hierarchized by the hierarchized numerical values d1, d2, and d3, respectively. Similarly, the data is hierarchized into four partial spaces by the hierarchized numerical values r1, r2, and r3.

  The area a in each graph shown in FIG. 2 has the highest degree of interest and is the area determined to detect the collision in detail, and the degree of interest decreases in the order of the areas b, c, and d. The smaller the ratio r1 of the partial space of the area a occupying the entire space VR, the fewer calculation resources are input in the entire system. Further, in the actual collision detection calculation, as shown in FIG. 2D, for example, as the value of the collision detection detail level ND with respect to the degree of interest ND of c1 <ND ≦ c2, for example, in the corresponding hierarchy The upper limit value d2 is selected, and collision detection calculation is performed using an environmental object model determined by the value d2.

  Next, an outline of the process of the collision detection method for changing the level of detail according to the interaction in space will be described. FIG. 3 shows a functional relationship that links interest level information and detail level of collision detection. The user's interest level information is information that identifies a space such as the user's pointing position, the user's viewpoint, and the user's visual field center (gaze direction) (# 1). When such interest level information about the space is acquired, the collision detection evaluation function I (x, y, z) reflecting the information is determined as a function of the spatial coordinates (# 2). For example, see formulas (1) and (2) below. As will be described later, the collision detection evaluation function has a maximum value at a point or a set of points defined by spatial information as interest level information, and decays with distance from the point (a matrix function). (# 3). This means that the degree of interest is weighted for a specific space.

  Since collision detection is performed between objects, it is necessary to assign interest level information to the objects. When the user's interest level information is obtained for the space and the collision detection evaluation function I (x, y, z) is obtained, the occupied space information (# 4) of the object m and the collision detection evaluation function are used. Thus, the degree of interest O (m) for the object m is calculated. This is obtained by integrating the collision evaluation function with respect to the space occupied by the object m (# 5, # 6), as shown in equation (3) described later. Based on the degree of interest O (m) for the object m thus obtained, the detail level of the object and the detail level of collision detection related to the object are determined.

  When the user's interest is toward the object itself, not to the space, and the interest level O (m) for the object m is obtained as the interest level information, the interest level O (m) is used (# 7). . The degree of interest O (m) of the object m obtained by any one of these methods takes any value from Omin to Omax depending on the spatial position.

  The object m has its expression format determined in advance or dynamically, and hierarchical levels L (0) to L (n−1), for example, are determined hierarchically based on the expression format ( # 8). For the object m, the current level of detail L (j) for the object m is determined using the current interest level O (m) for the object m (# 9, # 10). Since the level of detail L (j) is hierarchized, its value is usually given as an integer value. For an object with a high degree of interest, an object representation having a high level of detail L (j) is selected, and therefore the collision calculation is also detailed (# 11). In addition, for a dynamic change in the level of detail of collision detection in the octree method, a value indicating the depth of the octree division level for collision detection truncation is output as the output of the collision detection evaluation function.

  Next, a processing flow in a real-time virtual space system that detects a collision by dynamically changing the level of detail (detection depth) of collision detection according to the above-described interaction will be described. FIG. 4 shows the processing flow. This process flow is an improvement of the process of the collision detection S302 in the process flow generally performed conventionally shown in FIG. In the collision detection S302 in the prior art, the interest level of the user is taken into consideration, and the interest level acquisition S3, the collision evaluation function generation and update S4, the detail level determination S5, the collision level are newly calculated so as to detect the collision by analyzing the user interaction. Detection is S6.

  In the flow shown in FIG. 4, first, initial setting is performed (S1). The initial settings include condition settings for operating the system and environment settings. When the system is operated and the user moves in the virtual space generated in the real-time virtual space system or manipulates an object in the virtual space, interaction between the user and the space / object occurs (S2). ).

  In a series of subsequent processes, the user interaction is analyzed in consideration of the user's interest level, and a collision is detected. First, in the interest level acquisition step, the user's interest level information in all spatial positions and all objects is measured and examined for each evaluation item of interest level (S3). Subsequently, in the evaluation function generation step, a collision detection evaluation function is generated as a function of spatial coordinates (S4), and in the detail level determination step, using the collision detection evaluation function, the detail level of the object and the detail level of collision detection in each partial space Is determined (S5). A collision is detected based on the level of detail determined in the collision detection step (S6).

