Multi-channel anti-jamming communication system and method for agv cluster

Abstract

The application discloses an AGV cluster-oriented multi-channel anti-interference communication system and method, and belongs to the technical field of communication processing. The method specifically comprises the following steps: constructing a cluster communication topology based on a predicted time window, time-series modeling the path intersection relationship, relative distance change and shielding evolution of an AGV node, combining a decoy monitoring channel to form an interference migration sensing mechanism, generating a channel risk image and constructing a forbidden channel set on this basis, and when the main communication channel is abnormal or the predicted risk exceeds a threshold, implementing hierarchical data switching and cluster peak-shifting takeover control according to the forbidden constraint, task urgency and the relay capability of adjacent AGV nodes, realizing path-carrying of scheduling control data, relay links and non-control data; the application reduces the channel congestion and interference superposition risk caused by multi-node synchronous switching, and enhances the continuity and anti-interference capability of the communication link under the condition of rapid topology change.

Description

Technical Field

[0001] This invention belongs to the field of communication processing technology, specifically a multi-channel anti-interference communication system and method for AGV clusters. Background Technology

[0002] In intelligent manufacturing and flexible logistics systems, automated guided vehicles (AGV) clusters are widely used for material handling, warehouse scheduling, and production line collaborative operations. As the number of AGVs increases and the operational scenarios become more complex, continuous high-frequency control and status information exchange is required between AGVs and between AGVs and the scheduling center, which places high demands on the reliability and low latency of wireless communication systems.

[0003] Existing AGV wireless communication solutions typically employ single-channel optimization or dynamic switching mechanisms based on channel quality metrics, such as selecting the optimal channel based on received signal strength, signal-to-noise ratio, or channel occupancy. These methods primarily rely on local measurements at the current moment, lacking the ability to predict topology changes and interference migration trends within future time windows. In scenarios with dense AGV cluster operation, multiple AGVs may synchronously switch to the same high-quality channel within a short period, leading to channel congestion and secondary interference, resulting in a decrease in overall communication stability. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a multi-channel anti-interference communication system and method for AGV clusters, which can improve the continuity and anti-interference capability of AGV cluster communication by combining cluster topology prediction, multi-channel risk assessment and hierarchical switching control.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Multi-channel anti-interference communication methods for AGV clusters include:

[0007] The pose information, movement path, task urgency, adjacent link quality, and interference sampling data of each available wireless channel of each AGV node are obtained. Based on the path intersection relationship, relative distance change and occlusion change of each AGV node within the preset prediction time window, a cluster prediction communication map is constructed. The cluster prediction communication map is used to characterize the future adjacency relationship, future communication load and future link interruption risk of each AGV node within the preset prediction time window.

[0008] Based on the interference sampling data, cluster prediction communication map, and pre-set decoy listening channels, a channel risk profile for each AGV node is generated, and a set of forbidden channels is constructed based on the channel risk profile. The decoy listening channels are used to carry low-priority detection data or non-control data, and are used to obtain the migration trend of non-task interference between different channels.

[0009] When the main communication channel of any AGV node meets the anomaly criterion, or when the risk value of its main communication channel exceeds the preset risk threshold within the preset prediction time window, a hierarchical switching instruction is generated based on the forbidden channel set, task urgency and relay capability of adjacent AGV nodes. The hierarchical switching instruction is used to switch scheduling control data to the backup communication channel or the relay link of adjacent AGV nodes, and to switch non-control data to the delay-tolerant communication channel or the decoy listening channel.

[0010] Based on the cluster prediction communication graph, multiple AGV nodes with adjacency or path intersection relationships are identified, and different switching time slots, different target communication channels, or different relay AGV nodes are assigned to these multiple AGV nodes, so that the multiple AGV nodes do not switch to the same target communication channel synchronously in the same switching time slot.

[0011] Specifically, the step of constructing a cluster prediction communication map based on the path intersection relationships, relative distance changes, and occlusion changes of each AGV node within a preset prediction time window includes:

[0012] Based on the current position, direction of movement, and task path of each AGV node, determine the predicted movement area of ​​each AGV node within the preset prediction time window;

[0013] The predicted movement areas of different AGV nodes are compared in time series to identify AGV node pairs that have overlapping paths, adjacent paths, or shared passage areas within the same time period, and corresponding path intersection markers are generated.

[0014] The relative distance changes between the AGV node pairs are determined according to the time sequence, and corresponding adjacency relationship tags are generated based on the relative distance changes.

[0015] Based on the work area map and preset occlusion object information, the communication path between AGV node pairs is occluded, and corresponding occlusion markers are generated.

[0016] Based on the path intersection markers, adjacency markers, and occlusion markers, a cluster prediction communication graph including multiple time layers is generated in chronological order.

[0017] Specifically, based on the path intersection markers, adjacency markers, and occlusion markers, a cluster prediction communication graph including multiple time layers is generated in chronological order, including:

[0018] The preset prediction time window is segmented according to the preset time granularity to obtain multiple continuous time periods, and the path intersection mark, adjacency mark and occlusion mark are respectively assigned to the corresponding continuous time periods;

[0019] Within each consecutive time period, the path intersection markers, adjacency markers, and occlusion markers belonging to the same AGV node pair are merged to generate the communication edge state of the corresponding AGV node pair within that consecutive time period.

[0020] Based on the communication edge status of each AGV node within the same continuous time period, a corresponding time layer is generated. Each time layer records the communication edge status between each AGV node within the continuous time period and the channel candidate range corresponding to the communication edge status.

[0021] Multiple time layers are associated in chronological order, and edge state continuation markers, edge state disappearance markers, or edge state addition markers are written for the same AGV node in adjacent time layers to obtain the cluster prediction communication graph.

[0022] Specifically, based on the interference sampling data, the cluster predicted communication map, and the pre-set decoy listening channels, a channel risk profile for each AGV node is generated, and a set of forbidden channels is constructed based on the channel risk profile, including:

[0023] The interference sampling data of each available wireless channel is assigned to the corresponding time layer of the cluster prediction communication graph according to the sampling time, and the interference sampling data in the pre-set decoy listening channel is marked as a decoy interference event.

[0024] Track decoy interference events in continuous time layers. When interference sampling data in the decoy monitoring channel appears continuously, intermittently, or cross-channel between different time layers, generate corresponding interference migration markers.

[0025] The interference migration markers are time-aligned with the path intersection markers, adjacency markers, and occlusion markers in the cluster prediction communication graph to determine the candidate channel risk entries corresponding to each AGV node in each time layer.

[0026] Candidate channel risk entries are collected according to AGV node labels, and a corresponding channel risk profile is generated for each AGV node. The channel risk profile includes candidate channel identifier, corresponding time layer, decoy interference association label, path intersection association label, adjacency change association label, and occlusion association label.

[0027] Based on the channel risk profile of each AGV node, candidate channels are marked as prohibited and the candidate channels marked as prohibited are written into the prohibited channel set of the corresponding AGV node.

[0028] Specifically, the step of aggregating candidate channel risk entries according to AGV node labels and generating a corresponding channel risk profile for each AGV node includes:

[0029] Based on AGV node labeling, candidate channel risk entries belonging to the same AGV node are grouped into the same node entry set;

[0030] Within each node entry set, candidate channel risk entries corresponding to the same candidate channel identifier are arranged in chronological order to generate the corresponding candidate channel time chain entry.

[0031] The associated markers in each time chain entry are organized. When multiple types of markers, such as decoy interference, path intersection, adjacency change, or occlusion, appear sequentially in adjacent time layers for the same candidate channel, a channel shadow label is affixed to the candidate channel. The channel shadow label is used to characterize that although the candidate channel meets the availability conditions in the current time layer, it has a basis for disqualification triggering in subsequent time layers.

[0032] The node entry set, time chain entry, and channel shadow label are encapsulated according to the AGV node identifier to generate the corresponding AGV node channel risk profile.

