A cooperative grouping monitoring method based on real-time positioning

Abstract

The application relates to the technical field of virtual reality multi-person cooperative monitoring, in particular to a cooperative grouping monitoring method based on real-time positioning, which comprises the following steps: acquiring real-time positioning streams reported by experience terminals in the same group and content event scripts; determining event response windows corresponding to each content event according to event time, and constructing kinematic characteristic sequences of each experimenter in the event response windows based on the real-time positioning streams; performing cross-individual cross-correlation operation on the kinematic characteristic sequences of different experimenters in the group to obtain group coherence metrics; and determining experimenters with content step loss in the group and corresponding step loss amounts according to peak time delay, zero-lag coherence collapse degree relative to a group baseline, and the like.

Description

Technical Field

[0001] This application relates to the field of virtual reality multi-person collaborative monitoring technology, and in particular to a collaborative group monitoring method based on real-time positioning. Background Technology

[0002] With the development of virtual reality technology, large-space, multi-person, free-walking virtual reality experience projects have emerged in game venues, exhibition halls, and other settings. In these projects, multiple participants wearing experience terminals walk freely in groups within a physical space of hundreds of square meters, while a monitoring terminal monitors the operational status of each group and coordinates the broadcast.

[0003] Traditionally, monitoring of multi-person experience environments relies primarily on the following methods: First, real-time positioning and map display, marking the location of each participant on a site map, using the location display as the monitoring endpoint; second, electronic fences and collision detection, determining the geometric relationship between individuals and preset boundaries; third, content synchronization detection based on heartbeats or timestamps, with each terminal periodically reporting playback progress; and fourth, video crowd statistics and equipment status monitoring. Furthermore, regarding positioning errors, the drift in Simultaneous Localization and Mapping (SLAM) accumulates continuously over time, traditionally requiring calibration using additional anchor points or external reference hardware for mutual ranging.

[0004] However, the above methods have obvious drawbacks: the walking behavior of the experiencer who is out of sync with the content is completely normal. They never cross the boundary, their position trajectory is smooth, and their heartbeat can appear normal due to the blocking of the rendering thread. Therefore, in principle, position display, electronic fence, heartbeat detection and video statistics cannot detect the fault of "walking normally but the content timeline is lagging or ahead", which makes it impossible to detect and quantify the content out of sync in a timely manner. Summary of the Invention

[0005] Therefore, it is necessary to provide a collaborative group monitoring method based on real-time positioning to address the aforementioned technical problems.

[0006] This application provides a collaborative group monitoring method based on real-time positioning, the method comprising: Obtain real-time location streams reported by each experience terminal within the same group, as well as content event scripts; wherein, the content event scripts record the event moments of content events used to guide the movement of experiencers within the group; The event response window corresponding to each content event is determined based on the event time, and the kinematic feature sequence of each experiencer within the event response window is constructed based on the real-time positioning stream; Cross-individual cross-correlation calculations were performed on the kinematic feature sequences of different participants within a group to obtain the intra-group coherence metric; the intra-group coherence metric includes at least peak delay and zero-hysteresis coherence. Based on the peak latency and the degree of collapse of the zero-hysteresis coherent relative group baseline, determine the users within the group who have content out-of-sync and the corresponding amount of out-of-sync.

[0007] The aforementioned collaborative group monitoring method based on real-time positioning uses the content events recorded in the content event script as a common excitation signal injected into the motion channels of the entire group. The kinematic feature sequences of group members within the event response window serve as motion benchmarks for each other. By performing cross-individual cross-correlation operations on the kinematic feature sequences of different participants, peak coherence, peak delay, and zero-hysteresis coherence are obtained. This allows participants who are walking normally but whose content timeline is lagging or ahead to exhibit a drop in peak delay and a collapse of the relative group baseline under the plot clock reference system. Thus, content layer faults can be quantitatively inverted from the positioning data. This fault is theoretically invisible under electronic fences, heartbeat synchronization, and video statistics. However, this application can detect and quantify the amount of step loss in real time using only the existing real-time positioning stream of the experience terminal and deploying it purely in software on the server side, without adding any new sensors or mutual ranging hardware. Furthermore, the quantified amount of step loss is directly used as the target content time offset of the resynchronization command to drive precise resynchronization without restarting the experience. This is beneficial for improving the real-time monitoring, fault detection capability, and handling accuracy of large-space multi-person virtual reality experiences. Attached Figure Description

[0008] Figure 1 This is an application environment diagram of a collaborative group monitoring method based on real-time positioning in one embodiment; Figure 2 This is a flowchart illustrating a collaborative group monitoring method based on real-time positioning in one embodiment; Figure 3 This is a schematic diagram of cross-individual cross-correlation and coherence curves within an event response window in one embodiment; Figure 4 This is a schematic diagram of differential drift estimation and recharge correction in one embodiment; Figure 5 This is a schematic diagram of dynamic group adjustment and coherent reference write-back in one embodiment; Figure 6 This is a schematic diagram illustrating the linkage between group health monitoring and broadcast control in one embodiment; Figure 7 This is a structural block diagram of a collaborative group monitoring device based on real-time positioning in one embodiment; Explanation of reference numerals in the attached figures: 101. Experience terminal; 102. Server; 103. Monitoring and broadcast control terminal. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0010] The collaborative group monitoring method based on real-time positioning provided in this application can be applied to, for example... Figure 1 The application environment shown. Figure 1 This is an application environment diagram of a collaborative group monitoring method based on real-time positioning in one embodiment. (Refer to...) Figure 1 The application environment includes an experience terminal 101, a server 102, and a monitoring and control terminal 103. The experience terminal 101 and the monitoring and control terminal 103 communicate with the server 102 via the network. The same group includes multiple experience terminals 101.

[0011] The experience terminal 101 is a virtual reality display terminal worn on the user's head. In one embodiment, it is a VR (Virtual Reality) headset running the Android operating system. It integrates sensing components for positioning and calculates the user's position in the physical environment in real time using SLAM (Simultaneous Localization and Asynchronous Sensory Management) and Ultra Wide Band (UWB) fusion positioning methods. It should be noted that SLAM refers to a positioning technology where, when the experience terminal moves in an unknown environment, it estimates its own pose and incrementally builds an environmental map based on sensor data from its onboard cameras, inertial measurement units, etc.; UWB refers to a wireless positioning technology that uses nanosecond-level narrow pulse signals for time-of-flight ranging to achieve absolute positioning within a pre-deployed anchor point area. The experience terminal 101 can be, but is not limited to, various VR headsets, augmented reality headsets, smart glasses, and other wearable terminal devices.