  Thereafter, processing is performed in the same manner as the conventional processing flow shown in FIG. That is, a response calculation corresponding to the collision detection result is performed (S7). In this response calculation, physical properties and occurrence phenomena such as reflection, elasticity, reaction force, sound, and destruction at the collision site are considered based on the physical law. The response calculation result obtained in this way is presented to the user by a predetermined presentation function (S8). In response to the presented response, the user performs a new movement or operation (S9). In a real-time virtual space system with user interaction, such a series of processes is continuously performed dynamically and in real time, for example, at a frequency of 10 times per second until the system is stopped (Y in S10). Is called.

  Next, the above-described interaction analysis process (S3 to S6) will be further described. FIG. 5 shows the analysis processing flow part. Originally, collision detection is performed to present to a user using a virtual space a collision between a moving object and another object in a real-time virtual space system that is generated by a computer and can be interactively engaged by the user. Therefore, after the system is activated and the initial setting is performed, first, interest level information acquisition for the user's space or object is performed in the interest level acquisition step (S103).

  The interest level information acquired in this interest level acquisition step (S103) is basically information indicating which part of the virtual three-dimensional space the user is interested in. Information to identify. For example, the pointing position of the user, the position of the pointer, the viewpoint of the user, the distance from the user or an important object position in space, the center of the user's visual field (gaze direction), the angle from the pointing, pointing direction, etc. The moving speed of the object that crosses the region or the field of view is acquired as the interest level information. The point or direction that becomes the center of interest of the user changes from time to time with the user's activities. The pointing direction is obtained by mounting a sensor for measuring the position and azimuth angle (for example, a product called Fasttrack manufactured by Pohimas Co., Ltd.) on the finger and measuring the finger coordinate origin position and the direction of the pointing axis. If such a sensor is attached to the head, the position and direction of the head and face can be measured. The direction of the line of sight can be detected by, for example, a sensor called Visiontrack from Pohimas. Similar sensors are also available from other companies.

  In addition, as information representing the level of interest, a state quantity representing a physical state including a user's heart rate, sweating rate, respiratory rate, and the like is measured as necessary. The measurement values themselves of these state quantities themselves or the results of analyzing the fluctuations thereof are acquired as interest level information. The heart rate, the amount of sweat, the respiration rate, and the like can be measured by a general medical measurement device that is generally widely used. The so-called emotional interest level information based on this physical state quantity is complementary information that reinforces and complements the direct and spatial interest level information that is clearly shown in the above-mentioned spatial points. is there. That is, (interesting degree information) = (explicit spatial interest information + emotional interest degree information).

  Subsequently, in the evaluation function generation step (S104), a collision detection evaluation function for hierarchically associating the interest level information and the detail level of the collision detection according to the user interest level information acquired as described above is provided. , Generated as a function of spatial coordinates.

  The collision detection evaluation function is, for example, the distance from the user's pointing position, the distance from the user's viewpoint, the angle from the user's visual field center (gaze direction), the user's work area or the moving speed of the object across the visual field, etc. The function value increases or decreases accordingly.

  Subsequently, in the level of detail determination step (S105), the level of detail of the object and the level of detail of collision detection in each subspace are dynamically changed using the collision detection evaluation function, that is, in real time, corresponding to the situation change. To be determined. At this time, the expression format of the object that is the target of collision detection is determined. That is, each object in the virtual space has a plurality of expression formats with different levels of detail created in advance offline or in real time at the time of execution (described later). A representation format of each object for collision detection is dynamically selected based on the function value (evaluation function value).

  Subsequently, in a collision detection step (S106), collision detection between objects is performed using an object having an expression format selected based on the subspace and the level of detail of collision detection for each object. The representation format of each object used for collision detection is a detailed representation format for detailed (high-precision) collision detection, and a simple representation for collision detection for non-detailed (coarse) collision detection. Is used. For example, an object with less unevenness has a simpler expression format than an object with more unevenness, and has a larger number of constituent polygons, and collision detection is usually easier.