[0033] Specifically, the step of marking candidate channels as prohibited based on the channel risk profile of each AGV node, and writing the candidate channels with the prohibited marking into the prohibited channel set of the corresponding AGV node, includes:

[0034] Candidate channels with channel shadow tags are extracted from the channel risk profile to form a set of candidate channels to be determined for the corresponding AGV node;

[0035] The current availability of candidate channels in the set of candidate channels to be determined is verified. When any candidate channel meets the channel availability condition at the current time layer and has a channel shadow label at a subsequent time layer, the candidate channel is determined as a reverse-disabled channel.

[0036] Based on the associated marker sequence corresponding to the reverse forbidden channel in the channel risk profile, a forbidden channel is labeled with a forbidden reason tag. The forbidden reason tag includes one or more of the following: decoy interference forbidden tag, path intersection forbidden tag, adjacency change forbidden tag, and occlusion forbidden tag.

[0037] According to the AGV node identifier, the reverse disallowed channel with the disallowed reason label is written into the disallowed channel set of the corresponding AGV node.

[0038] Specifically, when the main communication channel of any AGV node meets the anomaly criterion, or when the risk value of its main communication channel within the preset prediction time window exceeds a preset risk threshold, a hierarchical switching instruction is generated based on the forbidden channel set, task urgency, and relay capability of adjacent AGV nodes, including:

[0039] Trigger judgment is performed on the main communication channel of each AGV node. When the main communication channel of any AGV node meets the abnormality criterion, or when the main communication channel has a prohibition reason label within the preset prediction time window, the AGV node is determined as the AGV node to be switched.

[0040] Read the forbidden channel set corresponding to the AGV node to be switched, and exclude the candidate channels included in the forbidden channel set from the available wireless channels to obtain the set of optional bearer channels corresponding to the AGV node to be switched;

[0041] Based on the task urgency of the AGV node to be switched, the data to be transmitted by the AGV node to be switched is divided into scheduling control data and non-control data. Priority switching flags are assigned to the scheduling control data, and delayed switching flags or decoy carrying flags are assigned to the non-control data.

[0042] Based on the cluster prediction communication graph, candidate relay AGV nodes are selected from among the AGV nodes that are adjacent to the AGV node to be switched and are not in the time layer corresponding to the forbidden channel set of the AGV node to be switched. Relay takeover markers are generated based on the edge state continuation markers of the candidate relay AGV nodes in adjacent time layers.

[0043] Based on the set of selectable bearer channels, priority switching flag, delayed switching flag or decoy bearer flag, and relay takeover flag, a hierarchical switching instruction is generated. The hierarchical switching instruction includes the target bearer mode for scheduling control data, the target bearer mode for non-control data, the switching execution time layer, and the candidate relay AGV node identifier.

[0044] Specifically, based on the cluster prediction communication graph, candidate relay AGV nodes are selected from among the AGV nodes that are adjacent to the AGV node to be switched but are not in the time layer corresponding to the forbidden channel set of the AGV node to be switched, and a relay takeover marker is generated, including:

[0045] Extract AGV nodes that have an adjacency relationship with the AGV node to be switched in the current time layer or a subsequent time layer, and generate an initial relay candidate set;

[0046] Based on the forbidden channel set of the AGV node to be switched, AGV nodes that use the same forbidden channel as the initial relay candidate set, have the same forbidden reason label, or are under the same occlusion mark in the time layer corresponding to the forbidden channel set are eliminated to obtain the filtered relay candidate set.

[0047] In the set of candidate relay nodes, the edge state continuation markers of each AGV node and the AGV node to be switched in adjacent time layers are read, and the AGV nodes with continuous edge state continuation markers and no edge state disappearance markers are determined as candidate relay AGV nodes.

[0048] The node identifier, corresponding time layer identifier, channel identifier, and edge state continuation marker between the candidate relay AGV node and the AGV node to be switched are bound together to generate a relay takeover marker.

[0049] Specifically, the step of generating hierarchical handover instructions based on the selectable bearer channel set, priority handover flag, delayed handover flag or decoy bearer flag, and relay takeover flag includes:

[0050] The first bearer object for carrying scheduling control data is determined from the set of optional bearer channels according to the priority switching flag. The first bearer object is either a backup communication channel or a bearable channel corresponding to a candidate relay AGV node specified by the relay takeover flag.

[0051] Based on the delay switching flag or the decoy bearer flag, a second bearer object for carrying non-control data is determined. When the delay switching flag is read, the second bearer object is a delay-tolerant communication channel; when the decoy bearer flag is read, the second bearer object is a decoy listening channel.

[0052] The first bearer object and the second bearer object are separated and verified. When the first bearer object and the second bearer object correspond to the same wireless channel, the same candidate relay AGV node or the same switching time layer, the bearer channel or the switching time layer of the second bearer object is reassigned.

[0053] The first carrier object, the second carrier object, the relay takeover marker, the delay switching marker or the decoy carrier marker, and the corresponding switching time layer are encapsulated to generate a layered switching instruction. The layered switching instruction includes a scheduling control data switching field, a non-control data switching field, a relay takeover field, and a time layer execution field.

[0054] A multi-channel anti-interference communication system for AGV clusters, used to implement the multi-channel anti-interference communication method for AGV clusters, includes: a communication graph construction module, a profile generation module, an instruction generation module, and a communication switching module;

[0055] The communication graph construction module is used to acquire the pose information, movement path, task urgency, adjacent link quality, and interference sampling data of each AGV node and each available wireless channel. Based on the path intersection relationship, relative distance change, and occlusion change of each AGV node within the preset prediction time window, a cluster prediction communication graph is constructed. The cluster prediction communication graph is used to characterize the future adjacency relationship, future communication load, and future link interruption risk of each AGV node within the preset prediction time window.

[0056] The profile generation module is used to generate a channel risk profile for each AGV node based on the interference sampling data, cluster prediction communication map and pre-set decoy listening channel, and to construct a set of forbidden channels based on the channel risk profile. The decoy listening channel is used to carry low-priority detection data or non-control data and to obtain the migration trend of non-task interference between different channels.

[0057] The instruction generation module is used to generate a hierarchical switching instruction based on the forbidden channel set, task urgency and relay capability of adjacent AGV nodes when the main communication channel of any AGV node meets the anomaly criterion, or when the risk value of its main communication channel exceeds the preset risk threshold within the preset prediction time window. The hierarchical switching instruction is used to switch scheduling control data to the backup communication channel or the relay link of adjacent AGV nodes, and to switch non-control data to the delay-tolerant communication channel or the decoy listening channel.

[0058] The communication switching module is used to determine multiple AGV nodes that have an adjacency relationship or a path intersection relationship based on the cluster predicted communication graph, and to allocate different switching time slots, different target communication channels or different relay AGV nodes to the multiple AGV nodes, so that the multiple AGV nodes do not switch to the same target communication channel synchronously in the same switching time slot.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] This invention proposes a multi-channel anti-interference communication system and method for AGV clusters. It constructs a cluster communication topology based on a prediction time window, performs time-series modeling of path intersection, adjacency changes, and occlusion processes, and generates a channel risk profile by combining interference sampling data and decoy listening channels. Based on this, a set of forbidden channels is constructed. When the main communication channel is abnormal or the predicted risk exceeds a threshold, a hierarchical switching command is generated based on forbidden channel constraints, task urgency, and the relay capability of adjacent AGVs. This enables AGV clusters to avoid misjudging local optima based on single channel quality in complex and dynamic industrial environments, reduces the risk of channel congestion and interference superposition caused by multi-node synchronous switching, and enhances the continuity and anti-interference capability of communication links under conditions of rapid topology changes. Attached Figure Description

[0061] Figure 1 The flowchart of the multi-channel anti-interference communication method for AGV clusters provided by the present invention is shown below.

[0062] Figure 2 This invention provides a schematic diagram of an industrial warehousing operation area.

[0063] Figure 3 This is a schematic diagram of the channel risk profile provided by the present invention;

[0064] Figure 4 This is a diagram of the multi-channel anti-interference communication system architecture for AGV clusters provided by the present invention. Detailed Implementation

[0065] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0066] 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.