[0012] The server 102 is used to collect real-time location streams reported by each experience terminal 101 within the same group, and to perform coherent calculation, out-of-step detection, drift estimation, backfeed correction, dynamic group adjustment, and group health synthesis in the collaborative group monitoring method provided in this application embodiment. That is, the server 102 runs as a coherent engine. The server 102 can be implemented using a standalone server or a server cluster composed of multiple servers.

[0013] The monitoring and control terminal 103 is used to obtain monitoring data such as group health, coherence distribution, out-of-sync volume, and drift rate of each group from the server 102 and present it to the on-site operations personnel. When it is confirmed that the content has lost sync or drifted, it triggers broadcast control linkage actions, such as issuing a resynchronization command to the designated experience terminal 101. The monitoring and control terminal 103 can be, but is not limited to, various personal computers, laptops, workstations, etc. In one embodiment, it is PC-based monitoring and control software.

[0014] The aforementioned application environment is anchored to VR large-space multi-person experience projects in venues such as game stadiums and exhibition halls. In such projects, the experience venue is a physical space of several hundred square meters. Participants are grouped into groups of 4 to 8 people, and multiple groups experience the experience in parallel within the venue. Each participant wears an experience terminal 101 and moves freely within the venue. The timeline of the narrative content determines the participant's movement goal and rhythm. Specifically, the opening direction of the virtual door determines the participant's displacement goal, the direction of the monster's attack determines the participant's evasion speed vector, the guide light strip determines the participant's movement path, and the convergence aperture determines the participant's convergence coordinates. The server 102 collects the real-time location streams from each experience terminal 101, and the monitoring and control terminal 103 monitors the operating status of each group and coordinates the broadcast control.

[0015] It should be noted that the core mechanism underlying the above scenario is that, since all participants in the same group are stimulated by the same content timeline, the content event is equivalent to a common stimulus signal injected into the entire group's motion channel. Therefore, the walking-level kinematic signals (including speed, turning, starting, and stopping) of group members within the event response window corresponding to the content event serve as motion benchmarks for each other. Based on this mechanism, the embodiments of this application can use the existing real-time positioning stream of the experience terminal 101 as raw material to quantitatively invert content layer faults and positioning layer drift from positioning data without adding any new sensors or mutual ranging hardware. This mechanism is the overall principle behind the following steps.

[0016] To ensure that the formulas used below are strictly unambiguous, this embodiment adopts a unified convention for symbols and dimensions throughout the text: Indicates the experiencer At any moment The real-time location of the stream is measured in meters (m). ; , , These represent velocity vector, velocity, and smooth velocity, respectively, with the dimension being meters per second (m / s). , , These represent normalized cross-correlation, coherence, and group health, respectively, all of which are dimensionless quantities. , , These represent time delay, step loss, and event time, respectively, with the unit being seconds (s). , , These represent the position offset, cumulative travel distance, and rendezvous point, respectively, with the unit being meters (m). This represents the drift rate, a dimensionless quantity expressed in meters per meter (m / m). Index for content events; The number of content events in the set class that participate in the drift estimation (i.e., the number of samples of each member's position offset). Grouping The number of members within.

[0017] Among them, the content event index is uniformly used throughout the entire text. This is to avoid using the symbol for the unit of length, the meter. This produces ambiguity. (Subscript) , Used for indexing testers, Used for index grouping.

[0018] Under the above-mentioned notation, the methods provided in the embodiments of this application will be described in detail below.

[0019] In one embodiment, such as Figure 2 As shown, Figure 2 This is a flowchart illustrating a collaborative group monitoring method based on real-time location in one embodiment. The method is applied to... Figure 1 Taking server 102 as an example, the process includes the following steps S202 to S208.

[0020] Step S202: Obtain the real-time location streams and content event scripts reported by each experience terminal within the same group; Among them, the real-time location stream refers to the time series of the user's location periodically reported by the experience terminal 101. Specifically, the user's location... The corresponding experience terminal 101 uses a sampling frequency Report location The unit of measurement is meters (m). This reporting stream is generated by the fusion positioning module inside the experience terminal 101: the output frequency of the SLAM positioning channel is 60 to 90 Hz, and the output frequency of the UWB positioning channel is 10 Hz. The two are then fused and thinned to a unified frequency. Report it to higher authorities.

[0021] Based on this, the location streams reported by each experience terminal 101 also need to be time-aligned. Specifically, the experience terminal 101 and the server 102 synchronize their time using NTP (Network Time Protocol) or PTP (Precision Time Protocol).

[0022] Within this group, the same group is identified through group registration. Specifically, server 102 divides the entering participants into several initial static groups. Each group The initial static grouping consists of 4 to 8 people. The initial static grouping can be determined by means of on-site formation, ticket batching, registration relationship, etc. It should be emphasized that the initial static grouping is only the starting point for coherent calculation, and it will be dynamically rewritten based on coherence measurement in the following example of dynamic grouping.

[0023] The content event script records the event times of content events used to guide the movement of participants within a group. The event time is the moment the content event occurs on the content timeline. Specifically, the meaning of content events guiding participant movement is that the content script imposes hard constraints on the participant's walking-level movement—guiding events specify the participant's path and starting time, avoidance events specify the participant's avoidance velocity vector, startle events specify the participant's sudden stop or turn, and convergence events specify the participant's convergence coordinates. In a free-roaming, large space, the plot cannot progress without displacement; movement itself is the interaction channel. Therefore, the kinematic characteristic sequence of group members within the event response window must exhibit calculable cross-individual coherence.

[0024] The output of step S202 is the time-aligned real-time location stream for each participant. And the event time sequence recorded by the content event script. In the above embodiments, by using a unified reporting frequency and a relative time base within the group, the real-time location streams of each experience terminal are comparable across individuals, and there is no need for a global absolute clock or terminal-by-terminal hardware synchronization, which provides a data foundation for subsequent cross-individual intercorrelation calculations.

[0025] Step S204: Determine the event response window corresponding to each content event based on the event time, and construct the kinematic feature sequence of each experiencer within the event response window based on the real-time positioning stream.