  In this way, if the object representation format is determined based on the collision detection evaluation function determined according to the degree of interest, and collision detection is performed using the object in the representation format, collision detection between the collision objects is performed. Even if the calculation of is completely performed, it is the collision detection of the degree of detail of the expression form of the object at most. Therefore, the detection operation with the accuracy exceeding the level of detail that is not based on the level of interest of the object is automatically canceled. The level of detail of collision detection may be determined by either the level of detail when the calculation determined in such an expression format is completely performed or the level of detail when a limit is further added to the number of calculation steps for collision detection in the expression format. .

  Next, the expression of the object in the virtual space described above will be described. FIGS. 6A to 6C show examples of representation formats of objects having different degrees of detail. The objects shown in FIGS. 6 (a) and 6 (b) are obtained by changing the level of detail hierarchically, that is, stepwise, not continuously, depending on the number of polygons used for object representation. The objects shown in FIG. 6 (c) Represents an object as a collection of spheres, and the level of detail changes hierarchically depending on the number of spheres used for object representation. As in these examples, all objects in the space are expressed in a plurality of expression formats with different levels of detail. These may be created in advance offline or in real time during execution.

  As expressions with different levels of detail of objects, hierarchical shape expressions such as Octree and LOD used to reduce the drawing load may be used in combination. The idea of LOD (level of detail) and LOD values that express LOD numerically is originally widely used in the field of computer graphics, and reduces the number of polygons of objects far from the viewpoint position. In other words, it is a technique for drawing a realistic picture quickly and efficiently by reducing the level of detail. 6A to 6C show the most detailed object representation when the LOD value is LOD = 0, and the detail level decreases as the LOD value increases, resulting in a rough object representation. Yes.

  Next, the change and determination of the detail level of collision detection in the above-described detail level determination step will be described. To what extent the collision is examined in detail, that is, at which stage the detailed collision detection is terminated depends on the level of detail of collision detection based on the collision detection evaluation function as described above. FIG. 7A shows the state of the user's work (interaction) in the virtual space, and FIG. 7B shows an example of the user's interest level information in the work. When collisions between objects occur in space during the user's interactive work, for example, the value of the collision detection evaluation function generated as a result of the interaction up to that point is integrated, and the corresponding object's The expression format and the level of detail of collision detection are determined, and this is used to detect (calculate) a collision. In FIG. 7B, a point P (X, Y, Z) is a point on the visual field center line (line of sight), and the visual field separation angle θ is an angle from the visual line direction. In this case, the smaller the visual field separation angle θ, the higher the degree of interest and thus the collision detection evaluation value. The collision detection evaluation function is generated for a plurality of types of interest levels for each object, and is used in a composite manner by performing statistical processing such as weighting on the plurality of results.

  The collision detection using the hierarchical object shape representation and the octree method will be described. FIG. 8 shows the relationship between the octree hierarchy and the collision detection detail level. When the user's degree of interest is high, the hierarchy j of the octree is deepened and the collision phenomenon is examined more finely. That is, more detailed collision calculation and response calculation are performed. The depth of the hierarchy for collision detection and the object representation are determined based on the above-described collision detection evaluation value.

  Next, a case where the concept of LOD is used for collision detection will be described. FIG. 9A shows an object expression used for collision detection in a virtual three-dimensional space, and FIG. 9B shows an object expression presented to the user. FIG. 9A shows that the object representation of the LOD value selected according to the interest level and the detail level of collision detection based on the collision detection evaluation value is used for collision detection. As shown in FIG. 9B, if the objects a0, b0, c0 in the expression format of LOD = 0 with a high degree of detail float in the three-dimensional space and collide, based on the motion vector and the normal vector of the collision surface. The user is presented with an image of the reverse motion.