[0067] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0068] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0069] Example 1

[0070] Please see Figures 1-3 This invention provides an embodiment of a multi-channel anti-interference communication method for AGV clusters, applied to an industrial warehousing operation area. The operation area includes an inbound area, a shelving area, a buffer area, a sorting area, and main and branch aisles connecting the various areas. The shelving area contains multiple rows of metal shelves, and conveying and lifting equipment are installed near the main aisles. A scheduling node and twenty-four AGV nodes are deployed in the operation area. Each AGV node receives path control instructions, speed control instructions, and task scheduling instructions sent by the scheduling node through a wireless communication link, and uploads its current position, running status, and task execution information to the scheduling node.

[0071] Each AGV node's wireless communication module supports four wireless channels: CH1, CH2, CH3, and CH4. During the current scheduling cycle, CH1 serves as the primary communication channel used by most AGV nodes, CH2 and CH3 serve as backup communication channels, and CH4 serves as a decoy listening channel. CH4 does not carry scheduling control data such as emergency braking commands or path change commands. Instead, it carries the operation log index, status summary, cached confirmation information, and periodic probe data. The scheduling node obtains the occurrence and migration of non-task interference in different time periods by using the channel occupancy status of CH4, the confirmation status of probe data, and the changes in interference energy.

[0072] like Figure 2 As shown, taking AGV node A07 as the node to be analyzed, A07 is currently located in the first branch channel and is performing the task of conveying materials to the sorting area. Its task urgency level is first. According to the current task path, A07 will enter the intersection area between the first branch channel and the main channel in three seconds, and will move towards the sorting area along the main channel in six seconds. AGV node A12 is expected to enter the same intersection area from another branch channel in four seconds. AGV node A18 runs in the same direction along the main channel and will remain in the adjacent area in front of A07 for the next eight seconds. Figure 2 Solid lines represent the operating path, and dashed lines represent communication / relay links.

[0073] The scheduling node receives the AGV node status every 500 milliseconds and executes a communication decision every two seconds. This communication decision uses a preset 12-second prediction time window, which is divided into 12 consecutive time periods with a preset time granularity of one second, corresponding to 12 time layers. The current moment corresponds to the first time layer, and the following 11 seconds correspond to the second to twelfth time layers respectively. The length of the preset prediction time window is determined based on the average driving speed of the automated guided vehicle, the average travel time of a single channel, the communication decision execution cycle, and the total delay of channel switching. The value range is usually set to eight to twenty seconds. The working area has long channels and the AGV driving speed is relatively high. A larger value is used when the current level is low, and a smaller value is used when the topology of the work area is complex and path switching is frequent. The preset time granularity is aligned with the automated guided vehicle (AGV) status reporting cycle and the channel interference sampling cycle, and the value range is usually set to 500 milliseconds to 2 seconds. The smaller the granularity, the higher the topology prediction accuracy, but the computational cost increases accordingly. This embodiment selects a time granularity of one second, consistent with the AGV status reporting cycle, balancing prediction accuracy and real-time computation. The prediction time window and time granularity are dynamically adjusted according to the communication stability statistics of the work scenario. When the communication interruption rate is higher than the preset upper limit for several consecutive decision cycles, the prediction time window is appropriately extended to improve the prediction lead time.

[0074] The specific steps include the following:

[0075] Step S1: Obtain the pose information, movement path, task urgency, adjacent link quality, and interference sampling data of each AGV node. Based on the path intersection relationship, relative distance change, and occlusion change of each AGV node within the preset prediction time window, construct a cluster prediction communication map. The cluster prediction communication map is used to characterize the future adjacency relationship, future communication load, and future link interruption risk of each AGV node within the preset prediction time window.

[0076] The pose information includes AGV node identifier, current position, current orientation, current speed, and acquisition time. The position information is obtained by existing laser positioning, visual positioning, ultra-wideband positioning, or a combination of positioning methods. The motion path is a sequence of path nodes generated and distributed by the scheduling node, including the position of each path node, channel identifier, and estimated arrival time. The quality of adjacent links is obtained through periodic probe frames sent between AGV nodes, specifically recording the received signal strength, acknowledgment frame return delay, number of consecutive frame drops, and number of retransmissions. The interference sampling data is reported by the wireless communication module of each AGV node and the fixed access equipment after scanning the channels from CH1 to CH4, including the channel identifier, sampling time, sampling area, channel busy ratio, background energy, effective frame ratio, and number of retransmissions.

[0077] The task path of A07 is segmented according to time layers. Based on A07's current position, current speed, path speed limit, and estimated arrival time of each path node, it is determined that A07 is located in the first branch channel in the first to third time layers, enters the intersection area between the first branch channel and the main channel in the fourth to sixth time layers, and moves towards the sorting area along the main channel in the seventh to twelfth time layers. The scheduling node determines the predicted movement area of ​​A07 by combining the expected path interval of A07 in each time layer with the positioning allowable deviation.

[0078] Using the same method, the predicted movement areas of A12, A18, and other AGV nodes within twelve time layers were determined. A12 entered the same intersection area as A07 in the fourth to sixth time layers; A18 moved along the main channel in the third to tenth time layers and was within the preset adjacency range of A07 in the fifth to eighth time layers. The preset adjacency range was calculated and determined based on the transmit power and receive sensitivity of the AGV's onboard wireless communication module and the average path loss in the industrial warehousing environment. In the scenario of unobstructed line of sight, it was set to 20 meters by default, that is, when the distance between two AGVs does not exceed 20 meters, they are determined to be within the adjacency range.

[0079] Subsequently, a time-series comparison is performed on the predicted movement areas of different AGV nodes. The time-series comparison satisfies both spatial and temporal conditions; that is, a path intersection marker is only generated when two AGV nodes enter overlapping areas, adjacent areas, or shared passage areas within the same time layer.

[0080] For A07 and A12, their predicted movement areas in the fourth to sixth time layers both cover the intersection area between the first branch channel and the main channel. Therefore, the path intersection markers corresponding to A07 and A12 are written in the fourth, fifth and sixth time layers, and the intersection type is recorded as a shared intersection area. For A07 and A18, their task paths do not directly intersect, but they are both located in the main channel in the fifth to eighth time layers. Their path relationship is recorded as a shared channel in the same direction.

[0081] After determining the path intersection, the relative distance change between AGV node pairs is determined based on the predicted movement areas within each time layer. Specifically, the scheduling node reads the shortest distance between two predicted movement areas and, combined with the movement directions of the two AGV nodes, records the relative distance change as approaching, maintaining, or moving away.

[0082] A07 and A18 were not yet within the preset adjacency range in the third time layer. They began to approach each other in the fourth time layer, remained within the preset adjacency range in the fifth to eighth time layers, and gradually moved away from each other after the ninth time layer. Therefore, an adjacency addition marker was written in the fourth time layer, an adjacency continuation marker was written in the fifth to eighth time layers, and an adjacency disappearance marker was written in the ninth time layer. A07 and A12 were within the adjacency range in the fourth to sixth time layers, and left the adjacency range after the seventh time layer due to their different driving directions.

[0083] Furthermore, the communication path between each AGV node pair is determined based on the work area map and the preset obstruction object information. The preset obstruction objects include the areas where metal shelves, walls, lifting equipment and large fixed equipment are located, which are pre-recorded in the work area map. For mobile or temporary storage areas, the warehouse management equipment updates the location and validity period of the mobile equipment to the scheduling node.

[0084] In the fifth time layer, A07 and A12 are both located in the intersection area, and there are no metal shelves obstructing them, so no obstruction mark is written. In the fifth to eighth time layers, A07 and A18 are both located in the main aisle, and there are no fixed obstructions between them. The scheduling node records the communication path between A07 and A18 as unobstructed. A07 is separated from AGV node A09 located in the second branch aisle by a metal shelf, so an obstruction mark is written in the corresponding time layer.

[0085] Path intersection markers, adjacency markers, and occlusion markers are assigned to their respective time layers. For multiple types of markers within the same time layer and for the same AGV node pair, they are merged to form the communication edge state of that AGV node pair.