[0026] The event response window refers to the time interval during which the user responds to a content event, denoted as . , and These represent the start and end times of the participant group's response to this type of event, respectively, with a typical window length of 5 seconds. The kinematic feature sequence refers to the time series extracted from the real-time location stream that characterizes the participant's walking-level motion state. Specifically, for each participant... Perform the following processing: First, perform a first-order difference on the position flow to obtain the velocity vector. ,in Secondly, the speed is obtained by taking the magnitude of the velocity vector. The dimension is meters per second (m / s).

[0027] Furthermore, to suppress the high-frequency noise introduced by positioning jitter, a first-order low-pass filter is applied to the rate sequence to obtain a smoothed rate.

[0028]

[0029] Where the filter coefficients .

[0030] In addition, pause markers are constructed based on smooth rates. That is, when the smoothing rate is below 0.05 m / s, the sampling point is determined to be in a pause state, which is used for subsequent calculation of the pause ratio; in one embodiment, the turning rate can also be further calculated. (Measured in radians per second (rad / s)) as a supplementary component of the kinematic characteristics. The output of step S204 is the kinematic characteristic sequence of each participant within each event response window. .

[0031] Step S206: Perform cross-individual cross-correlation calculation on the kinematic feature sequences of different participants within the group to obtain the intra-group coherence metric; Specifically, cross-individual correlation operations are performed only within the event response window of the emphasized event. Combined with... Figure 3 , Figure 3 This is a schematic diagram of cross-individual cross-correlation and coherence curves within an event response window in one embodiment. Specifically, for any two participants within a group... , The kinematic feature sequence is slided relative to the event response window with each time delay, and the normalized cross-correlation of the two sequences at each time delay is calculated. The result is the result of the cross-individual cross-correlation operation. The specific calculation formula of the normalized cross-correlation and the details of the overlapping segment processing are shown in the corresponding embodiment below.

[0032] The intragroup coherence metric includes at least peak delay and zero-hysteresis coherence. Specifically, the functions of each quantity in the intragroup coherence metric are as follows: Peak coherence is the maximum value of the cross-individual cross-correlation result at each delay, measuring the consistency of the waveforms of two kinematic characteristic sequences, independent of delay, and used for the calculation of group affinity in dynamic grouping below; Peak delay is the delay corresponding to the maximum value of the cross-individual cross-correlation result, measuring the experience of the user. relative experiencer The motion response delay; zero-lag coherence is the result of cross-correlation operation across individuals when the time delay is zero, which measures the degree of coherence between two sequences under strict temporal alignment conditions and reflects the degree of temporal alignment.

[0033] Understandably, the user experiencing content synchronization issues exhibits completely normal walking behavior, with their rate waveform consistent with that of their group members. The only difference is that the entire user is shifted by a synchronization loss amount within the content clock reference frame. This waveform shift does not alter the waveform's shape; therefore, the peak coherence between the synchronized user and other members remains high, with only the zero-hysteresis coherence collapsing due to the shift. In other words, synchronicity can only be measured by zero-hysteresis coherence, and peak coherence is inherently insensitive to synchronization issues—this is precisely the physical fulcrum for "inverting content-layer faults from positioning data": content synchronization issues are invisible under methods such as electronic fences, video statistics, and heartbeat synchronization, but only manifest as zero-hysteresis coherence collapse and peak delay jumps within the story clock reference frame.

[0034] The output of step S206 is the set of in-group coherence metrics. Based on this, the following embodiments further define member coherence. Group coherence In the above embodiments, by dividing the work of three metrics—peak coherence, peak latency, and zero-hysteresis coherence—the experiencer who walks normally but loses their footing in the content appears in the form of peak latency jumps and zero-hysteresis coherence collapses in the context of the story clock reference system, thereby enabling the observability of content-layer faults inverted from positioning data.

[0035] Step S208: Based on the peak latency and the degree of collapse of the zero-hysteresis coherent relative group baseline, determine the experiencers within the group who have content out-of-sync and the corresponding out-of-sync amount.

[0036] Content out-of-sync refers to the content timeline presented by the experience terminal lagging behind or ahead of the content timeline of other experience terminals in the group; the out-of-sync amount represents the duration of this lag or advance; the group baseline refers to the coherence reference level of normal responding members in the group, which is taken as the median of the coherence of each member in the group in one embodiment. The content out-of-sync discrimination process includes two-stage processing: first, response gating is performed, and then out-of-sync discrimination is performed on those who pass the gating. The above causal order must be clarified first: response gating is a preliminary discrimination step - experiencers who do not respond to content events (such as tying their shoelaces in place, answering the phone, or deliberately staying still) will also have their zero-lag coherence collapse due to the lack of motion response. If such members are not removed first, their collapse will be misjudged as content out-of-sync, thus confusing the two completely different situations of "no response" and "response but delay". The specific limitations of response gating and out-of-sync criteria are shown in the corresponding embodiments below.

[0037] The output of step S208 is: a step loss alarm and a quantified step loss amount for the step loss experiencer. Alternatively, it can provide on-site handling prompts for unresponsive users. In the above embodiments, through a two-level structure of "gating first and then judging," it is possible to cross-confirm content synchronization loss based on two independent pieces of evidence: the estimation of the amount of synchronization loss and the coherence collapse. At the same time, it eliminates the interference of unresponsive members from a mechanism perspective, providing a reliable parameter source for subsequent accurate resynchronization.

[0038] It should be noted that the above steps share the same coherent computation kernel for cross-individual cross-correlation, and there is a causal dependency between each step and the scheme below: the out-of-step amount quantified in step S208 is the parameter source of the resynchronization instruction, the confidence of the backfeed correction is directly taken from the member coherence defined in step S206 and its embodiment, and the group registration information updated by dynamic grouping is written back to step S202 as the coherence benchmark for subsequent cross-individual cross-correlation calculation, thus forming a circular closed loop of "measurement-grouping-remeasurement".

[0039] In one embodiment, cross-individual cross-correlation is performed on the kinematic feature sequences of different participants within a group to obtain an intra-group coherence metric, including the following process: For each delay Two participants were selected. , Overlapping segments of kinematic feature sequences within the event response window .in, Indicates the event response window Overall translation delay The interval obtained afterwards The geometric meaning is: when the experiencer The sequence relative experiencer Sequence sliding delay At that time, the two sequences still cover the same time interval. Only in the overlapping segment... The mean is removed and the energy is calculated. The product sum of the two kinematic feature sequences within the overlapping segment is normalized to obtain the normalized cross-correlation at this time delay:

[0040]

[0041] in, , These are two sequences in the overlapping segment. The mean of the above.

