Industrial vehicle node data entry method based on human-computer interaction

Abstract

The invention discloses an industrial vehicle node data entry method based on man-machine interaction, and relates to the technical field of automation control, and the method comprises the steps: obtaining the pose data of an industrial vehicle, building a track recording session, forming a track sequence, and recording a bookmark sequence; track and bookmark data are solidified, and time axis and canvas linkage playback is achieved on the same interface; generating candidate points based on the tracks and bookmarks, calculating candidate scores, ranking and presenting the candidate scores in a candidate list, and clicking the candidate points to drive a time axis to jump and position on a canvas; aiming at the positioning candidate points, determining node types, inputting and generating nodes according to minimum necessary fields, enabling node records to be adsorbed to corresponding moments and poses, and writing the node records into a node sequence; track subsequences between adjacent nodes are intercepted according to node moments, track segment constraint fitting is carried out on edges to generate curve parameters, error verification is carried out, if the curve parameters do not pass, rollback and remarking are carried out, and after the error verification, a road network is published and loaded. And the road network quality and the deployment stability are improved through trajectory segment constraint fitting and verification of a rollback closed loop.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and in particular to a method for inputting node data of industrial vehicles based on human-machine interaction. Background Technology

[0002] In recent years, industrial vehicles have been rapidly adopted in warehousing and logistics, flexible manufacturing, and automated delivery scenarios, with vehicles gradually shifting from traditional manual driving to a hybrid mode of "manned collaboration and unmanned autonomy." A key emerging need is to quickly transform the routes, workstations, and task constraints generated during on-site operations into computable road network data to support scheduling, route planning, and operation traceability. Existing engineering practices typically combine vehicle positioning, map playback, and graphical annotation tools, using human-machine interaction to input nodes and edges, which are then further used for road network deployment and version iteration, thereby shortening deployment cycles and reducing operation and maintenance costs.

[0003] However, existing methods still have two limitations: First, most solutions focus on "online and real-time" annotation, requiring operators to simultaneously select nodes and fill in fields while driving or working, which is susceptible to communication fluctuations, positioning jitter, and operational distraction, making it difficult to strictly align node time and pose with the actual trajectory; Second, the generation of road network edges often adopts direct connection between nodes, manual drawing, or weak constraint smoothing, lacking trajectory segment constraint fitting and error verification closed loop based on node time, resulting in deviation between edge geometry and actual driving trajectory, and instability of turning radius and path feasibility. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a human-computer interaction-based method for inputting industrial vehicle node data to solve the problems of insufficient alignment of node input at the pose level at the actual trajectory moment and lack of trajectory constraint error closed-loop verification in edge generation, which leads to road network deviation from reality and difficulty in stable publishing.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for recording node data of industrial vehicles based on human-computer interaction, which includes: acquiring industrial vehicle pose data through real-time communication, establishing a trajectory recording session, starting recording to generate a trajectory sequence, and receiving bookmarks triggered by the operator to generate a bookmark sequence.

[0008] Solidify the trajectory sequence and bookmark sequence, and establish a timeline and canvas linkage playback on the same interface;

[0009] A candidate point set is generated based on the trajectory sequence and bookmark sequence. The candidate scores are calculated, and the candidate points are sorted according to the candidate scores and presented in the candidate list. Clicking on a candidate point drives the timeline to jump and locates the candidate point on the canvas.

[0010] For the candidate points to be located, determine the node type and complete the input according to the minimum required field corresponding to the node type, then confirm the generation of nodes, so that the generated node records the time and pose corresponding to the candidate points and write them into the node sequence.

[0011] Based on the node sequence, the trajectory subsequence between adjacent nodes is extracted in the order of node time. The trajectory subsequence is used to constrain and fit the edges formed by adjacent nodes, generate the curve parameters of the edges, and perform error verification. When all edges pass the verification, the road network is published and loaded.

[0012] As a preferred embodiment of the human-computer interaction-based industrial vehicle node data input method of the present invention, the specific steps for acquiring industrial vehicle pose data through real-time communication are as follows:

[0013] Collect vehicle pose data frames and continuously update the latest frames to calculate the communication availability score;

[0014] Recording can begin when the communication availability score is greater than or equal to the minimum score threshold. Recording will not be allowed and a communication error will be indicated when the communication availability score is less than the minimum score threshold.

[0015] As a preferred embodiment of the human-computer interaction-based industrial vehicle node data entry method of the present invention, the steps of establishing a trajectory recording session, initiating recording to form a trajectory sequence, and receiving bookmarks triggered by the operator to form a bookmark sequence are as follows:

[0016] Establish a trajectory recording session, perform frame taking once at each sampling time, and select the frame with the largest timestamp among all pose frames in the pose buffer as the current sampling frame;

[0017] Write the position and heading in the current sampling frame into the trajectory sequence. Perform a consistency check on adjacent sampling points. If the check passes, write the trajectory point; otherwise, discard it and wait for the next sampling period to continue writing.