[0086] For example, the communication edge status between A07 and A18 in the fifth time layer includes a shared channel marker, an adjacency continuation marker, and an unobstructed marker; the communication edge status between A07 and A12 in the fifth time layer includes a shared intersection area marker, an adjacency continuation marker, and an unobstructed marker; and the communication edge status between A07 and A09 includes a path adjacency marker and an obstruction marker.

[0087] Each communication edge state is also associated with a channel candidate range. The channel candidate range is determined based on the channels supported by the wireless communication modules of the relevant AGV nodes and the basic coverage of the corresponding work area. Both A07 and A18 support CH1 to CH4, so the initial channel candidate range corresponding to their communication edge states includes CH1, CH2, CH3 and CH4. This channel candidate range only represents the channels participating in subsequent risk assessment, not the final channel switching.

[0088] The corresponding time layer is generated based on the AGV nodes and their communication edge states within each time layer, and the twelve time layers are associated in chronological order. For the same AGV node pair, if its communication edge state exists in both adjacent time layers, a continuation flag for the edge state is written; if it only appears in the later time layer, a new flag for the edge state is written; if it only appears in the previous time layer, a disappearance flag for the edge state is written.

[0089] The communication edge between A07 and A18 is added in the fourth time layer, continues in the fifth to eighth time layers, and disappears in the ninth time layer. The communication edge between A07 and A12 is added in the fourth time layer, continues in the fifth and sixth time layers, and disappears in the seventh time layer. After processing all AGV node pairs in the above manner, a cluster prediction communication graph including twelve time layers is obtained.

[0090] Based on the number of AGV nodes entering the same work area, having the same channel candidate range, and connected to the same fixed access device within each time layer, the future communication load of the corresponding time layer is recorded. In the fifth time layer, A07, A12, and two other AGV nodes are all expected to enter the intersection area, and all use CH1 and CH2 as candidate channels. Therefore, the intersection area in the fifth time layer is recorded as a state of rising communication load.

[0091] The future communication load is quantified using regional channel load density. The calculation method is as follows: within the same time layer and the same local operating area, the number of AGV nodes expected to use the same candidate channel is divided by the theoretical maximum number of access nodes for the corresponding channel in that area. The resulting ratio is the load density of that channel in that area. The theoretical maximum number of access nodes is determined based on the channel capacity of the fixed access equipment in that area and the average service bandwidth per node. In this embodiment, the theoretical maximum number of access nodes per channel in the intersection area is 6. The load density is divided into three levels: less than or equal to 0.3 indicates a low load state, between 0.3 and 0.6 indicates a medium load state, and greater than 0.6 indicates a high load state, i.e., a rising communication load state. In the intersection area of ​​the fifth time layer, a total of 4 AGV nodes are expected to use the second backup channel, corresponding to a load density of 4 / 6≈0.67, which is greater than 0.6, therefore it is determined to be a rising communication load state.

[0092] For AGV node pairs whose edge states disappear in subsequent time layers, appear with occlusion markers, or change from adjacency continuation to adjacency disappearance, the corresponding future link interruption risk is recorded; the cluster predictive communication graph simultaneously records future adjacency relationships, future communication load, and future link interruption risk.

[0093] Step S2: Based on the interference sampling data, cluster prediction communication map and pre-set decoy listening channels, generate a channel risk profile for each AGV node, and construct a set of forbidden channels based on the channel risk profile. The decoy listening channels are used to carry low-priority detection data or non-control data, and are used to obtain the migration trend of non-task interference between different channels.

[0094] After the cluster predicts the communication map, the interference sampling data of CH1 to CH4 are assigned to the corresponding time layer according to the sampling time and sampling area. For example, in the second time layer, the fixed access device located at the entrance of the shelf area detects that the channel busy ratio of CH4 continues to increase, and the confirmation rate of CH4 detection data reported by A07 and A12 decreases at the same time. The scheduling node assigns this group of sampling data to the second time layer and marks it as a CH4 decoy interference event.

[0095] In the third time layer, the background energy of CH4 briefly decreased, but the effective detection frame ratio was still lower than normal. The scheduling node continued to record intermittent decoy interference events. In the fourth time layer, CH4 returned to normal, while CH2 experienced an increased channel busy ratio and continuous retransmissions in the intersection area. Since the anomalies of CH4 and CH2 occurred in adjacent time layers, and the sampling area moved from the shelf area entrance to the intersection area along the expected movement direction of A07, a cross-channel interference migration marker pointing from CH4 to CH2 was generated.

[0096] The interference migration marker records the starting channel, target channel, starting time layer, target time layer, and corresponding sampling area. The interference migration marker does not need to identify a specific interference source; it is formed only based on the continuous relationship between interference events on different channels in time and space.

[0097] The interference migration markers from CH4 to CH2 are time-aligned with the cluster's predicted communication map. A07 will enter the convergence region in the fourth to sixth time layers. Therefore, the interference migration range of CH2 corresponds to the predicted movement region of A07 in the fourth time layer. At the same time, A07 and A12 have path convergence markers in the fourth to sixth time layers, and the future communication load in the convergence region is increasing.

[0098] Based on this, candidate channel risk entries for CH2 are generated for A07. The CH2 risk entries at the fourth time layer include decoy interference association markers, path intersection association markers, and adjacency change association markers; the CH2 risk entries at the fifth time layer include path intersection association markers and communication load association information; the CH2 risk entries at the sixth time layer include path intersection association markers. If a candidate channel is also related to occlusion changes at the corresponding time layer, an occlusion association marker is added to the risk entry.

[0099] According to the node identifier of A07, the candidate channel risk entries corresponding to CH1 to CH4 are included in the node entry set of A07. Within this node entry set, the risk entries of the same candidate channel are arranged in chronological order.

[0100] The CH2 time chain entry corresponding to A07 records the following in sequence: First time layer: no risk association; Second time layer: pre-interference information of decoy interference; Third time layer: interference migration association; Fourth time layer: decoy interference association, path intersection association, and adjacency change association; Fifth time layer: path intersection association and communication load increase information; Sixth time layer: path intersection association.

[0101] The scheduling node checks the order of association markers in the time chain entry. Since CH2 sequentially exhibits decoy interference association, path convergence association, and adjacency change association in adjacent time layers, and at least two types of association markers appear consecutively, the scheduling node writes a channel shadow label for CH2 corresponding to A07. This channel shadow label records the first trigger time layer as the fourth time layer, and is expected to continue until the sixth time layer. The triggering criteria include decoy interference migration, path convergence, and adjacency change.

[0102] The triggering of channel shadow tags requires both quantity and timing conditions to be met. The quantity condition is as follows: within at least two consecutive adjacent time layers, at least two different types of risk-associated tags must appear on the same candidate channel. Decoy interference and obstruction-associated tags are high-weight tags; if a single type appears consecutively for at least three time layers, the quantity condition is also met. The timing condition requires that risk tags take effect sequentially according to the time progression. A specific fixed order of tag types is not required; as long as the types of risk tags within adjacent time layers do not regress and the cumulative number of types increases or remains constant, it is considered sequential. If a tagging interruption occurs in a single time layer, but the same type or a new type of tag reappears in subsequent time layers, and the interruption duration does not exceed one time layer, it is considered temporally continuous. The effective duration of a channel shadow tag begins from the time layer where the triggering condition is first met and ends in the first time layer after all associated tags have disappeared.

[0103] The generation of channel shadow labels is not predicated on whether CH2 is currently unavailable. At the current first time layer, the received signal strength, acknowledgment frame delay, and retransmission count obtained by A07 after channel probing of CH2 are all within the allowable range, and CH2 still meets the current availability conditions. However, CH2 has formed continuous risk entries in the subsequent fourth to sixth time layers, so the channel is marked as having a basis for future disqualification.