  In the collision detection, in the partial space (inside the circle of radius r0 in the figure) where the collision detection evaluation value is high as a result of the interaction between the user and the space, the objects a0 and b0 having the highest level of detail are LOD = 0. , C0, and collision detection is performed using the objects a1, b1, c1 with LOD = 1 in the ring-shaped subspace between the circle with the radius r1 and the circle with the radius r0. Similarly, in the remaining space, collision detection is performed in association with the LOD value. In this example, the collision detection evaluation value decreases as the distance from the center of the circle increases, and collision detection is performed using a roughly shaped object. When it is desired to reduce the burden of object presentation, the object representation shown in FIG. 9A may be presented to the user instead of the object representation shown in FIG. 9B.

  Next, the collision detection evaluation function will be described more specifically. FIG. 10 shows an example of the collision detection evaluation function. The collision detection evaluation function is a function of spatial coordinates for hierarchically associating the interest level information and the detail level of the object, and thus the detail level of the collision detection, according to the interest level information of the user as described above. The interest level information is basically information indicating which part of the virtual three-dimensional space the user is interested in, and is information for specifying a point or direction in the space. The collision detection evaluation function indicates the level of interest distributed in the space.

  When the collision detection evaluation function I (x, y, z) is characterized by distinguishing the type of user's interest, the collision detection evaluation function is expressed as I (x, y, z; α). Here, x, y, and z are the coordinates of a spatial point, and α is an interest type that gives interest level information of the user that characterizes the collision detection evaluation function. As described above, the interest degree type α includes the user's pointing position, viewpoint (view center), and direction thereof. Since the collision detection evaluation function I (x, y, z, α) measures and evaluates the degree of interest, the degree of interest of the user with respect to the degree of interest type α is expressed numerically, and the degree of interest is spatially distributed. As a function. This function numerically represents how much the focused subspace is the focus of the user's work (interest level), that is, what value the interest level is.

  As the simplest collision detection evaluation function I (x, y, z: α), as shown in FIG. 10 (a), a point P1 in the space is the center and decreases as the distance Δr from the point P1 increases. The collision detection evaluation function I (x, y, z; Δr) having the function value to be increased is raised. The collision detection evaluation function I (x, y, z; Δr) is spherically symmetric around the point P1. The point P1 is specified by the measuring means described in the above-described interest level acquisition step as the pointing of the user 2 or the viewpoint position of the user 2. As the function form, for example, the following equation (1) can be used.

  Here, r0 is a position vector from the user 2 to the point P1, r is a position vector from the user 2 to the point P2, Δr is an absolute value of the difference between the two vectors, and d is I (x, y, z; Δr). A parameter for setting a finite value, k is a function value adjustment parameter. Note that these parameters may have a distance Δr dependency. The collision detection evaluation function I (x, y, z; Δr) represented by the equation (1) has a maximum (maximum interest level) at the position of the distance r0 as shown in FIG. ~ W3 etc. The function of equation (1) is squared with the denominator Δr to make it a function that rapidly decreases as Δr increases.

  Differences between the curves w1 to w3 of the above-described collision detection evaluation function I (x, y, z; Δr) will be described. The collision detection evaluation function is generally a function of time, and reflects that the degree of interest of the user 2 changes as the situation changes. For example, assuming that the state of the curve w1 is a normal state and the moving speed of a certain object suddenly changes in the vicinity of the point P1 that the user 2 is paying attention to, the user 2 is paying attention to the change, so that the degree of interest is normal. The collision detection evaluation function I (x, y, z; Δr) represented by the curve w2 is obtained. In the curve w3, for example, a case where all the objects are dispersed from the periphery of the point P1 can be considered, and at this time, the degree of interest of the user 2 is lower. In addition, the strength of the collision detection evaluation function can be changed so as to reflect the emotional interest level information obtained from the physical state quantity of the user 2. As described above, these emotional interest level information can be acquired by measuring state quantities representing the body state including the user's heart rate, sweating rate, respiratory rate, and the like in the interest level acquisition step.