[0018] Find the track point index that is closest to the bookmark trigger timestamp in the written track sequence, bind the bookmark to the track point, and obtain the bookmark sequence.

[0019] As a preferred embodiment of the human-computer interaction-based industrial vehicle node data entry method of the present invention, the specific steps for solidifying the trajectory sequence and bookmark sequence, and establishing a timeline and canvas-linked playback on the same interface are as follows.

[0020] Solidify the trajectory sequence and bookmark sequence, and read the start and end times of the playback timeline from the trajectory point sequence;

[0021] Replay the canvas to draw the complete trajectory line and all bookmark icons, and display the vehicle position indicator at the coordinates of the first trajectory point in the trajectory point sequence;

[0022] The playback frame index is obtained by using a frame retrieval function. Each bookmark is then judged based on the bookmark highlight half-width threshold to determine whether it should be highlighted.

[0023] As a preferred embodiment of the human-computer interaction-based industrial vehicle node data entry method of the present invention, the steps for generating a candidate point set based on trajectory sequences and bookmark sequences, and calculating candidate scores, are as follows:

[0024] Based on the trajectory sequence, the discrete velocity and discrete heading rate of change between adjacent trajectory points are calculated, and a fixed-length moving average smoothing is performed to obtain the smoothed velocity sequence and the smoothed heading rate of change sequence.

[0025] A set of candidate evaluation times is generated with a fixed step size. A bookmark indicator is established. For each evaluation time, a time window integration is performed and the bookmark indicator is superimposed to obtain the evidence value. The evidence value is then used to obtain the candidate score through a bounded mapping.

[0026] As a preferred embodiment of the human-computer interaction-based industrial vehicle node data entry method of the present invention, the following steps are taken: Candidate points are sorted according to their scores and presented in a candidate list; clicking on a candidate point drives a jump along the timeline and positions the candidate point on the canvas.

[0027] Sort the candidate points in descending order of their candidate scores and iterate through them sequentially. If the planar distance between the current candidate point and any candidate point in the retained candidate point set is not greater than the distance threshold, discard the current candidate point; otherwise, add it to the retained set to obtain the final candidate point set.

[0028] The final set of candidate points is sorted from high to low according to the candidate scores to generate a candidate list. The canvas is synchronized to mark the positions of all candidate points with a uniform icon, and the candidate point with the highest candidate score is selected and its position is highlighted.

[0029] When the candidate point with the highest candidate score is selected, the current playback time on the timeline is set as the candidate point time, and the vehicle position indicator is positioned at the coordinates of the trajectory point whose timestamp is no later than the candidate point time and has the largest timestamp.

[0030] As a preferred embodiment of the human-computer interaction-based industrial vehicle node data entry method of the present invention, the steps are as follows: for the candidate points to be located, the node type is determined, and after completing the input according to the minimum required fields corresponding to the node type, the node is confirmed to be generated. The generated node records the time and pose corresponding to the candidate point and are written into the node sequence.

[0031] After selecting a candidate point, a list of node type options is displayed. When the current candidate point has a bookmark, a business type node is selected; when the current candidate point does not have a bookmark, a path point is selected.

[0032] Based on the determined node type, the minimum set of required fields corresponding to the type is deterministically activated, and the node is only allowed to be generated and the node record is obtained when all required fields pass the validation.

[0033] An adsorption rule is applied to the node records to label the current candidate points;

[0034] The adsorption rules include fixing the timestamp of the node to the candidate time of the current candidate point, and fixing the coordinates and heading of the node to the coordinates and heading of the playback frame trajectory corresponding to the candidate time.

[0035] As a preferred embodiment of the human-computer interaction-based industrial vehicle node data entry method of the present invention, the minimum set of required fields corresponding to the activation type is defined in the following steps.

[0036] When the node type is path point, the required field is remarks;

[0037] When the node type is pickup point and drop-off point, the required fields are remarks and workstation number;

[0038] When the node type is standby point or standby receiving point, the required fields are remarks and orientation lock mark;

[0039] When the node type is a delivery point, the required fields are remarks, workstation number, and orientation lock mark.

[0040] As a preferred embodiment of the human-computer interaction-based industrial vehicle node data entry method of the present invention, the method involves: extracting trajectory subsequences between adjacent nodes according to the node time sequence, and using the trajectory subsequences to constrain and fit the edges formed by adjacent nodes to generate curve parameters for the edges. The specific steps are as follows.