[0104] like Figure 3 As shown, other candidate channels corresponding to A07 are processed in the same way, and the time chain entries and channel shadow labels of each channel are encapsulated into a channel risk profile of A07. This channel risk profile includes A07 node identifier, candidate channel identifier, time layer identifier, various associated markers and channel shadow labels.

[0105] Subsequently, CH2 with channel shadow tags is extracted from the channel risk profile of A07 to form a set of candidate channels to be judged, and the current availability of CH2 is verified. In this embodiment, the current availability conditions are that the channel busy ratio is lower than the preset upper limit, the number of consecutive retransmissions does not exceed the preset number, and the probe frame can be confirmed normally. CH2 meets the above conditions and is therefore a currently available channel.

[0106] Since CH2 is currently available, but has channel shadow labels in subsequent fourth to sixth time layers, the scheduling node determines CH2 as the reverse prohibited channel of A07. According to the risk entries corresponding to the channel shadow labels, the scheduling node writes decoy interference prohibited labels, path intersection prohibited labels, and adjacency change prohibited labels for CH2, and writes CH2, the prohibited label, and the prohibited effective range of the fourth to sixth time layers into the prohibited channel set of A07.

[0107] Using the same method, the scheduling node generates channel risk profiles and forbidden channel sets for other AGV nodes. Since the task paths and predicted movement areas of different AGV nodes are different, CH2 belongs to the forbidden channel of A07, but does not belong to the forbidden channel of AGV nodes located in other work areas.

[0108] Step S3: When the main communication channel of any AGV node meets the anomaly criterion, or when the risk value of its main communication channel exceeds the preset risk threshold within the preset prediction time window, a hierarchical switching instruction is generated based on the forbidden channel set, task urgency, and relay capability of adjacent AGV nodes. The hierarchical switching instruction is used to switch scheduling control data to the backup communication channel or the relay link of adjacent AGV nodes, and to switch non-control data to the delay-tolerant communication channel or the decoy listening channel.

[0109] A07 currently communicates with the scheduling node via CH1, continuously monitoring whether CH1 meets the anomaly criteria. These criteria include: failure to receive acknowledgment frames for multiple consecutive communication cycles; exceeding a preset number of retransmissions; a valid frame ratio below a preset ratio; or control data transmission delay exceeding the allowable range. The thresholds for these anomaly criteria are determined based on the closed-loop response requirements and safety braking delay requirements of the automated guided vehicle (AGV) motion control. The retransmission threshold corresponds to the maximum permissible delay of control commands; when the retransmission count reaches this threshold, the command issuance delay will exceed the motion control closed-loop cycle, leading to excessive path tracking deviation. The valid frame ratio threshold corresponds to the minimum command arrival rate required for stable operation of the control system; below this ratio, the continuity of motion control cannot be guaranteed. The transmission delay threshold corresponds to the maximum permissible transmission time for emergency braking commands; exceeding this value will pose a risk of delayed safety response. Different thresholds correspond to different task urgency levels, set based on the varying real-time control requirements of different tasks; higher urgency levels necessitate stricter communication quality requirements.

[0110] Specifically, the transmission period for scheduling control data is set to 100 milliseconds, and the acknowledgment frame waiting time is set to 80 milliseconds. If the same scheduling control data fails to receive an acknowledgment frame within the specified waiting time, retransmission is performed. For tasks of the first emergency level, if the same scheduling control data is retransmitted twice, or if the cumulative retransmission count within the most recent 1-second statistical time window reaches 3 times, the retransmission anomaly criterion is met. The effective frame ratio is determined by the ratio of the number of effectively received service frames to the number of service frames that should be received within the most recent 1-second period. A valid service frame is one whose frame verification is correct and whose node identifier and data sequence identifier match. Furthermore, for service frames arriving within the specified time, when A07 performs the first emergency level task, the effective frame ratio threshold is set to 95%, and for the second, third, or fourth emergency level tasks, it is set to 90%, 85%, and 80%, respectively. The control data transmission delay is the time interval between the scheduling node sending scheduling control data and A07 completing the reception and submission of the data. The allowable transmission delay for ordinary scheduling control data in the first emergency level task is set to 50 milliseconds, and for the second, third, or fourth emergency level tasks, it is set to 80 milliseconds, 100 milliseconds, and 150 milliseconds, respectively.

[0111] The urgency of a task is determined based on its remaining allowable time, the status of subsequent tasks, the cache status of the target work area, and the task's deferability status. When the remaining allowable time does not exceed a first time threshold, or the corresponding downstream work equipment has entered a waiting-for-materials state, the task is classified as a first-urgency task. When the remaining allowable time is greater than the first time threshold but not greater than the second time threshold, or there is a directly connected subsequent work task after the task is completed, the task is classified as a second-urgency task. When the remaining allowable time is greater than the second time threshold but not greater than the third time threshold, and there is no directly connected subsequent work task, the task is classified as a third-urgency task. When the task does not have a rigid completion deadline and allows for suspension, postponement, or replacement by other AGV nodes, the task is classified as a fourth-urgency task. The first, second, and third time thresholds are set to 5 minutes, 15 minutes, and 30 minutes, respectively.

[0112] In addition to the current anomaly criteria, the channel risk profile of A07 is also read to assess the risk of CH1 within the prediction time window. Risk assessment is based on the type of risk marker, the number of consecutive time layers, and the combination of markers. A low risk is defined as the appearance of only one type of isolated marker in a single time layer; a medium risk is defined as the appearance of two types of risk markers in adjacent time layers; and a high risk is defined as the appearance of decoy interference associations, obstruction associations, or three or more types of risk markers in adjacent time layers. In this embodiment, a preset risk threshold of 6 points is set for high risk.

[0113] The channel risk value is calculated by superimposing the label base weights and the time continuity coefficient. For a single candidate channel of a single AGV node, the risk value is calculated separately for each time layer, and the maximum value within the prediction time window is taken as the overall risk value of the channel. The specific calculation rules are as follows: First, a basic weight score is set for each type of risk-related marker. The basic weight scores for each type of risk-related marker are calibrated using the analytic hierarchy process (AHP) combined with field measurement data. Specifically, the basic score for decoy interference-related markers is 3 points, the basic score for occlusion-related markers is 3 points, the basic score for path intersection-related markers is 2 points, the basic score for adjacency change-related markers is 1 point, and the basic score for communication load increase-related information is 1 point. Second, a temporal continuity coefficient is set. When the same type of marker appears consecutively in adjacent time layers, the coefficient increases by 0.2 from the initial value of 1.0 for each subsequent time layer, up to a maximum of 2.0. If a marker reappears after being interrupted for one time layer, the continuity coefficient is reset to 1.0. The channel risk value within a single time layer is the sum of the basic scores of all effective risk markers within that time layer, multiplied by the corresponding temporal continuity coefficient.

[0114] Taking the first main communication channel corresponding to node A07 as an example, in the fifth time layer, this channel simultaneously has path intersection association markers and communication load increase association information, and both types of markers appear continuously starting from the fourth time layer, with a temporal continuity coefficient of 1.2, corresponding to a base score of 3 points. Therefore, the risk value of the fifth time layer is 3 × 1.2 = 3.6 points. In the sixth time layer, the channel adds an adjacency change association marker, increasing the base score to 4 points and the temporal continuity coefficient to 1.4, corresponding to a risk value of 4 × 1.4 = 5.6 points. Therefore, it is judged to be close to medium risk.

[0115] At the third time level, CH1 has not yet met the current anomaly criterion, but its risk at the fifth and sixth time levels is judged to be high. Therefore, the scheduling node identifies A07 as the AGV node to be switched and reads A07's forbidden channel set. Since CH2 has already been written into the forbidden channel set, CH2 is excluded from the available wireless channels. CH4 is a decoy listening channel and is not directly used as the preferred bearer channel for scheduling control data; therefore, CH3 is retained in A07's optional bearer channel set.

[0116] Data to be transmitted is processed according to the urgency level of task A07. A07 executes the first urgency level task, classifying speed control instructions, path node update instructions, stop instructions, task change instructions, and switchover confirmation information into the scheduling control data queue and writing a priority switchover flag; it classifies the operation log, historical trajectory, non-real-time diagnostic information, and map verification data into the non-control data queue.