[0042] It should be noted that both the mean removal and energy reduction occur only in the overlapping segment. The reason for calculating on the upper part of the window, rather than on the entire window, is that: with time delay... As the values ​​change, the position and length of the overlapping segments also change. If the mean and energy are fixed at the entire window diameter, then the differences will be significant. The normalization benchmarks are inconsistent, resulting in different coherence curves. Different time delays can lead to systematic biases; however, by re-estimating the mean and energy on the overlapping segments with each time delay, the estimation accuracy can be kept consistent under different time delays, thus obtaining an unbiased coherence curve.

[0043] Furthermore, by the Cauchy-Schwarz inequality, for any sequence of real numbers... and Hengyou Applying this inequality to the two sequences after removing the mean in the above equation, the square of the numerator is always no greater than the product of the energies of the two terms in the denominator. This property holds true consistently. The physical meaning of this boundedness is that the coherence degree has a uniform and comparable range of values—1 indicates that the two sequences are completely linearly identical. 0 indicates complete inversion, and 0 indicates complete incoherence, thus enabling coherence measurements between different member pairs, different event windows, and different groups to be compared and thresholded on the same scale.

[0044] Among these, the absolute value of the delay is no greater than the preset delay upper limit, i.e., the effective delay range. τ ≤τmax; and the number of samples in the overlapping segment is not less than the preset minimum number of samples. The reason for limiting the preset upper limit of latency and the minimum number of overlapping samples is because when When the overlap segment shortens drastically beyond half the window length, the statistical degrees of freedom of the cross-correlation estimate are insufficient, and the estimation variance is unacceptable. The two constraints mentioned above together ensure the statistical validity of the coherent estimate.

[0045] Specifically, in the normalized cross-correlation at each time delay, the maximum value is taken as the peak coherence. The delay corresponding to the maximum value is taken as the peak delay. (The unit of measurement is seconds (s), by convention) Lagging behind (For positive values), the normalized cross-correlation with zero delay is taken as the zero-lag coherence. .

[0046] Based on the aforementioned pairwise coherence measures, we further define membership coherence and group coherence. Among them, membership coherence... experiencer The mean of zero-hysteresis coherence among all participants in the group serves two purposes: firstly, it helps the content loss-of-step assessment stage determine the degree of coherence collapse for that participant; secondly, it provides the correction confidence level for the refeedback correction stage. (Group coherence)

[0047]

[0048] That is, grouping The mean of zero-lag coherence among all members, where The reciprocal of the number of members with respect to the total number. Furthermore, the weighted group coherence throughout the process.

[0049]

[0050] That is, the group coherence corresponding to each event response window is weighted and averaged according to the event weight, and the free exploration period is due to Automatically excluded from synthesis. In the above embodiments, re-estimating the mean and energy on the overlapping segment time-delay ensures the unbiasedness and comparability of coherent estimation, enabling coherent metrics to be compared and thresholded on the same scale among members, between event windows, and between groups.

[0051] In one embodiment, based on the peak latency and the degree of collapse of the zero-hysteresis coherence relative to the group baseline, the users within the group who are experiencing content out-of-sync and the corresponding amount of out-of-sync are determined, including the following process: For testers The median of the peak latency from the remaining participants in the group to the participant in question is taken as the participant's step loss.

[0052]

[0053] The median was chosen instead of the mean because when out-of-sync or abnormal members are mixed in within a group, the peak delay of individual contamination will be significantly increased. The proportion of outliers skews the mean, while the median collapse point is 50%. As long as the outlier count is less than half, the outlier estimation is not contaminated by single points. Secondly, the participant's membership coherence is determined based on the zero-lag coherence between the participant and the other participants in the group. (The definitions in the above embodiments are used).

[0054] The criterion for missing synchronization consists of three elements: first, First, the absolute value of the step loss is greater than the step loss threshold; second, member coherence. The collapse of the relative baseline is greater than the collapse threshold. Third, at least two consecutive event response windows must satisfy both of the above conditions (persistence criterion). When all three conditions are met simultaneously, the user is determined to have lost synchronization with the content. For example, the synchronization threshold... Its statistical meaning is as follows: The distribution of response delays to content events among the normal population is shown, with the peak delay of normal members approximately following a zero mean and a zero standard deviation. The distribution, with a threshold of 3 times the scatter, means that the probability of a normal member exceeding the threshold is controlled to the thousandths level, and false alarms are manageable; collapse threshold Its value is much smaller than the actual collapse caused by the loss of synchronization, and the discrimination interval is sufficient. The persistence criterion requires that at least two consecutive event windows be established. Its purpose is to prevent instantaneous disturbances (such as an accidental pause or a positioning jump) within a single event window from triggering false judgments. Since the content clock loss of synchronization persists in all subsequent event windows once it occurs, while instantaneous disturbances do not have cross-window persistence, this criterion can distinguish between persistent faults and instantaneous disturbances from a mechanism perspective. In the above embodiments, the median loss of synchronization estimation is resistant to single-point contamination, the three-fold scatter threshold makes false alarms controllable, and the persistence criterion distinguishes between persistent faults and instantaneous disturbances from a mechanism perspective, ensuring the robustness of the loss of synchronization determination.

[0055] In one embodiment, before determining the users within a group who have lost synchronization with the content and the corresponding amount of synchronization loss, the following process is also included: Response gating: Calculates the velocity energy of each participant within the event response window.

[0056]

[0057] That is, the mean of the squared smooth rate within the event response window, and the percentage of pauses.

[0058]

[0059] This refers to the percentage of samples where the pause flag within the event response window is true; it is dimensionless. When the experiencer's speed energy is below an energy threshold... If the pause percentage exceeds a threshold of 0.5, the user is deemed unresponsive to the content event and transferred to the on-site handling channel, but is not included in the content out-of-sync judgment. If the user's speed energy is not lower than the energy threshold and the pause percentage is not higher than the percentage threshold, the user is deemed responsive, and subsequent out-of-sync judgment is applied. In the above embodiment, the response gating first removes members who are unresponsive to the content event to avoid their zero-hysteresis coherence collapse being misjudged as content out-of-sync, ensuring that out-of-sync judgment only applies to responsive members and guaranteeing the mechanistic purity of the judgment result.

[0060] In one embodiment, the method further includes: for a collection-type content event, obtaining the position of each participant at the time of the event, and calculating the position offset of each participant relative to the group reference; the group reference is the median of the positions of the other participants in the group; regressing the position offset of each participant along the participant's cumulative travel distance to obtain the participant's drift rate; determining the drift vector based on the drift rate and cumulative travel distance, and performing backfeed correction on the real-time positioning stream according to the drift vector.