[0041] In the trajectory point sequence, all trajectory points with timestamps no earlier than the start of the interval and no later than the end of the interval are selected and arranged in chronological order to form a trajectory subsequence. When the number of trajectory points in the trajectory subsequence is less than the minimum number of points required for fitting, the edges corresponding to adjacent node pairs are marked as edges with insufficient data.

[0042] For each pair of adjacent nodes that is not marked as an edge with insufficient data, the start and end coordinates of the adjacent nodes are used as endpoint constraints of the edge curve.

[0043] Each trajectory point in the corresponding trajectory subsequence is used as a fitting sample point. A third-order Bézier curve is used as the geometric representation of the edge. The endpoints of the curve are fixed as the coordinates of two nodes. The curve parameters are obtained by minimizing the objective function.

[0044] As a preferred embodiment of the human-computer interaction-based industrial vehicle node data entry method of the present invention, the following steps are taken: Error verification is performed, and road network data is generated when all edges pass the verification.

[0045] For each edge with generated curve parameters, calculate the root mean square fitting error on the sample points of the corresponding trajectory subsequence.

[0046] An edge is considered a valid edge if and only if the root mean square fitting error is not greater than the fitting error threshold; an edge is considered a problematic edge if the root mean square fitting error is greater than the fitting error threshold.

[0047] When all edges pass the verification, the ordered node sequence and the set of verified edges are combined to generate road network data.

[0048] The beneficial effects of this invention are as follows: By using communication availability gating to initiate recording and solidify trajectory and bookmark data, stable, reliable, and traceable data collection is achieved; by generating candidate points from trajectories and bookmarks and using time window integration to superimpose bookmark intent candidate scores for ranking, key locations can be quickly focused on and the amount of annotation operations reduced; by extracting trajectory segments by node time and performing trajectory segment constraint fitting on edges, and backtracking and re-marking if error verification fails until it passes before publishing the road network, the road network is consistent with actual driving and achieves closed-loop convergence, improving road network quality and deployment stability. Attached Figure Description

[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart of a method for inputting node data of industrial vehicles based on human-computer interaction.

[0051] Figure 2 A flowchart for trajectory recording and bookmark generation.

[0052] Figure 3 The flowchart for initializing the linked playback.

[0053] Figure 4 A flowchart for entering candidate points and nodes. Detailed Implementation

[0054] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0057] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for inputting industrial vehicle node data based on human-computer interaction, including the following steps:

[0058] S1. Acquire industrial vehicle pose data through real-time communication, establish a trajectory recording session, start recording to generate a trajectory sequence, and receive bookmarks triggered by the operator to generate a bookmark sequence.

[0059] After establishing a WebSocket connection, vehicle pose data frames are collected and continuously updated with the latest frames. To prevent collection failures due to continuous frame drops despite a established connection, a communication availability gating is set. Recording is only allowed to start when communication remains available within a time window. The expression is:

[0060] ;

[0061] in, Indicates communication availability score, Indicates the length of the integration time window. The timestamp indicating the time of the determination. Represents the timestamp of the integral independent variable. This indicates that the received instruction value has been received.

[0062] It should be noted that if in [ If at least one frame of pose is received within the specified time, the receive indication value is set to 1; otherwise, the receive indication value is set to 0. This indicates the receive timeout threshold, set to a fixed number of seconds (e.g., 0.2 seconds).

[0063] when Recording can begin when the score is greater than or equal to the minimum score threshold. If the score is below the minimum score threshold, recording will be disabled and a communication error will be displayed.

[0064] It should be noted that the minimum score threshold is obtained by continuously running the recording function for no less than 30 minutes and calculating a set of communication availability score samples at a frequency of 1Hz. After sorting the samples from smallest to largest, the value corresponding to the 5th percentile is taken as the minimum score threshold.

[0065] Establish a trajectory recording session, perform frame taking once at each sampling moment, select the frame with the largest timestamp among all pose frames in the pose buffer as the current sampling frame, obtain the current position and heading, write the current position and heading into the trajectory sequence, perform consistency checks on adjacent sampling points, including whether the time interval is abnormal, whether the speed calculated from the displacement of adjacent points exceeds the vehicle configuration limit, and whether the angular velocity calculated from the change of adjacent headings exceeds the vehicle's steering capability limit; if the check passes, write the trajectory point; if it fails, discard it and keep it waiting for the next sampling period to continue writing.

[0066] Find the track point index in the written track sequence that is closest to the bookmark trigger timestamp, and bind the bookmark to the track point. The expression is:

[0067] ;

[0068] in, Indicates the first Second bookmark trigger timestamp Indicates and index of the nearest trajectory point Indicates the first Each trajectory point timestamp.

[0069] The x-coordinate of the bound trajectory point, the y-coordinate of the bound trajectory point, the heading angle of the bound trajectory point, the trajectory sequence, and... The index of the closest trajectory point and the first The bookmark trigger timestamps are combined to obtain the bookmark records. All bookmark records are combined to obtain the bookmark sequence.