[0117] For runtime logs and historical trajectories that are allowed to be transmitted in subsequent time layers, the scheduling node writes a delay switching flag; for state summaries and probe data used to continue sampling interference states, a decoy bearer flag is written.

[0118] Subsequently, AGV nodes that are adjacent to A07 in the current time layer and subsequent time layers are extracted from the cluster prediction communication graph, and A12, A15 and A18 are written into the initial relay candidate set.

[0119] First, the initial relay candidate set is filtered based on channel and risk association. A12 uses CH2 in the fourth to sixth time layers, while CH2 belongs to the forbidden channel set of A07 in the same time layer, so A12 is eliminated. Although A15 can use CH3, it is in the same metal shelf obstruction area as A07 in the fifth time layer and has the same obstruction prohibition criteria, so A15 is eliminated. A18 uses CH3, is not in the obstruction area corresponding to A07, and CH3 is not written into the forbidden channel set of A07 in the fifth to eighth time layers, so A18 is retained.

[0120] Further reading reveals the communication edge status of A07 and A18 in adjacent time layers. A07 and A18 show a newly added edge status marker in the fourth time layer, and have edge status continuation markers in the fifth to eighth time layers, until the edge status disappearance marker appears in the ninth time layer. Therefore, A18 meets the relay candidate conditions in the fifth to eighth time layers and is identified as a candidate relay AGV node.

[0121] The relay capability of adjacent AGV nodes is comprehensively evaluated through four quantitative indicators: remaining forwarding bandwidth percentage, expected relay stability duration, self-task load coefficient, and historical forwarding success rate. Each indicator is normalized to a score from 0 to 1, and the total relay capability score is obtained by weighting and summing according to preset weights. The qualification threshold is set at 0.7, and nodes with a total score greater than or equal to 0.7 are qualified to relay. The weights of the four evaluation indicators are determined by orthogonal experimental design combined with engineering measurement data. Among them, the remaining forwarding bandwidth and relay stability duration directly determine the carrying capacity and duration of the relay link, so they are given higher weights. Self-task load and historical forwarding success rate are auxiliary evaluation items with relatively lower weights. The qualification threshold is set based on the following: when the total relay capability score reaches the threshold, the transmission latency and packet loss rate of the scheduling control data are guaranteed to meet the requirements of the first emergency level task. The additional score rule is used to reward adjacent nodes with more stable link quality and guide the priority selection of unobstructed and long-duration relay links.

[0122] The weights and calculation rules for each indicator are as follows: The remaining forwarding bandwidth ratio has a weight of 0.3, calculated as the ratio of the difference between the node's current available wireless bandwidth and its own service-occupied bandwidth to the node's total available bandwidth; the expected relay stability duration has a weight of 0.3, calculated as the ratio of the duration of the adjacency relationship between the two nodes within the prediction time window to the total prediction time window duration; the self-task load coefficient has a weight of 0.2, using the inverse score of the node's current task urgency: 1.0 for the fourth urgency level, 0.7 for the third urgency level, 0.55 for the second urgency level, and 0.4 for the first urgency level. The higher the task urgency, the lower the self-task load coefficient score; the historical forwarding success rate has a weight of 0.2, calculated as the packet delivery rate of the node's relay forwarding tasks within the past thirty days. Additionally, if the adjacent edge status between the two nodes is uninterrupted and unobstructed during the effective relay period, an extra 0.05 points are awarded. A relay node can simultaneously carry the forwarding tasks of a maximum of two nodes; nodes exceeding this number are no longer included in the candidate set.

[0123] Taking node A18 as an example, its remaining forwarding bandwidth accounts for 0.8, corresponding to a score of 0.8; its adjacency duration with A07 is four time layers, accounting for approximately 0.33 of the twelve-layer prediction window, corresponding to a score of 0.33; its own task is at the third emergency level, corresponding to a load factor score of 0.7; its historical forwarding success rate is 0.95, corresponding to a score of 0.95; the weighted basic score is 0.8×0.3+0.33×0.3+0.7×0.2+0.95×0.2=0.669, combined with the additional score of 0.05 for continuous unobstructed edge states, the final score is 0.719, which is greater than the threshold of 0.7, thus qualifying it as a relay node.

[0124] The A18 node identifier, the fifth to eighth time layers, the CH3 channel identifier, and the edge state continuation flag between A07 and A18 are bound together to generate a relay takeover flag. This relay takeover flag specifies that A18 takes over the forwarding task of A07's scheduling control data in the fifth to seventh time layers.

[0125] Based on the priority switching flag, the first bearer of scheduling control data is determined from the set of optional bearer channels of A07. Since there is partial obstruction on the direct CH3 communication path between A07 and the fixed access equipment at the sixth time layer, while the CH3 communication paths between A18 and A07 and between A18 and the fixed access equipment are not marked as obstructed, A18 and its CH3 channel are determined as the first bearer.

[0126] For state summaries and probe data with decoy bearer tags, CH4 is identified as the second bearer. For runtime logs and historical tracks with delayed switching tags, their execution time is set after the ninth time layer, and the bearer channel is re-identified from channels that were never banned at that time.

[0127] The scheduling node performs separation verification on the first and second bearer objects. The scheduling control data of A07 is carried through A18 and CH3, and the probe data is carried through CH4. The two do not use the same radio channel. The scheduling control data is switched from the fifth time layer, and the probe data is sent in the sixth time layer. The two are not executed in the same switching time layer. Delayed data is arranged to the ninth time layer and does not compete with the scheduling control data for the relay forwarding time of A18.

[0128] After the separation verification is completed, a layered handover instruction is generated. The scheduling control data handover field in this instruction records the A07 node identifier, scheduling control data queue identifier, A18 node identifier, CH3 channel identifier, and the handover time at the fifth time layer; the non-control data handover field records that probe data is sent via CH4 at the sixth time layer, and delayed data is sent after the ninth time layer; the relay takeover field records the effective relay time range for A18; and the time layer execution field records the sequence of each handover action.

[0129] During the same time period, A12 and another AGV node A15 also need to switch channels. The scheduling node confirms that A07, A12 and A15 have path intersection relationships in the fourth to sixth time layers based on the cluster prediction communication map, and therefore does not allow the three to switch to the same target channel in the same handover time slot.

[0130] In this embodiment, each one-second time layer is divided into four handover time slots. The scheduling node arranges A07 in the first handover time slot of the fifth time layer and switches to CH3 via A18; it arranges A12 in the third handover time slot of the fifth time layer and switches to another channel outside its forbidden channel set; it arranges A15 to perform the handover in the sixth time layer, thereby forming the allocation results of different handover time slots, different target channels, or different relay AGV nodes.

[0131] During the handover execution, A18 first enters the takeover state based on the relay takeover field and sends a takeover confirmation on CH3. After receiving the takeover confirmation, the scheduling node sends handover execution information to A07. A07 stops sending new non-control data on CH1 and sends scheduling control data with priority handover flags to A18 via CH3. A18 then forwards the data to the scheduling node. Path control instructions sent by the scheduling node to A07 are also forwarded via A18.

[0132] In the sixth time layer, A07 transmits a state summary and probe data with decoy bearer tags via CH4. The scheduling node continues to record the channel busy ratio, probe frame acknowledgment status, and background energy of CH4, and uses the collected results for the next communication decision cycle. The operation log with the delay handover tag continues to be stored in A07's transmission queue until it is transmitted again after the ninth time layer.

[0133] After completing the handover and relay takeover, A07 and A18 respectively return an execution confirmation to the scheduling node. The execution confirmation records the actual handover time layer, the actual carried channel, the relay AGV node identifier, and the current link quality. If A18 does not return a takeover confirmation within the specified handover time slot of the fifth time layer, the scheduling node will not perform the handover action dependent on A18, but will instead reselect other candidate relay AGV nodes from the selected relay candidate set, or set CH3 as the direct backup communication channel for A07.