  Next, a collision detection evaluation function according to another interest level will be described. FIG. 11 shows an example of an angle-dependent collision detection evaluation function I (x, y, z; θ). When the user 2 is gazing in a certain direction of the space, the degree of interest of the user 2 in the separated space decreases as the angle is away from the line-of-sight vector n. Therefore, the collision detection evaluation function I (x, y, z; θ) is defined as a function whose function value decreases as the line-of-sight separation angle θ increases. The collision detection evaluation function I (x, y, z; θ) is axisymmetric about the line-of-sight vector n. The line-of-sight direction is specified by a normal line-of-sight measuring device. As a function form of the collision detection evaluation function I (x, y, z; θ), for example, the following equation (2) can be used.

  Here, as described above, g is a parameter for setting I (x, y, z; θ) to a finite value, and m is a function value adjustment parameter. Note that these parameters may have a gaze separation angle θ dependency. The collision detection evaluation function I (x, y, z; θ) represented by the equation (2) is a function that monotonously decreases with the line-of-sight separation angle θ, similarly to the equation (1) described above.

  FIG. 12 shows a collision detection evaluation function I (x, y, z; θ, ψ) having a dependency on the line-of-sight separation angle θ. As shown by the curves w5 and w6, the user's interest level is usually concentrated in the front, and when the visual line separation angle θ exceeds a certain angle ψ, the user's interest level is rapidly attenuated. As such a function, for example, a function such as I (x, y, z; θ, ψ) = a / {exp [b (θ−ψ) +1]}, a> 0, b> 0 is used. Can do. The magnitudes of these curves w5 and w6 reflect, for example, the temporal change in the interest level of the user 2 as in the case described above.

  Next, combinations of collision detection evaluation functions having different interest types will be described. The collision detection evaluation function determined by pointing as described above may be expressed by, for example, a function I (x, y, z; φ) that attenuates as the angle φ from the finger pointing direction increases. In addition, it may be expressed by I (x, y, z; s) that attenuates depending on the distance s from the finger position. It is also possible to combine the two. For example, using the constant a, I (x, y, z; φ, s) = a / (φ · φ · s · s).

  Furthermore, the collision detection evaluation function If = I (x, y, z; f) determined by pointing, the collision detection evaluation function Ih = I (x, y, z; h) determined by the head or face orientation, It is also conceivable to combine a collision detection evaluation function Ie = I (x, y, z; e) determined by For example, for the collision detection evaluation functions If, Ih, and Ie that can be regarded as having a substantially equal relationship as described above, I (x, y, z; f, h, e) is obtained by the sum of three types of collision detection evaluation functions. ) = If + Ih + Ie A collision detection evaluation function can be used.

  Next, the interest level (object interest level) O (m) for the virtual object m existing in the space will be described. The “object interest level” is defined for an object in order to hierarchically relate the user interest level information and the detail level of collision detection. In the sense that an object moves, “object interest” is a function of spatial coordinates. The collision detection evaluation function is a function in which an interest level is set for a specific partial space in the virtual three-dimensional space. The degree of interest in an object (object interest level) is that the object originally moves, so when an object appears in a space with a high degree of interest, the interest level of the object increases due to the influence of the degree of interest in that space In some cases, the object itself is of great interest to the user regardless of the position of the space where the object exists.

  In the above-mentioned case, that is, when the object is affected by the interest level of the space, the interest level of the object is, for example, the interest level in the space occupied by the object according to the position and orientation of the object in the space. Can be defined as the grand total. That is, the above-described collision detection evaluation function I (x, y, z; α) relating to the space can be calculated by integrating in the space region occupied by the object. Therefore, the degree of interest O (m) of the object m is obtained by the following equation (3).

  In the latter case described above, that is, when the object itself is of high interest to the user regardless of the position of the space in which the object exists, the degree of interest O (m1) of the object m1 is as follows: It is decided by the method. One is determined by the user setting the interest level O (m1) of the object at the start of operation of the system. Second, during the system operation, the interest level information of the user is acquired in the interest level acquisition step, and the interest level O (m1) of the object m1 is determined based on the information. Finally, in both cases, the interest level O (m1) of the changed object γ is determined by taking into account the temporal change in the interest level of the user. In any case, using the degree of interest O (m1) of the object m1 determined in this way, the detail level of the object expression and the detail level of collision detection are determined.