[0070] When recording ends, retrieve the trajectory sequence and bookmark sequence, and set the session state to recording complete.

[0071] S2. Solidify the trajectory sequence and bookmark sequence, and establish a timeline and canvas linkage playback in the same interface.

[0072] Upon completion of recording, immediately stop adding new trajectory points to the trajectory sequence and bookmark points to the bookmark sequence, and solidify the trajectory and bookmark sequences. Read the timestamp of the first trajectory point from the trajectory point sequence as the start time of the playback timeline, and read the timestamp of the last trajectory point from the trajectory point sequence as the end time of the playback timeline. Fix the draggable range of the timeline to the start time to the end time. Assign a sequential number to each trajectory point in the trajectory point sequence according to its order in the trajectory sequence.

[0073] The same interface displays a timeline control and a playback canvas. Specifically, the minimum time of the timeline control is fixed at the timestamp of the first trajectory point in the trajectory point sequence, and the maximum time of the timeline control is fixed at the timestamp of the last trajectory point in the trajectory point sequence. The current playback time is displayed on the timeline control. The playback canvas first draws a complete trajectory line, which is obtained by connecting the trajectory points in the trajectory point sequence in order. The playback canvas then draws all bookmark icons, where the position of each bookmark icon is fixed at the coordinate position of the trajectory point to which the bookmark is bound. After the drawing is completed, the current playback time is determined as the minimum time of the timeline control, and the vehicle position indicator is definitively displayed at the coordinate position of the first trajectory point in the trajectory point sequence, thus completing the initialization of the playback annotation interface.

[0074] The playback frame index is obtained using a deterministic frame-fetching function, expressed as follows:

[0075] ;

[0076] It should be noted that, This indicates the current moment on the timeline. Indicates the timeline index. Indicates to make The index that yields the maximum value .

[0077] Whether to highlight each bookmark is determined by the following criteria, expressed as:

[0078] ;

[0079] in, This indicates the half-width threshold for bookmark highlighting. Indicates the first An indicator of whether a bookmark is highlighted at the current playback moment.

[0080] It should be noted that, The value can be 0 or 1, where 1 indicates highlighting and 0 indicates no highlighting; The method involves collecting bookmark trigger samples for at least 30 minutes, calculating the absolute time difference between the trigger time of each bookmark and the timestamp of the bound trajectory point, forming a set of time difference samples, sorting them from smallest to largest, and taking the 95th percentile value of the time difference sample as the half-width threshold for bookmark highlighting.

[0081] When the operator drags the timeline, causing a change in the current playback time, the playback frame index is immediately updated and the vehicle indicator is located, and then the bookmark highlighting state is updated. When the operator triggers playback, the current time is incremented by a fixed time step (e.g., 0.05 seconds each time), and after each increment, the playback frame index is updated and the vehicle indicator is located, and then the bookmark highlighting state is updated, until the current playback time reaches the playback end time, at which point playback stops and the final positioning state is maintained.

[0082] S3. Generate a set of candidate points based on the trajectory sequence and bookmark sequence, calculate the candidate scores, sort the candidate points according to the candidate scores and present them in the candidate list, and drive the timeline to jump by clicking on the candidate points and locate the candidate points on the canvas.

[0083] Based on the trajectory sequence, the discrete velocity and discrete heading rate of change between each adjacent trajectory point are calculated in chronological order. The discrete velocity and discrete heading rate of change are then smoothed by a fixed-length moving average to obtain a smoothed velocity sequence and a smoothed heading rate of change sequence.

[0084] Starting from the timeline start time, a set of candidate evaluation times is generated with a fixed step size. An evaluation time is generated every fixed time interval (e.g., 0.20 seconds) until the timeline end time is reached. For each evaluation time, a bookmark indicator is established. The bookmark indicator is set to 1 if and only if there is a bookmark trigger time that falls within a closed interval of a fixed time interval (e.g., 0.30 seconds) before or after the evaluation time. Otherwise, it is set to 0.

[0085] For each evaluation moment, a fixed duration (e.g., 1.5 seconds) is taken before and after the evaluation moment as an integration window. Time window integration is performed, and the bookmark indicator is superimposed to obtain the evidence value. The evidence value is then mapped to obtain the candidate score, expressed as:

[0086] ;

[0087] ;

[0088] in, Indicates the evidence value. A timestamp representing a specific evaluation moment. Indicates the half-width of the integration window. Indicates time The corresponding smoothed rate of change of heading, This represents the normalized scaling parameter for the rate of change of heading. Indicates time The corresponding smoothing speed, This represents the velocity normalization scaling parameter. Indicates the number of bookmarks. This represents the bookmark gain coefficient. This indicates the candidate score.