[0134] Step S4: Based on the cluster prediction communication graph, identify multiple AGV nodes that have adjacency or path intersection relationships, and assign different switching time slots, different target communication channels, or different relay AGV nodes to the multiple AGV nodes so that the multiple AGV nodes do not switch to the same target communication channel synchronously in the same switching time slot.

[0135] After completing this switchover, the quality of adjacent links is updated based on the execution confirmations returned by A07 and A18, and the pose information and motion path of each AGV node are reacquired in the next communication decision cycle.

[0136] If A07 continues to run along its original task path, the actual adjacency relationship between A07 and A18 in time layers 5 through 8 will be related to the cluster's predicted communication. Figure 1 If A07 changes its movement path due to temporary avoidance, the scheduling node will redetermine the predicted movement area based on the updated path and regenerate the path intersection marker, adjacency marker, and occlusion marker.

[0137] Simultaneously update the interference sampling data of each channel. If CH2 no longer exhibits decoy interference association, path intersection association, or adjacency change association within the new prediction time window, no new channel shadow label will be generated for CH2. After the original forbidden selection effective time layer ends, CH2 will be removed from the forbidden channel set of A07. If CH3 exhibits continuous interference migration markers within the new prediction time window, CH3 will be written into the new candidate channel risk entry according to the same process, and it will be re-determined whether the first bearer needs to be replaced.

[0138] Through the above continuous processing, the scheduling node sequentially obtains the predicted mobile area, path intersection mark, adjacency mark, obstruction mark, cross-time layer communication edge status, channel risk profile, channel shadow label, reverse forbidden channel, forbidden channel set, relay takeover mark and hierarchical switching instruction of A07, and accordingly completes the separate carrying of A07 scheduling control data, detection data and delay data, as well as the peak-shifting switching of A07, A12 and A15.

[0139] Example 2

[0140] Please see Figure 4 Another embodiment of the present invention provides a multi-channel anti-interference communication system for AGV clusters, comprising: a communication graph construction module, a profile generation module, an instruction generation module, and a communication switching module;

[0141] The communication graph construction module is used to acquire the pose information, movement path, task urgency, adjacent link quality, and interference sampling data of each AGV node and each available wireless channel. Based on the path intersection relationship, relative distance change, and occlusion change of each AGV node within the preset prediction time window, a cluster prediction communication graph is constructed. The cluster prediction communication graph is used to characterize the future adjacency relationship, future communication load, and future link interruption risk of each AGV node within the preset prediction time window.

[0142] The profile generation module is used to generate a channel risk profile for each AGV node based on the interference sampling data, cluster prediction communication map and pre-set decoy listening channel, and to construct a set of forbidden channels based on the channel risk profile. The decoy listening channel is used to carry low-priority detection data or non-control data and to obtain the migration trend of non-task interference between different channels.

[0143] The instruction generation module is used to generate a hierarchical switching instruction based on the forbidden channel set, task urgency and relay capability of adjacent AGV nodes when the main communication channel of any AGV node meets the anomaly criterion, or when the risk value of its main communication channel exceeds the preset risk threshold within the preset prediction time window. The hierarchical switching instruction is used to switch scheduling control data to the backup communication channel or the relay link of adjacent AGV nodes, and to switch non-control data to the delay-tolerant communication channel or the decoy listening channel.

[0144] The communication switching module is used to determine multiple AGV nodes that have an adjacency relationship or a path intersection relationship based on the cluster predicted communication graph, and to allocate different switching time slots, different target communication channels or different relay AGV nodes to the multiple AGV nodes, so that the multiple AGV nodes do not switch to the same target communication channel synchronously in the same switching time slot.

[0145] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0146] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-channel anti-interference communication method for AGV clusters, applied to an AGV cluster communication system including a scheduling node, multiple AGV nodes, and multiple available wireless channels, characterized in that, include: The pose information, movement path, task urgency, adjacent link quality, and interference sampling data of each available wireless channel of each AGV node are obtained. Based on the path intersection relationship, relative distance change and occlusion change of each AGV node within the preset prediction time window, a cluster prediction communication map is constructed. The cluster prediction communication map is used to characterize the future adjacency relationship, future communication load and future link interruption risk of each AGV node within the preset prediction time window. Based on the interference sampling data, cluster prediction communication map, and pre-set decoy listening channels, a channel risk profile for each AGV node is generated, and a set of forbidden channels is constructed based on the channel risk profile. The decoy listening channels are used to carry low-priority detection data or non-control data, and are used to obtain the migration trend of non-task interference between different channels. When the main communication channel of any AGV node meets the anomaly criterion, or when the risk value of its main communication channel exceeds the preset risk threshold within the preset prediction time window, a hierarchical switching instruction is generated based on the forbidden channel set, task urgency and relay capability of adjacent AGV nodes. The hierarchical switching instruction is used to switch scheduling control data to the backup communication channel or the relay link of adjacent AGV nodes, and to switch non-control data to the delay-tolerant communication channel or the decoy listening channel. Based on the cluster prediction communication graph, multiple AGV nodes with adjacency or path intersection relationships are identified, and different switching time slots, different target communication channels, or different relay AGV nodes are assigned to these multiple AGV nodes, so that the multiple AGV nodes do not switch to the same target communication channel synchronously in the same switching time slot.

2. The multi-channel anti-interference communication method for AGV clusters as described in claim 1, characterized in that, The process of constructing a cluster prediction communication graph based on the path intersection relationships, relative distance changes, and occlusion changes of each AGV node within a preset prediction time window includes: Based on the current position, direction of movement, and task path of each AGV node, determine the predicted movement area of ​​each AGV node within the preset prediction time window; The predicted movement areas of different AGV nodes are compared in time series to identify AGV node pairs that have overlapping paths, adjacent paths, or shared passage areas within the same time period, and corresponding path intersection markers are generated. The relative distance changes between the AGV node pairs are determined according to the time sequence, and corresponding adjacency relationship tags are generated based on the relative distance changes. Based on the work area map and preset occlusion object information, the communication path between AGV node pairs is occluded, and corresponding occlusion markers are generated. Based on the path intersection markers, adjacency markers, and occlusion markers, a cluster prediction communication graph including multiple time layers is generated in chronological order.

3. The multi-channel anti-interference communication method for AGV clusters as described in claim 2, characterized in that, Based on the path intersection markers, adjacency markers, and occlusion markers, a cluster prediction communication graph comprising multiple time layers is generated in chronological order, including: The preset prediction time window is segmented according to the preset time granularity to obtain multiple continuous time periods, and the path intersection mark, adjacency mark and occlusion mark are respectively assigned to the corresponding continuous time periods; Within each consecutive time period, the path intersection markers, adjacency markers, and occlusion markers belonging to the same AGV node pair are merged to generate the communication edge state of the corresponding AGV node pair within that consecutive time period. Based on the communication edge status of each AGV node within the same continuous time period, a corresponding time layer is generated; Multiple time layers are associated in chronological order, and edge state continuation markers, edge state disappearance markers, or edge state addition markers are written for the same AGV node in adjacent time layers to obtain the cluster prediction communication graph.

4. The multi-channel anti-interference communication method for AGV clusters as described in claim 1, characterized in that, Based on the interference sampling data, the cluster predicted communication map, and the pre-set decoy listening channels, a channel risk profile is generated for each AGV node, and a set of forbidden channels is constructed based on the channel risk profile, including: The interference sampling data of each available wireless channel is assigned to the corresponding time layer of the cluster prediction communication graph according to the sampling time, and the interference sampling data in the pre-set decoy listening channel is marked as a decoy interference event. Track decoy interference events in continuous time layers. When interference sampling data in the decoy monitoring channel appears continuously, intermittently, or cross-channel between different time layers, generate corresponding interference migration markers. The interference migration markers are time-aligned with the path intersection markers, adjacency markers, and occlusion markers in the cluster prediction communication graph to determine the candidate channel risk entries corresponding to each AGV node in each time layer. Candidate channel risk entries are collected according to AGV node labels, and a corresponding channel risk profile is generated for each AGV node. The channel risk profile includes candidate channel identifier, corresponding time layer, decoy interference association label, path intersection association label, adjacency change association label, and occlusion association label. Based on the channel risk profile of each AGV node, candidate channels are marked as prohibited and the candidate channels marked as prohibited are written into the prohibited channel set of the corresponding AGV node.