  Next, it will be described how to determine the detail level of the object using the interest levels O (m) and O (m1) of the object m and the object m1 and further to determine the detail level of collision detection. The object m will be described as a representative of the object. The interest level O (m) of the object m usually takes a continuous value, and Omin ≦ O (m) ≦ Omax, where the upper and lower limits are Omax and Omin. Further, the object m has a plurality of expression formats with different levels of detail, and the levels of detail take discrete and finite values. Each different representation format of the object m is labeled with a level of detail L (j), j = 0 to (n−1). L (0) is the most detailed, and L (n-1) is the coarsest detail. Using the above symbols, an argument j that gives the level of detail L (j) of the object m for a certain degree of interest Oj is determined by the following equation (4). However, the symbol [] sandwiching the formula on the right side of the formula (4) represents the rounded value of the formula value sandwiched between the symbols [].

  An explanation of the above equation (4) is shown in FIG. An argument j that obtains a degree of detail L (j) is obtained by converting a numerical value range 0 to (n−1) into an integer from a value that is proportionally distributed by (Omax−Oj) / (Omax−Omin).

  In the above equation (4), the upper and lower limits of the degree of interest correspond to the upper and lower limits of the detail level of the object. However, by providing an offset, the fluctuation in strength between the interest level at other time points and other objects is changed. Elements may be reflected. For example, with respect to an object that exhibits a lower overall interest level (object interest level) than other objects, the detail levels L (1) to L (n−1) excluding the expression format of the maximum detail level are the above. Equation (4) is applied.

  Next, the above-described space collision detection evaluation function I (x, y, z, α), the degree of interest O (m) of the object m, and the level of detail L (j) for the degree of interest Oj of the object m are used. A series of processing for detecting a collision in the real-time virtual space system (the above-described interaction analysis processing in FIG. 4 or FIG. 5) will be described. First, the system is activated and interaction between the user and space / object occurs. Therefore, the interest level information of the user is checked for each evaluation item of the interest level, and the collision detection evaluation function I (X, Y, Z; α) is generated or the previously generated collision detection evaluation function I (X, Y, Z). ; Α) is updated.

  Subsequently, based on the collision detection evaluation function I (X, Y, Z; α) in space, the object interest levels O (m1), O (m2),... Are calculated for each object m1, m2,. Is performed, the level of detail (expression form) of the object used for collision detection, and therefore the level of detail of collision detection, is determined. Next, using the shape model for collision detection of each object determined by the level of detail (expression format) of the object, it is checked whether each object has collided with each other, and it is further determined that a collision has occurred. If so, it is examined where and how it occurred.

  The collision detection step performed in the above-described interaction analysis process is characterized in that the detection operation with accuracy exceeding the detail level is discontinued and no further detection is performed. FIG. 14 shows an example of collision detection with such a limited degree of detail as an octree. When investigating the collision of the objects M1 and M2, the collision detection is performed from the coarsest model to the second coarsest level of detail L (n-2), but the detection is not interrupted after that. Yes.

  FIG. 15 shows collision detection when the detail levels of the objects k1 and k2 represented by polygons are determined as LOD (k1, j1) and LOD (k2, j2), respectively. In such collision detection between polygons, calculation of collision detection is performed between polygons constituting both objects. For example, if the number of polygons is P (k1) and P (k2), P (k1) × P (k2) inter-polygon calculations are performed. Such a calculation amount between polygons is such that the degree of detail of each object, and hence the number of polygons, is limited based on the degree of interest.

  Next, a virtual space generation device that uses any of the collision detection methods described above will be described. FIG. 16 shows a block configuration of the virtual space generation apparatus. The virtual space generation device 10 includes an interface unit 40 for performing interaction with the user 2 and a computer 50 for performing data processing and device control. The user 2 and the interface unit 40 are coupled to each other by bidirectional interaction, and the interface unit 40 and the computer 50 are coupled by bidirectional data communication. Below, the structure of each part is demonstrated.

  First, the interface unit 40 includes an input I / F unit 41 and an output I / F unit 42. The input I / F unit 41 receives an input from the user 2 and transfers data to the computer 50. Further, the output I / F unit 42 receives the output from the computer 50 and presents the output to the user by an image, sound, or the like in a manner that the user 2 can recognize.