[0089] It should be noted that, The method involves continuously collecting trajectory data for at least 30 minutes, first calculating the rate of change of heading for each sampling interval and smoothing it with a fixed window to obtain a set of smoothed rate of change of heading samples, then sorting the smoothed rate of change of heading samples in ascending order, and taking the value corresponding to the 50th percentile as the normalization scale parameter of the rate of change of heading. The process involves continuously collecting trajectory data for at least 30 minutes, first calculating the velocity of adjacent trajectory points and smoothing it with a fixed window to obtain a set of smoothed velocity samples, then sorting the smoothed velocity samples from smallest to largest, and taking the velocity value corresponding to the 50th percentile as the velocity normalization scale parameter. Each candidate point is divided into a bookmark group and a non-bookmark group based on whether it falls within the interval of 0.30 seconds before and after the bookmark trigger time. Then, with the heading change rate normalization scale parameter and the velocity normalization scale parameter fixed in advance, the bookmark gain coefficient is calculated by taking values ​​of 0.05 in the interval [0.00, 1.00]. The difference between the median score of the candidate points in the bookmark group and the median score of the candidate points in the non-bookmark group is used as the evaluation index. The value that maximizes the evaluation index and at the same time satisfies that the 5th percentile of the bookmark group score is not less than 0.65 is selected as the bookmark gain coefficient.

[0090] In the evaluation time sequence arranged chronologically, candidate points are extracted using deterministic rules. Specifically, when the candidate score of a certain evaluation time simultaneously satisfies the following conditions: not less than the candidate score of the previous evaluation time, strictly greater than the candidate score of the next evaluation time, and not less than the lower limit threshold of the candidate score, the evaluation time is confirmed as a candidate point time. The spatial location of the candidate point is fixed by taking the coordinates of the playback frame trajectory point corresponding to the candidate point time, and the candidate score and bookmark indicator of the candidate point are written into the candidate point record.

[0091] It should be noted that the lower limit threshold is determined by marking each time point that should be a candidate point in the playback interface as a control set. Then, the scores of all peak times are used as candidate thresholds in ascending order, and the proportion of hits in the control set and the proportion of false hits in the non-control times are calculated. Finally, the candidate threshold that minimizes the proportion of false hits in the non-control times is selected as the candidate score lower limit threshold under the constraint that the proportion of hits in the control set is not lower than the predetermined proportion value. The predetermined proportion value is obtained by performing a complete annotation on all samples, counting the proportion of hits in the control set in each sample, obtaining a set of proportion samples, sorting them in ascending order, and taking the proportion corresponding to the 5th percentile of the proportion samples as the predetermined proportion value.

[0092] Sort the candidate points in the candidate point record from highest to lowest according to their candidate scores, and iterate through them sequentially. If the planar distance between the current candidate point and any candidate point in the already retained candidate point set is not greater than the distance threshold, discard the current candidate point; otherwise, add it to the retained set. After the iteration is completed, the retained set is the final candidate point set.

[0093] It should be noted that the distance threshold is achieved by generating initial candidate points on multiple typical recording samples, calculating the planar distance between each candidate point and sorting them in ascending order, and using candidate point pairs that are indistinguishable to the naked eye on the playback canvas and produce the same node result after labeling as a control set. The 95th percentile of the distance distribution corresponding to the control set is taken as the distance threshold.

[0094] The final set of candidate points is sorted from highest to lowest according to the candidate scores to generate a candidate list. The candidate list includes the candidate point number, candidate score, whether it is marked with a bookmark, and the candidate point time. The canvas is synchronized to mark the positions of all candidate points with a uniform icon, and the first candidate with the highest score is selected by default and its position is highlighted.

[0095] When the operator clicks on a candidate in the candidate list, the current playback time of the timeline is immediately set to the candidate point time recorded by the candidate; then the vehicle position indicator is positioned to the coordinates of the trajectory point whose timestamp is no later than the candidate point time and has the largest timestamp, and the trajectory point and candidate point icons are highlighted at the same time.

[0096] S4. For the candidate points to be located, determine the node type and complete the input according to the minimum required field corresponding to the node type. Then confirm the generation of nodes, so that the generated node records the time and pose corresponding to the candidate points and write them into the node sequence.

[0097] After clicking on a candidate point in the candidate list, it is set as the current annotation object, and the candidate point position is highlighted on the playback canvas. At the same time, the current playback time on the timeline is immediately set as the candidate time recorded by the candidate point. Then, the candidate time, candidate position coordinates, candidate heading, candidate score, and whether it is marked with a bookmark are displayed in read-only mode in the annotation panel. The operator clicks "Start Annotation" in the annotation panel to lock the current annotation object.