[0061] This embodiment addresses the cumulative drift of the SLAM positioning channel: SLAM positioning error monotonically accumulates with cumulative travel, while the distribution of participants' positions at the assembly point is a zero-mean random quantity. Statistically, these two are necessarily separable; therefore, differential drift can be separated solely from the continuous "member relative group offset" without the need for mutual ranging hardware. Combined with... Figure 4 , Figure 4 This is a schematic diagram of differential drift estimation and recharge correction in one embodiment. The processing in this embodiment is as follows.

[0062] Specifically, at the event moment of a collection-type content event. To obtain the location of each participant ; for testers Calculate its position offset relative to the group reference.

[0063]

[0064] The unit of measurement is meters (m). The group baseline is the median position of the remaining participants within the group. The median is used as the group baseline because its collapse point is 50%: when less than half of the participants in the group simultaneously exhibit drift or loss of synchronization, the positions of these aberrant participants do not contribute to determining the median position, thus preserving the group baseline and ensuring accurate positional offset. A true reflection of the experience The drift is due to its own positioning rather than the movement of the reference point.

[0065] Furthermore, the positional offset of each participant is regressed along their cumulative travel distance to obtain their drift rate. Among them, cumulative mileage (Dimensions are meters (m)). The drift vector is determined based on the drift rate and cumulative travel. The magnitude of the drift vector is the product of the drift rate and the cumulative travel, and its direction is the drift direction. Further, the real-time positioning stream is re-compensated based on the drift vector; that is, the reported real-time positioning stream is compensated in reverse using the estimated drift vector, gradually deducting the accumulated drift from the positioning stream. The specific implementation of the re-compensation is described in the corresponding embodiment below; the output of this embodiment is the drift vector for each participant. And its drift determination results.

[0066] It should be noted that the applicable conditions and failure modes of this embodiment are explicitly declared as follows: First, the station distribution must be approximately independent and identically distributed in the travel dimension and have zero mean—if the formation changes systematically with the plot, a spurious trend will occur; the countermeasures are the median benchmark and Theil-Sen robust regression; Second, SLAM loop jumps manifest as outliers in the regression sequence, which are suppressed by robust regression; Third, less than half of the members must drift simultaneously, otherwise the median benchmark will cross the collapse point; Fourth, station distribution... The system must be stable; even in unstable conditions, the false alarm rate can still be kept constant through significant verification. Fifth, common model drift (unidirectional translation or rotation of the entire group) is unobservable in intra-group differential analysis, and is covered by existing UWB anchor points—the accuracy of the common model is limited by the anchor points' own error of 10 to 30 cm. This solution is responsible for the differential layers below this error limit; the two have clearly defined roles and do not overlap. In the above embodiments, through the differential architecture of "group as benchmark," the monotonically accumulating SLAM drift with travel is statistically separated from the zero-mean station distribution, achieving quantitative estimation and re-feedback correction of drift without adding any new mutual ranging hardware.

[0067] In one embodiment, the positional offset of each participant is regressed along the participant's cumulative travel distance to obtain the participant's drift rate, including the following process: offset the position Along two orthogonal coordinate axes (site coordinate system) shaft and The axis is decomposed to obtain two orthogonal components. and The two orthogonal components obtained from the decomposition are then decomposed along the cumulative travel path. Perform linear regression to obtain the slope vector. Drift direction Dimensionless; drift rate Dimensionless; therefore, the drift vector In one embodiment, conventional least squares (OLS) regression is used; in another embodiment, Theil-Sen robust regression is used to suppress outliers from contaminating the slope estimate when there are SLAM loop jumps or stationary nonstationarities.

[0068] Norms for positional offsets are prohibited here. The reason for direct regression is that if we directly use the scalar norm... For cumulative travel regression, under zero-mean two-dimensional Gaussian noise, the two-dimensional vector norm follows a Rice distribution rather than a Gaussian distribution: the noise causes the norm to be positively skewed (even if the true offset is zero, the expectation of the norm is positive, i.e., the noise flattening effect), and the norm's response to the true offset exhibits curvature in the small offset region; when performing linear regression on this norm sequence containing positive skew and curvature, the intercept term absorbs part of the signal that increases with travel, resulting in the slope being systematically underestimated.

[0069] Furthermore, a significance test is performed on the slope vector. Specifically, the following methods are used: Verification: Molecular degree of freedom 2 corresponds to two slope parameters ( and (One for each), Degrees of freedom in the denominator Corresponding residual degrees of freedom—each of the two coordinate axes contains A total of observations Each observation, with two parameters (slope and intercept) per axis of regression, consumes a total of 4 degrees of freedom; therefore, the residual degrees of freedom are... For example, the significance level is taken as... It is important to emphasize that the reason for using a significance test instead of a fixed amplitude threshold is as follows: The test threshold is automatically determined by the residual variance and varies with the station location. Adaptive adjustment, thereby making the false alarm rate consistent with... Decoupling and maintaining constancy. It should be noted that the above... The test assumes that the residuals of the two orthogonal components are approximately independent and homoscedastic; in the embodiment using Theil-Sen robust regression, the significance test can be replaced by a permutation test. Verification is conducted to maintain consistency in the definition of false alarm rates.

[0070] The drift criterion is: the slope vector passes the significance test, and the drift rate... greater than the drift rate threshold Furthermore, the participant's step loss was approximately zero. When all three conditions are met, the user is determined to be drifting rather than losing their footing. The role of this element is: in (peak delay) Low-frequency envelope divergence On the two-dimensional feature plane, drift manifests as low-frequency spatial divergence—peak delay is approximately zero, and the drift rate increases with travel distance; loss of synchronization manifests as a delay phenomenon—peak delay jumps and zero-hysteresis coherent collapse. These two types of faults are naturally separable on this plane; therefore, introducing the condition that the loss of synchronization is approximately zero can avoid misinterpreting delay shifts as drift. In the above embodiment, biaxial regression eliminates the systematic underestimation of the norm, and the significance test decouples the false alarm rate from the station distribution. The two-dimensional criterion (peak delay, drift rate) separates drift and loss of synchronization faults mechanistically.