  The input I / F unit 41 includes an operation device input unit 411 that inputs data from a device that is operated spontaneously by a user to generate an interaction. As the operation device, a pointer, an operation lever, a handle, or the like can be used. Moreover, it has the measuring device input part 412 which inputs data from the apparatus which detects the state change which generate | occur | produces in the user 2 as a result of interaction with a user and the virtual space generation apparatus 10. FIG. As the measuring device, a gaze direction measuring device that captures the movement of the pupil, and a general measuring device that measures posture, body temperature, sweating, brain waves, and the like can be used.

  The output I / F unit 42 presents data so as to appeal to the sense in order for the user 2 to recognize the virtual space generated by the computer 50 and the objects and events therein. For example, two-dimensional or three-dimensional object display by the image output unit 421, collision sound presentation by the voice output unit 422, pressure presentation to the user 2 by the sensation output unit 423, and the like are performed.

  Next, the computer 50 includes an input unit 51, an object posture calculation unit 52, a collision calculation unit 53, an output unit 54, and a time management unit 55. The input unit 51 includes a collision detection evaluation function creation unit 511 that processes the data from the interface unit 40 and evaluates the degree of interest of the user. Here, the collision detection evaluation function I (X, Y, Z; α) is generated or updated. The object posture calculation unit 52 includes a physical condition storage unit 521 that stores data for reflecting a physical law in a collision phenomenon or a change in the state of an object. These data, past output results, and Based on the current input information, the position and orientation of the object in the space are calculated.

  The collision calculation unit 53 includes an object expression storage unit 531, a detail level storage unit 532, and a collision detection unit 533, and detects a collision by changing the detail level according to user interaction. The object expression storage unit 531 stores object shape models expressed with different degrees of detail used for collision detection. Further, the detail level storage unit 532 integrates the collision detection evaluation function I (X, Y, Z; α), generates an object interest level H (m) for each object, and determines the detail level. An object shape model used for collision detection is determined. The collision detection unit 533 detects where and how a collision has occurred or has not occurred based on the surface position and normal direction of each object shape model.

  The output unit 54 creates collision result data to be presented to the user 2 based on the data detected by the collision calculation unit 532 and output about where and how each object collided. To do. This collision result data is passed to the interface unit 40 and presented to the user by the output I / F unit 42. The collision result data is data obtained by normalizing the upper and lower limits by the visual unit 541, the auditory unit 542, the kinesthetic unit 543, and the skin sensor unit 544 according to the form presented to the user.

  The time management unit 55 generates an event that occurs in the virtual space in accordance with a physical law, and performs time management of repeated cycles so as to ensure real-time properties in the space.

  With the configuration of the virtual space generation device 10 described above, the computer 50 generates a virtual space, and collision detection is performed based on a collision detection evaluation function that is hierarchically determined according to the degree of interest of the user 2 with respect to the partial space or an object in the space. The level of detail is dynamically changed and a collision is detected. It is possible to realize a device in which calculation resources for detailed collision detection are concentrated on necessary places, and both the collision detection accuracy and the collision detection speed for an object of interest are improved. As a result, the interactive activities of the user 2 are supported more effectively. The user 2 can interactively use the virtual space in a state in which self-projection is realized in the virtual space generated by such a virtual space generation device while achieving both autonomy (consistency) and real-time property. it can.

  The present invention is not limited to the above-described configuration, and various modifications can be made. For example, voxels or other expressions can be used as the object shape expression method. In addition, for an object of particular interest, the collision detection evaluation function value may be kept high even when the object does not exist in the user's field of view. The present invention is effectively applied in a wide three-dimensional space or a large screen immersive display. In addition, the present invention includes collisions in a broad sense such as separation and convergence of objects, such as operation training under a microscope such as surgery, proficiency training for assembly and disassembly of mechanical products, and spacecraft docking training in outer space. Effectively applied in the work field.