[0098] The label panel displays a list of node type options, including path point, pickup point, drop-off point, standby point, standby receiving point, and drop-off and warehousing point. When the current candidate point is marked with a bookmark, select the business type node (which points to the pickup point by default); when the current candidate point is not marked with a bookmark, select the path point.

[0099] Based on the determined node type, the minimum set of required fields corresponding to the type is deterministically enabled, and the node generation is only allowed when all required fields pass validation. The minimum set of required fields is deterministically mapped using the following steps A1-A4:

[0100] A1. When the node type is a path point, the required field is Remarks.

[0101] A2. When the node type is pickup point and drop-off point, the required fields are remarks and workstation number.

[0102] A3. When the node type is standby point or standby receiving point, the required fields are remarks and orientation lock mark.

[0103] A4. When the node type is a delivery point, the required fields are remarks, workstation number, and orientation lock mark.

[0104] After clicking "Confirm Node Generation," a node record is generated, and a deterministic adsorption rule is adopted. Specifically, the node's timestamp is fixed at the candidate time of the current candidate point, and the node's coordinates and heading are fixed at the coordinates and heading of the playback frame trajectory corresponding to the candidate time. A sequential number is assigned to the node, and the node record is appended to the node sequence in the order of generation. At the same time, a node icon is drawn on the canvas, and the current candidate point is marked as labeled. After traversing all unlabeled candidate points in the candidate list, a deterministic check is performed. Specifically, the number of nodes is not less than the preset minimum number, and the timestamps corresponding to the sequential numbering of nodes satisfy strict incrementing.

[0105] It should be noted that the preset minimum quantity is determined by selecting multiple recorded samples, marking the set of nodes necessary to complete a full scheduling loop on the playback interface, counting the number of nodes in the necessary set of nodes in each sample and forming a set of quantity samples, sorting them from smallest to largest, and taking the 5th percentile of the quantity sample as the preset minimum quantity.

[0106] S5. Based on the node sequence, extract the trajectory subsequence between adjacent nodes in the order of node time. Use the trajectory subsequence to constrain and fit the edges formed by adjacent nodes, generate the curve parameters of the edges, and perform error verification. When all edges pass the verification, publish and load the road network.

[0107] The node sequence is set to be non-appendable, and the nodes are sorted in ascending order of their timestamps to obtain an ordered node sequence. At the same time, the trajectory point sequence is read and set to be non-appendable. If any two adjacent nodes have the same timestamp, only the node with the higher candidate score is retained and the node with the lower candidate score is deleted to ensure that the node timestamps are strictly increasing.

[0108] For any two adjacent nodes in an ordered node sequence, take the timestamp of the preceding node as the start of the interval and the timestamp of the following node as the end of the interval; in the trajectory point sequence, select all trajectory points whose timestamps are no earlier than the start of the interval and no later than the end of the interval, and form a trajectory subsequence of adjacent node pairs according to their order in the trajectory point sequence; if the number of trajectory points in a trajectory subsequence is less than the minimum number of points required for fitting, then mark the edge corresponding to the adjacent node pair as an edge with insufficient data.

[0109] It should be noted that the minimum number of points required for fitting is determined by gradually increasing the number of trajectory points involved in fitting for each segment of the trajectory subsequence between adjacent nodes and repeating the fitting process. At the same time, the root mean square distance from the fitted curve to the points of this segment is calculated, and the minimum number of points at which the root mean square distance first converges is recorded as the minimum number of usable points for this segment. The minimum number of usable points for all sample segments is then summed, and the 95th percentile of the sum is taken as the minimum number of points required for fitting.

[0110] For each pair of adjacent nodes not marked as an edge with insufficient data, the start and end coordinates of the adjacent nodes are used as endpoint constraints of the edge curve. Each trajectory point in the corresponding trajectory subsequence is used as a fitting sample point. A third-order Bézier curve is used as the geometric representation of the edge. The curve endpoints are fixed at the coordinates of the two nodes. The remaining curve parameters are determined by minimizing the sum of the squared distances from the curve to the trajectory sample points. The minimization objective function is expressed as:

[0111] ;

[0112] in, Represents a third-order Bézier curve. , Indicates the control point to be determined. The first subsequence of the trajectory A vector of coordinates of sample points, This represents the parameter obtained by normalizing the cumulative arc length based on the sample points.

[0113] By minimizing the objective function, the control points are obtained as real coordinates, and the curve parameters are acquired.

[0114] For each edge with generated curve parameters, calculate the root mean square fitting error at the sample points of the corresponding trajectory subsequence. The expression is:

[0115] ;

[0116] in, This represents the root mean square fitting error.

[0117] If and only if When the error is not greater than the fitting error threshold, the edge is considered to have passed the verification. When the error exceeds the fitting error threshold, the edge is identified as a problem edge, and the time interval corresponding to the problem edge is recorded.