5. The multi-channel anti-interference communication method for AGV clusters as described in claim 4, characterized in that, The process of aggregating candidate channel risk entries according to AGV node labels and generating a corresponding channel risk profile for each AGV node includes: Based on AGV node labeling, candidate channel risk entries belonging to the same AGV node are grouped into the same node entry set; Within each node entry set, candidate channel risk entries corresponding to the same candidate channel identifier are arranged in chronological order to generate the corresponding candidate channel time chain entry. The associated markers in each time chain entry are organized. When multiple types of markers, such as decoy interference, path intersection, adjacency change, or occlusion, appear sequentially in adjacent time layers for the same candidate channel, a channel shadow label is affixed to the candidate channel. The channel shadow label is used to characterize that although the candidate channel meets the availability conditions in the current time layer, it has a basis for disqualification triggering in subsequent time layers. The node entry set, time chain entry, and channel shadow label are encapsulated according to the AGV node identifier to generate the corresponding AGV node channel risk profile.

6. The multi-channel anti-interference communication method for AGV clusters as described in claim 5, characterized in that, The step of marking candidate channels as prohibited based on the channel risk profile of each AGV node, and writing the candidate channels with the prohibited markings into the prohibited channel set of the corresponding AGV node, includes: Candidate channels with channel shadow tags are extracted from the channel risk profile to form a set of candidate channels to be determined for the corresponding AGV node; The current availability of candidate channels in the set of candidate channels to be determined is verified. When any candidate channel meets the channel availability condition at the current time layer and has a channel shadow label at a subsequent time layer, the candidate channel is determined as a reverse-disabled channel. Based on the associated marker sequence corresponding to the reverse forbidden channel in the channel risk profile, a forbidden channel is labeled with a forbidden reason tag. The forbidden reason tag includes one or more of the following: decoy interference forbidden tag, path intersection forbidden tag, adjacency change forbidden tag, and occlusion forbidden tag. According to the AGV node identifier, the reverse disallowed channel with the disallowed reason label is written into the disallowed channel set of the corresponding AGV node.

7. The multi-channel anti-interference communication method for AGV clusters as described in claim 1, characterized in that, When the main communication channel of any AGV node meets the anomaly criterion, or the risk value of its main communication channel within the preset prediction time window exceeds the preset risk threshold, a hierarchical switching instruction is generated based on the forbidden channel set, task urgency, and relay capability of adjacent AGV nodes, including: Trigger judgment is performed on the main communication channel of each AGV node. When the main communication channel of any AGV node meets the abnormality criterion, or when the main communication channel has a prohibition reason label within the preset prediction time window, the AGV node is determined as the AGV node to be switched. Read the forbidden channel set corresponding to the AGV node to be switched, and exclude the candidate channels included in the forbidden channel set from the available wireless channels to obtain the set of optional bearer channels corresponding to the AGV node to be switched; Based on the task urgency of the AGV node to be switched, the data to be transmitted by the AGV node to be switched is divided into scheduling control data and non-control data. Priority switching flags are assigned to the scheduling control data, and delayed switching flags or decoy carrying flags are assigned to the non-control data. Based on the cluster's predicted communication graph, candidate relay AGV nodes are selected from among the AGV nodes that are adjacent to the AGV node to be switched and are not in the time layer corresponding to the forbidden channel set of the AGV node to be switched, and a relay takeover marker is generated. Based on the set of selectable bearer channels, priority switching flag, delayed switching flag or decoy bearer flag, and relay takeover flag, a hierarchical switching instruction is generated. The hierarchical switching instruction includes the target bearer mode for scheduling control data, the target bearer mode for non-control data, the switching execution time layer, and the candidate relay AGV node identifier.

8. The multi-channel anti-interference communication method for AGV clusters as described in claim 7, characterized in that, According to the cluster prediction communication graph, among the AGV nodes that are adjacent to the AGV node to be switched but are not in the time layer corresponding to the forbidden channel set of the AGV node to be switched, candidate relay AGV nodes are selected and a relay takeover marker is generated, including: Extract AGV nodes that have an adjacency relationship with the AGV node to be switched in the current time layer or a subsequent time layer, and generate an initial relay candidate set; Based on the forbidden channel set of the AGV node to be switched, AGV nodes that use the same forbidden channel as the initial relay candidate set, have the same forbidden reason label, or are under the same occlusion mark in the time layer corresponding to the forbidden channel set are eliminated to obtain the filtered relay candidate set. In the set of candidate relay nodes, the edge state continuation markers of each AGV node and the AGV node to be switched in adjacent time layers are read, and the AGV nodes with continuous edge state continuation markers and no edge state disappearance markers are determined as candidate relay AGV nodes. The node identifier, corresponding time layer identifier, channel identifier, and edge state continuation marker between the candidate relay AGV node and the AGV node to be switched are bound together to generate a relay takeover marker.

9. The multi-channel anti-interference communication method for AGV clusters as described in claim 8, characterized in that, The step of generating hierarchical handover instructions based on the selectable bearer channel set, priority handover flag, delayed handover flag or decoy bearer flag, and relay takeover flag includes: The first bearer object is determined from the set of optional bearer channels according to the priority switching flag. The first bearer object is either a backup communication channel or a bearable channel corresponding to a candidate relay AGV node specified by the relay takeover flag. The second carrier object is determined based on the delay switching flag or the decoy carrier flag. When the delay switching flag is read, the second carrier object is a delay-tolerant communication channel. When the decoy carrier flag is read, the second carrier object is a decoy listening channel. The first bearer object and the second bearer object are separated and verified. When the first bearer object and the second bearer object correspond to the same wireless channel, the same candidate relay AGV node or the same switching time layer, the bearer channel or the switching time layer of the second bearer object is reassigned. The first carrier object, the second carrier object, the relay takeover marker, the delay switching marker or the decoy carrier marker, and the corresponding switching time layer are encapsulated to generate a layered switching instruction.

10. A multi-channel anti-interference communication system for AGV clusters, used to implement the multi-channel anti-interference communication method for AGV clusters as described in any one of claims 1-9, characterized in that, include: The module includes a communication graph construction module, a profile generation module, an instruction generation module, and a communication switching module. The communication graph construction module is used to acquire the pose information, movement path, task urgency, adjacent link quality, and interference sampling data of each AGV node and each available wireless channel. Based on the path intersection relationship, relative distance change, and occlusion change of each AGV node within the preset prediction time window, a cluster prediction communication graph is constructed. The cluster prediction communication graph is used to characterize the future adjacency relationship, future communication load, and future link interruption risk of each AGV node within the preset prediction time window. The profile generation module is used to generate a channel risk profile for each AGV node based on the interference sampling data, cluster prediction communication map and pre-set decoy listening channel, and to construct a set of forbidden channels based on the channel risk profile. The decoy listening channel is used to carry low-priority detection data or non-control data and to obtain the migration trend of non-task interference between different channels. The instruction generation module is used to generate a hierarchical switching instruction based on the forbidden channel set, task urgency and relay capability of adjacent AGV nodes when the main communication channel of any AGV node meets the anomaly criterion, or when the risk value of its main communication channel exceeds the preset risk threshold within the preset prediction time window. The hierarchical switching instruction is used to switch scheduling control data to the backup communication channel or the relay link of adjacent AGV nodes, and to switch non-control data to the delay-tolerant communication channel or the decoy listening channel. The communication switching module is used to determine multiple AGV nodes that have an adjacency relationship or a path intersection relationship based on the cluster predicted communication graph, and to allocate different switching time slots, different target communication channels or different relay AGV nodes to the multiple AGV nodes, so that the multiple AGV nodes do not switch to the same target communication channel synchronously in the same switching time slot.

Images