[0071] In one embodiment, the real-time positioning stream is re-corrected according to the drift vector, including the following process: For those who confirmed the existence of drift Position correction is performed using the following formula:

[0072]

[0073] Among them, the correction coefficient This means performing soft correction based on a certain percentage of the drift vector, rather than a one-time full correction. The reason for using soft correction is that if the drift vector were deducted in full at once, the user's position trajectory would experience a step change, disrupting the continuity of the trajectory and the stability of downstream applications (such as map display and collision avoidance); gradual reinjection based on the travel ratio allows the correction amount to accumulate smoothly with the travel, maintaining the continuity of the trajectory.

[0074] Furthermore, the corrected confidence level is determined based on the member coherence of each participant, and the corrected confidence level is then used. This directly reuses the member coherence (mean of zero-hysteresis coherence between the participant and the other participants in the group) defined in the above embodiments. In one embodiment, the confidence threshold is, for example, 0.5 to 0.7: the member coherence of normal members is usually higher than 0.9, while the member coherence of out-of-sync or out-of-group members collapses significantly relative to the group baseline, falling below the threshold. When a participant's corrected confidence is lower than the confidence threshold, the refeedback correction for that participant is frozen, and that participant is excluded from the calculation of the group baseline; when a participant's corrected confidence is not lower than the confidence threshold, the real-time positioning stream of that participant is corrected according to the product of the correction coefficient and the drift vector.

[0075] The protection logic of confidence refeeding is as follows: First, the kinematic characteristics of out-of-sync or outlier members have lost coherence with the common excitation within the group, and their positional offset contains non-drift components such as time delay misalignment. The drift vector obtained from this regression is unreliable. If refeeding correction is performed on them, erroneous correction amounts will be injected into their positioning flow. Second, the median benchmark is only uncontaminated when less than half of the members participating in the calculation are anomalous. In this application, the group consists of 4 to 8 people, and the group benchmark consists of only 3 to 7 samples, with limited robustness margin. Excluding confirmed anomalous members from the calculation of the group benchmark in advance can prevent them from being superimposed with subsequent new anomalous members (such as another member drifting) and causing the median benchmark to exceed the 50% collapse point. Therefore, content out-of-sync discrimination and confidence freezing in this embodiment together constitute a dual verification of calibration—out-of-sync discrimination removes anomalous members from the time delay dimension, and confidence freezing removes anomalous members from the coherence dimension. Together, they ensure that the median benchmark and refeeding correction only apply to normal members. In the above embodiments, soft correction ensures trajectory continuity, confidence gating avoids the injection of erroneous correction amounts, and the corrected positioning stream can serve as a stable input for downstream applications.

[0076] In one embodiment, the method further includes: for each participant, calculating the peak coherence between the participant and participants in each group, and averaging the peak coherence using a rolling window to obtain the participant's group affinity for each group; when the difference between the participant's group affinity for another group and its group affinity for the current group is greater than the group adjustment threshold and continues for a preset window number, reclassifying the participant to another group; updating the group registration information, and using the updated group as the coherence benchmark for subsequent cross-individual cross-correlation calculations.

[0077] Among them, group affinity refers to the statistical measure of peak coherence between the participant and each member within a given group. Combined with... Figure 5 , Figure 5 This is a schematic diagram of dynamic group adjustment and coherent reference write-back in one embodiment. Specifically, for each user... Calculate its relationship with each group separately. Peak coherence among internal experiencers The peak coherence was averaged using a rolling window to obtain the participant's group affinity for each group.

[0078]

[0079] The number of windows in a scrolling window is An event response window. When the user interacts with another group... If the difference between the user's affinity for their current group and their affinity for the group they belong to exceeds the group adjustment threshold, and this continues for a preset number of windows, the user will be reassigned to another group. satisfy And continue The window will then allow the experiencer to... The judgment was changed to a group. Among them, the adjustment group threshold Preset number of windows .

[0080] It should be noted that peak coherence, rather than zero-latency coherence, is used for group adjustment. The reason is that the problem in group adjustment is whether the user follows the same storyline waveform of a particular group. Waveform consistency is not related to latency—even if the user is out of sync, their waveform shape still matches the storyline, and peak coherence remains valid. If zero-latency coherence were used, the group affinity of out-of-sync members would be systematically low, leading to misjudgment as being in the wrong group, thus confusing "out-of-sync" and "wrong group" faults. Group Adjustment Threshold With continued The purpose of setting up windows is to prevent shaking: the coherent fluctuations of a single window are random, requiring the difference in affinity to exceed the threshold and for three consecutive windows to be valid, which can prevent members from repeatedly shaking and changing their judgments between two groups.

[0081] Typical scenarios handled by dynamic grouping include: a participant joining the wrong group and following another group's storyline; a participant joining a group midway through the process; and misclassification of ticket batches with actual group assignments. Further, after grouping is completed, the group registration information is updated, and the updated group is written back to step S202, using the updated group as the coherence benchmark for subsequent cross-individual cross-correlation calculations. The reason for writing back is that in this scheme, the group is not a broadcast control permission label, but rather a benchmark set for coherent calculations: misclassified or mixed members will cause a systematic deviation in the benchmark's storyline clock reference system, contaminating the peak delay and zero-hysteresis coherence of the entire group. The content out-of-step judgment and drift estimation stages will misinterpret the benchmark error as a member fault; therefore, dynamic grouping is a self-maintenance mechanism for the benchmark set, thus forming a closed loop of "measurement-grouping-remeasurement". In the above embodiment, peak coherence affinity drives grouping and writes back the group registration, ensuring the coherent benchmark set remains self-consistent during operation and guaranteeing the purity of the benchmark for the entire group's coherent measurements.

[0082] In one embodiment, the method further includes: performing dimensionless processing on the statistics of group coherence, the amount of step loss within the group, and the drift rate, and weighting the processing results to synthesize a group health score; the group coherence is the mean of zero-hysteresis coherence between each pair of participants within the group; when the group health score is lower than the warning threshold, outputting warning information for the group; when it is confirmed that a participant has content step loss, sending a resynchronization command to the participant's corresponding experience terminal; the resynchronization command carries the target content time, which is the sum of the current content time and the amount of step loss.