[0118] It should be noted that the fitting error threshold is determined by selecting multiple segments of trajectory subsequences between adjacent nodes, confirming a set of curves for which the fitting result of each segment is acceptable (i.e., the overall curve is consistent with the segment trajectory line and does not produce erroneous detours). For each segment in the set of references, the root mean square distance from the fitting curve to the segment trajectory point is calculated to obtain a set of error samples, which are then sorted in ascending order. The 95th percentile value of the error sample is taken as the fitting error threshold.

[0119] Generate a problem list by ordering all problem edges in the node sequence. After selecting a problem edge, the playback timeline is deterministically jumped to the starting point of the problem edge's time interval at the current playback time. The time interval is then highlighted on the timeline. Simultaneously, the canvas is positioned at the position corresponding to the start point of the interval, allowing the operator to reselect candidate points and regenerate adjacent nodes within the time interval (at least one of the two nodes connected by the original problem edge is replaced). After completion, the trajectory subsequence is re-extracted using the new node sequence and refitted for verification.

[0120] When all edges pass the verification, the ordered node sequence and the set of verified edges (including curve parameters) are combined to generate road network data. The road network is then set to the current available version. Each edge is drawn as a curve on the same interface and all node icons are displayed, thus completing the road network loading.

[0121] In summary, this invention achieves stable, reliable, and traceable data collection by: initiating recording and solidifying trajectory and bookmark data through communication availability gating; quickly focusing on key locations and significantly reducing the amount of annotation operations by generating candidate points from trajectories and bookmarks and presenting them using time window integration and superimposed bookmark intent candidate scores; and ensuring that the road network is consistent with actual driving conditions and achieves closed-loop convergence, thereby improving road network quality and deployment stability by: extracting trajectory segments by node time and performing trajectory segment constraint fitting on edges, and reverting to re-marking if error verification fails until it passes before publishing the road network.

[0122] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for inputting node data of industrial vehicles based on human-computer interaction, characterized in that: include, The system acquires industrial vehicle pose data through real-time communication, establishes a trajectory recording session, starts recording to generate a trajectory sequence, and receives bookmarks triggered by the operator to generate a bookmark sequence. Solidify the trajectory sequence and bookmark sequence, and establish a timeline and canvas linkage playback on the same interface; A candidate point set is generated based on the trajectory sequence and bookmark sequence. The candidate scores are calculated, and the candidate points are sorted according to the candidate scores and presented in the candidate list. Clicking on a candidate point drives the timeline to jump and locates the candidate point on the canvas. For the candidate points to be located, determine the node type and complete the input according to the minimum required field corresponding to the node type, then confirm the generation of nodes, so that the generated node records the time and pose corresponding to the candidate points and write them into the node sequence. Based on the node sequence, the trajectory subsequence between adjacent nodes is extracted in the order of node time. The trajectory subsequence is used to constrain and fit the edges formed by adjacent nodes, generate the curve parameters of the edges, and perform error verification. When all edges pass the verification, the road network is published and loaded.

2. The method for inputting industrial vehicle node data based on human-computer interaction as described in claim 1, characterized in that: The specific steps for acquiring industrial vehicle pose data through real-time communication are as follows: Collect vehicle pose data frames and continuously update the latest frames to calculate the communication availability score; Recording can begin when the communication availability score is greater than or equal to the minimum score threshold. Recording will not be allowed and a communication error will be indicated when the communication availability score is less than the minimum score threshold.

3. The method for inputting industrial vehicle node data based on human-computer interaction as described in claim 2, characterized in that: The steps for establishing a trajectory recording session, initiating recording to form a trajectory sequence, and receiving bookmarks triggered by the operator to form a bookmark sequence are as follows: Establish a trajectory recording session, perform frame taking once at each sampling time, and select the frame with the largest timestamp among all pose frames in the pose buffer as the current sampling frame; Write the position and heading in the current sampling frame into the trajectory sequence. Perform a consistency check on adjacent sampling points. If the check passes, write the trajectory point; otherwise, discard it and wait for the next sampling period to continue writing. Find the track point index that is closest to the bookmark trigger timestamp in the written track sequence, bind the bookmark to the track point, and obtain the bookmark sequence.

4. The method for inputting industrial vehicle node data based on human-computer interaction as described in claim 3, characterized in that: The specific steps for solidifying the trajectory sequence and bookmark sequence, and establishing a timeline and canvas-linked playback on the same interface are as follows: Solidify the trajectory sequence and bookmark sequence, and read the start and end times of the playback timeline from the trajectory point sequence; Replay the canvas to draw the complete trajectory line and all bookmark icons, and display the vehicle position indicator at the coordinates of the first trajectory point in the trajectory point sequence; The playback frame index is obtained by using a frame retrieval function. Each bookmark is then judged based on the bookmark highlight half-width threshold to determine whether it should be highlighted.