The conceptual diagram of the relationship between the interest level in the virtual three-dimensional space which concerns on one Embodiment of this invention, and the detail level of collision detection. (A) is a three-dimensional graph for explaining the size of a partial region of a virtual three-dimensional space determined from the degree of interest and the level of detail of collision detection in the virtual three-dimensional space according to an embodiment of the present invention, (b) to (c). Is a two-dimensional projection graph of the three-dimensional graph of (a). The conceptual explanatory view of the collision detection evaluation function concerning one embodiment of the present invention. The processing flowchart of the real-time virtual space system which changes a detail level according to the interaction which concerns on one Embodiment of this invention, and detects a collision. The flowchart which shows the interaction analysis process in the virtual three-dimensional space which concerns on one Embodiment of this invention. (A) (b) (c) is a figure of the shape expression of the object from which the detail differs based on one Embodiment of this invention. (A) and (b) are the figures which show the measuring method of the interest level which concerns on one Embodiment of this invention. FIG. 5 is a relationship diagram between the octree hierarchy and the collision detection detail according to an embodiment of the present invention. (A) is the relationship figure which applied the interest level at the time of the collision detection in the virtual three-dimensional space which concerns on one Embodiment of this invention, the detail level of collision detection, and the detail level of object expression, (b) is (a). The figure of the object shown to a user in a state. (A) is a perspective view which shows the structure of the collision detection evaluation function in the virtual three-dimensional space which concerns on one Embodiment of this invention, (b) is a graph of the collision detection evaluation function of (a). The perspective view which shows the structure of the other collision detection evaluation function in the virtual three-dimensional space which concerns on one Embodiment of this invention. The graph of the other collision detection evaluation function in the virtual three-dimensional space which concerns on one Embodiment of this invention. Description of the relationship between the degree of interest and the degree of detail according to an embodiment of the present invention. The figure of the octree showing the example of the collision detection which concerns on one Embodiment of this invention. The figure explaining the example of the collision detection which concerns on one Embodiment of this invention. 1 is a block diagram of a virtual space generation device according to an embodiment of the present invention. The field map figure explaining the application field of collision and interference detection. (A) (b) is explanatory drawing of the object collision in virtual space. The processing flow figure of the conventional real-time system in the virtual space with a user interaction. (A) is a perspective view showing object display and object collision in a virtual space, (b) is a perspective view of the virtual space divided into eight. (A) and (b) are diagrams of an octree that explains object collision detection by an octree.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Virtual space 2 User 5, 6, 7, 8 Object 10 Virtual space production | generation apparatus 50 Computer I Collision detection evaluation function O Interest degree of object

Claims (5)

A collision detection method for detecting a collision between a moving object and another object in a virtual space generated by a computer and interactively involved by a user in order to present the user to the virtual space,
An interest level acquisition step of acquiring user interest level information for a subspace of the virtual space and / or an object in the virtual space;
A level of detail determination step for dynamically changing and determining the level of detail of the object and the level of detail of collision detection according to the user's level of interest information acquired in the step, Method. An evaluation function generation step for generating a collision detection evaluation function as a function of spatial coordinates for hierarchically associating the interest level information and the level of detail of collision detection;
A collision detection step of detecting a collision based on the level of detail determined in the level of detail determination step,
In the detail level determination step, using the generated collision detection evaluation function, the detail level of the object in each subspace and the detail level of collision detection are dynamically changed and determined,
2. The collision detection method according to claim 1, wherein in the collision detection step, the detection operation with an accuracy exceeding the level of detail is stopped and no further detection is performed.   Each object in the virtual space has a plurality of representation formats having different degrees of detail, and the representation format of the object in each subspace is dynamically selected based on the collision detection evaluation function in the detail level determination step, The collision detection method according to claim 2, wherein collision detection is performed using the selected expression format in the collision detection step.   In the evaluation function generation step, the user's pointing, the distance from the pointer position, the distance from the user's viewpoint, the distance from the user or an important object in space, or the user's visual field center (gaze direction) A collision detection evaluation function is generated by increasing / decreasing a collision detection evaluation function value according to an angle from a user, an angle from a user's pointing or pointing direction, or a moving speed of an object crossing a user's work area or field of view. The collision detection method according to claim 2 or claim 3. A virtual space generation apparatus using the collision detection method according to claim 1.

Images