[0083] Combination Figure 6 , Figure 6 This is a schematic diagram illustrating the linkage between group health monitoring and broadcast control in one embodiment. Specifically, group health is calculated using the following formula:

[0084]

[0085] in, Group coherence (the mean of zero-hysteresis coherence between each pair of participants within a group, or the full-weighted group coherence as defined in the above embodiments). The statistical measure of the step loss within a group is, in one embodiment, the mean of the absolute values ​​of the step loss of each member; The drift rate is a statistical measure, and in one embodiment, the mean of the drift rates of each member is taken; Δmax=2s is the cutoff criterion for the step loss, and εmax=1% is the cutoff criterion for the drift rate. , , As weight, satisfying The coherence of the group is dimensionless; in the out-of-step term, min( Δ k Δmax / Δmax is in seconds per second, dimensionless; in the drift term, min(ε) k,εmax / εmax is a dimensionless ratio, therefore... It is a dimensionless weighted sum, with a range of values ​​of 1. .

[0086] in, The significance of truncation is to limit the negative contribution of a single fault item to the health score to a maximum value of 1, preventing a single severe fault (such as a very large step loss) from causing the group health score to drift negatively without bounds, thereby preserving the additivity of multiple faults. When coherence decline, step loss and drift exist simultaneously in the group, the contribution of each of the three items is bounded and the health score can still distinguish the severity difference between "single fault" and "multiple faults occurring concurrently".

[0087] Furthermore, when the group's health... Below the warning threshold (In one embodiment, For example, when the value is 0.8, a warning message is output for that group, triggering a yellow warning in one embodiment.

[0088] The broadcast control linkage includes three processing channels: First, when it is confirmed that a user has lost synchronization with the content, a resynchronization command is sent to the user's corresponding user terminal 101; the resynchronization command carries the target content time, which is the sum of the current content time and the amount of synchronization loss, i.e. The sign direction is explained as follows: step loss. First, the system takes the user's lag as a positive factor; that is, the playback control system advances the user's content clock forward by the same amount as the terminal's content clock lags, ensuring that the terminal's content timeline is precisely aligned with the entire set, achieving accurate resynchronization without requiring a restart. Second, when user drift is confirmed, a soft calibration is performed according to the above-mentioned refeedback correction embodiment, and an anchor point inspection prompt is output to the site (because the accuracy of the common module depends on the UWB anchor point, frequent drift may indicate an abnormal anchor point status). Third, when the user does not respond to content events, a site handling prompt is output, allowing on-site operations personnel to intervene manually.

[0089] In the above embodiments, the group health status closes the coherent measurement into a consumable processing action: the step loss is the only parameter source of the resynchronization command. Without quantification of the step loss, the broadcast control can only restart the entire experience; the synthesis and presentation of the group health status enables on-site operators to grasp the overall group operation status with a single scalar and process it separately according to the three types of channels.

[0090] In one embodiment, the server 102 registers an event table in the content event script. Each event in the event table must record at least the following fields: event time. The unit of measurement is seconds (s), that is, the first... The time when a content event occurs on the content timeline; event type This is used to mark the emphasis category of the content event; event weight. Dimensionless, used for weighted synthesis of overall coherence; response window The typical length is 5 seconds, which is the time interval during which the user reacts to the content event, i.e., the event response window; [This refers to] the target point. The unit of measurement is meters (m), and it is only valid for set-type events; expected rate. The unit of measurement is meters per second (m / s), which is the target rate at which the experiencer responds to movement as specified in the content script.

[0091] Among them, event types The events are categorized into four types: guidance, avoidance, fright, and assembly. Guiding events, through the appearance of a guiding light band, define the participant's path and starting point, causing them to accelerate along the designated path. Avoidance events, through the direction of a monster's attack, define the participant's avoidance speed vector, causing them to move quickly away from the attack direction. Fright events, through sudden stimuli, mandate abrupt stops or turns for the participant. Assembly events, through the appearance of an assembly circle, mandate that the participant converge at a designated coordinate. The common thread among these four types of events is that the content script imposes hard constraints on the participant's walking-level movements—in a large, freely navigable space, without displacement, the plot cannot progress; movement itself is the interaction channel. Therefore, the kinematic sequence of group members within the event response window inevitably exhibits calculable cross-individual coherence.

[0092] Furthermore, each content event is assigned a corresponding event weight according to its event type. The event weights for content events during free exploration periods are zero. This is because the movement of participants during free exploration periods is not subject to hard constraints from the plot, and its cross-individual coherence is close to random. Including it in coherent synthesis would dilute coherence and submerge correlation peaks. Therefore, this scheme only uses the periods of "script-forced synchronization" for coherence measurement, without relying on the full synchronization of participants. This is a substantial relaxation of the stronger assumption that "movements within the same group must be synchronized."

[0093] Specifically, based on the event weight of each content event. The coherence calculation results corresponding to each event response window are weighted: the coherence of the full-process weighted group.

[0094]

[0095] That is, the group coherence corresponding to each event response window is weighted and averaged according to the event weight, and the free exploration period is due to Automatically excluded from synthesis. In the above embodiments, through structured registration of the event table and weighting by event type, coherence measurement only utilizes the "script-forced synchronization" period, avoiding the dilution of coherence synthesis during the free exploration period, and not imposing a stronger assumption of full synchronization on the experiencer.

[0096] In one embodiment, the method further includes: presenting group health, coherence distribution among participants within a group, step loss and drift rate for each participant, by group.

[0097] Specifically, the monitoring and control terminal 103 displays the group health status of each group. The coherence distribution among participants within a group (presented as a coherence heatmap in one embodiment), and the step loss of each participant. and drift rate In the above embodiments, on-site operators can grasp the overall operational status of the group with a single scalar, and locate specific abnormal members from three dimensions: coherence distribution, step loss, and drift rate, thereby improving the operability of monitoring.

[0098] Based on the same inventive concept, this application also provides a collaborative group monitoring device for implementing the collaborative group monitoring method based on real-time positioning involved in the above method embodiments. The solution provided by this device is similar to the implementation solution described in the above method. Therefore, the specific limitations of one or more collaborative group monitoring device embodiments provided below can be found in the limitations of the collaborative group monitoring method based on real-time positioning above, and will not be repeated here.

[0099] In one embodiment, such as Figure 7 As shown, Figure 7 This is a structural block diagram of a collaborative group monitoring device based on real-time positioning in one embodiment. A collaborative group monitoring device 700 based on real-time positioning is provided, including: an acquisition module 701, a sequence construction module 702, a coherence calculation module 703, and a loss-of-step detection module 704, wherein: The acquisition module 701 is used to acquire the real-time location streams reported by each experience terminal within the same group, as well as the content event scripts; wherein, the content event scripts record the event times of content events used to guide the movement of experiencers within the group; The sequence construction module 702 is used to determine the event response window corresponding to each content event based on the event time, and to construct the kinematic feature sequence of each experiencer within the event response window based on the real-time positioning stream. The coherence calculation module 703 is used to perform cross-individual cross-correlation calculations on the kinematic feature sequences of different participants within a group to obtain intra-group coherence metrics; wherein, the intra-group coherence metrics include at least peak delay and zero-hysteresis coherence; The out-of-sync detection module 704 is used to determine the experiencer and the corresponding out-of-sync amount within the group based on the peak latency and the degree of collapse of the zero-hysteresis coherent relative group baseline.