5. The method for inputting industrial vehicle node data based on human-computer interaction as described in claim 4, characterized in that: The specific steps for generating a candidate point set based on the trajectory sequence and bookmark sequence, and calculating the candidate score, are as follows: Based on the trajectory sequence, the discrete velocity and discrete heading rate of change between adjacent trajectory points are calculated, and a fixed-length moving average smoothing is performed to obtain the smoothed velocity sequence and the smoothed heading rate of change sequence. A set of candidate evaluation times is generated with a fixed step size. A bookmark indicator is established. For each evaluation time, a time window integral is performed and the bookmark indicator is superimposed to obtain the evidence value. The evidence value is then used to obtain the candidate score through a bounded mapping.

6. The method for inputting industrial vehicle node data based on human-computer interaction as described in claim 5, characterized in that: The candidate points are sorted by their scores and presented in a candidate list. Clicking on a candidate point drives a jump along the timeline and positions the candidate point on the canvas. The specific steps are as follows: Sort the candidate points in descending order of their candidate scores and iterate through them sequentially. If the planar distance between the current candidate point and any candidate point in the retained candidate point set is not greater than the distance threshold, discard the current candidate point; otherwise, add it to the retained set to obtain the final candidate point set. The final set of candidate points is sorted from high to low according to the candidate scores to generate a candidate list. The canvas is synchronized to mark the positions of all candidate points with a uniform icon, and the candidate point with the highest candidate score is selected and its position is highlighted. When the candidate point with the highest candidate score is selected, the current playback time on the timeline is set as the candidate point time, and the vehicle position indicator is positioned at the coordinates of the trajectory point whose timestamp is no later than the candidate point time and has the largest timestamp.

7. The method for inputting industrial vehicle node data based on human-computer interaction as described in claim 6, characterized in that: For the candidate points to be located, the node type is determined, and after completing the input according to the minimum required fields corresponding to the node type, the node generation is confirmed. The generated node records the time and pose corresponding to the candidate point and writes them into the node sequence. The specific steps are as follows: After selecting a candidate point, a list of node type options is displayed. When the current candidate point has a bookmark, a business type node is selected; when the current candidate point does not have a bookmark, a path point is selected. Based on the determined node type, the minimum set of required fields corresponding to the type is deterministically activated, and the node is only allowed to be generated and the node record is obtained when all required fields pass the validation. An adsorption rule is applied to the node records to label the current candidate points; The adsorption rules include fixing the timestamp of the node to the candidate time of the current candidate point, and fixing the coordinates and heading of the node to the coordinates and heading of the playback frame trajectory corresponding to the candidate time.

8. The method for inputting industrial vehicle node data based on human-computer interaction as described in claim 7, characterized in that: The minimum set of required fields corresponding to the activation type, specifically the following steps. When the node type is a path point, the required field is Remarks; When the node type is pickup point and drop-off point, the required fields are remarks and workstation number; When the node type is standby point or standby receiving point, the required fields are remarks and orientation lock mark; When the node type is a delivery point, the required fields are remarks, workstation number, and orientation lock mark.

9. The method for inputting industrial vehicle node data based on human-computer interaction as described in claim 8, characterized in that: The method involves extracting trajectory subsequences between adjacent nodes based on the node sequence in time order, and then using these trajectory subsequences to constrain and fit the edges formed by adjacent nodes to generate curve parameters for the edges. The specific steps are as follows: In the trajectory point sequence, all trajectory points with timestamps no earlier than the start of the interval and no later than the end of the interval are selected and arranged in chronological order to form a trajectory subsequence. When the number of trajectory points in the trajectory subsequence is less than the minimum number of points required for fitting, the edges corresponding to adjacent node pairs are marked as edges with insufficient data. For each pair of adjacent nodes that is not marked as an edge with insufficient data, the start and end coordinates of the adjacent nodes are used as endpoint constraints of the edge curve. Each trajectory point in the corresponding trajectory subsequence is used as a fitting sample point. A third-order Bézier curve is used as the geometric representation of the edge. The endpoints of the curve are fixed as the coordinates of two nodes. The curve parameters are obtained by minimizing the objective function.

10. The method for inputting industrial vehicle node data based on human-computer interaction as described in claim 9, characterized in that: The process involves performing error verification. When all edges pass the verification, road network data is generated. The specific steps are as follows: For each edge with generated curve parameters, calculate the root mean square fitting error on the sample points of the corresponding trajectory subsequence. An edge is considered a valid edge if and only if the root mean square fitting error is not greater than the fitting error threshold; an edge is considered a problematic edge if the root mean square fitting error is greater than the fitting error threshold. When all edges pass the verification, the ordered node sequence and the set of verified edges are combined to generate road network data.

Images