[0100] The modules in the aforementioned real-time positioning-based collaborative group monitoring device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0101] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in any of the above method embodiments.

[0102] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in any of the above method embodiments.

[0103] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in any of the above method embodiments.

[0104] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0105] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0106] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A collaborative group monitoring method based on real-time positioning, characterized in that, The method includes: The system acquires real-time location streams reported by each experience terminal within the same group, as well as content event scripts; wherein, the content event scripts record the event times of content events used to guide the movement of experiencers within the group. The event response window corresponding to each content event is determined based on the event time, and the kinematic feature sequence of each experiencer within the event response window is constructed based on the real-time positioning stream. Cross-individual cross-correlation is performed on the kinematic feature sequences of different participants within the group to obtain intra-group coherence metrics; wherein, the intra-group coherence metrics include at least peak delay and zero hysteresis coherence; Based on the peak latency and the degree of collapse of the zero-hysteresis coherent relative group baseline, determine the users in the group who have content out-of-sync and the corresponding out-of-sync amount.

2. The method according to claim 1, characterized in that, The step of performing cross-individual cross-correlation calculations on the kinematic feature sequences of different participants within the group to obtain intra-group coherence metrics includes: For each time delay, the overlapping segment of the kinematic feature sequences of the two participants within the event response window is taken, and the mean is removed and the energy is calculated only on the overlapping segment. The product of the two kinematic feature sequences within the overlapping segment is normalized to obtain the normalized cross-correlation under that time delay. In the normalized cross-correlation under each of the aforementioned time delays, the maximum value is taken as the peak coherence, the time delay corresponding to the maximum value is taken as the peak time delay, and the normalized cross-correlation when the time delay is zero is taken as the zero-hysteresis coherence.

3. The method according to claim 1, characterized in that, The step of determining the users within the group who are experiencing content synchronization loss and the corresponding loss amount based on the peak latency and the degree of collapse of the zero-hysteresis coherent relative group baseline includes: For each participant, the median of the peak delays of the other participants in the group pointing to that participant is taken as the participant's step loss. The member coherence of the participant is determined based on the zero-hysteresis coherence between the participant and the other participants in the group. When the absolute value of the out-of-step amount is greater than the out-of-step threshold, the collapse amount of the member coherence relative to the group baseline is greater than the collapse threshold, and at least two consecutive event response windows meet the above conditions, it is determined that the user has content out-of-step.

4. The method according to claim 1, characterized in that, Before determining the users within the group who have experienced content synchronization loss and the corresponding amount of synchronization loss, the method further includes: Calculate the speed energy and pause percentage for each participant within the event response window; When a participant's speed energy is below the energy threshold or the percentage of pauses is above the percentage threshold, it is determined that the participant is not responding to the content event and stops participating in the content out-of-step judgment. When the speed energy of the user is not lower than the energy threshold and the percentage of pauses is not higher than the percentage threshold, the user is determined to be responsive, and the content out-of-step judgment is performed on the responsive user.

5. The method according to claim 1, characterized in that, The method further includes: For content events involving collections, the positions of each participant at the time of the event are obtained, and the positional offset of each participant relative to the group baseline is calculated; wherein, the group baseline is the median of the positions of the remaining participants within the group; The positional offset of each participant is regressed along the cumulative distance traveled by that participant to obtain the drift rate of that participant; The drift vector is determined based on the drift rate and cumulative travel, and the real-time positioning flow is re-flushed and corrected according to the drift vector.

6. The method according to claim 5, characterized in that, The step of regressing the positional offset of each participant along their cumulative travel distance to obtain the participant's drift rate includes: The position offset is decomposed along two orthogonal coordinate axes, and the two orthogonal components obtained from the decomposition are regressed along the cumulative travel to obtain the slope vector; Before performing the re-feedback correction on the real-time positioning stream according to the drift vector, the method further includes: Perform a significance test on the slope vector; When the slope vector passes the significance test, the drift rate is greater than the drift rate threshold, and the user's step loss is zero within the error range, it is determined that the user has drift, and the refeeding correction is performed on the user.

7. The method according to claim 6, characterized in that, The step of re-feeding back the real-time positioning stream according to the drift vector includes: The corrected confidence level of each participant is determined based on the member coherence of each participant; wherein, the member coherence is the mean of the zero-hysteresis coherence between the participant and the other participants in the group; When the corrected confidence level of the participant is lower than the confidence threshold, the refeedback correction for the participant is frozen, and the participant is excluded from the calculation of the group baseline. When the corrected confidence level of the user is not lower than the confidence level threshold, the user's real-time positioning stream is corrected by multiplying the correction coefficient by the drift vector.

8. The method according to claim 7, characterized in that, The method further includes: The group coherence, the statistics of step loss within the group, and the statistics of drift rate are all dimensionless, and the processing results are weighted and synthesized into a group health score; wherein, the group coherence is the mean of zero-hysteresis coherence between each pair of participants within the group; When the health status of the group is lower than the warning threshold, a warning message is output for the group. When it is confirmed that a user has lost synchronization with the content, a resynchronization command is sent to the user's corresponding terminal.

9. The method according to claim 1, characterized in that, The method further includes: For each participant, the peak coherence between the participant and participants in each group is calculated, and the peak coherence is averaged using a rolling window to obtain the participant's group affinity for each group. When the difference between the experiencer's group affinity to another group and their group affinity to their current group is greater than the group adjustment threshold and continues for a preset number of windows, the experiencer will be reassigned to the other group. Update the group registration information and use the updated group as the coherence benchmark for the subsequent cross-individual cross-correlation calculation.

10. The method according to claim 1, characterized in that, The content event script includes an event table, and each content event in the event table records at least: event time, event type, event weight, event response window, set target point, and expected rate; The event types include four categories of emphasis events: guidance, avoidance, fright, and assembly. Each content event is assigned a corresponding event weight according to its event type, and the event weight of the content events corresponding to the free exploration period is zero.

Images