State synchronization method and device for agent cluster and computer device

Abstract

The present disclosure provides a state synchronization method and device for an agent cluster and a computer device, relates to the technical field of data communication, in particular to the fields of artificial intelligence, embodied intelligence, multi-agent system, etc., and achieves the scheme as follows: obtaining a plurality of state information of each agent node; determining a priority score of each state information based on weighted summation of a first component, a second component and a third component; determining a size of available bandwidth for state synchronization; determining a synchronization strategy for state synchronization based on the size of available bandwidth, in response to determining that the size of available bandwidth indicates bandwidth permission, determining to adopt a first synchronization strategy, periodically synchronizing the plurality of state information according to a preset synchronization period, and event-triggered synchronizing target state information with a higher priority score; and in response to determining that the size of available bandwidth indicates bandwidth limitation, determining to adopt a second synchronization strategy, and only event-triggered synchronizing the target state information.

Description

Technical Field

[0001] This disclosure relates to the field of data communication technology, and in particular to the fields of artificial intelligence, embodied intelligence, and multi-agent systems. It is applicable to scenarios such as communication, interaction, and collaboration of intelligent agent clusters, and in particular to a method, apparatus, computer device, computer-readable storage medium, and computer program product for state synchronization of intelligent agent clusters. Background Technology

[0002] In multi-agent systems, agents need to share state information, such as pose, velocity, and sensor data, via a network. However, in real-world deployment environments, network links are unstable, exhibiting issues such as bandwidth fluctuations, packet loss, and latency jitter. Traditional state synchronization methods struggle to balance real-time performance with bandwidth utilization, easily leading to information redundancy, conflict control issues, or command failures. Therefore, a more flexible state synchronization mechanism is urgently needed to improve system robustness, communication efficiency, and collaborative performance. Summary of the Invention

[0003] This disclosure provides a method, apparatus, computer device, computer-readable storage medium, and computer program product for state synchronization of intelligent agent clusters.

[0004] According to one aspect of this disclosure, a method for state synchronization of an intelligent agent cluster is provided. The intelligent agent cluster includes multiple intelligent agent nodes. The method includes: acquiring multiple state information of each intelligent agent node, including first state information associated with behavior control, second state information associated with task state, and third state information associated with environmental perception; determining a priority score for each state information based on a weighted sum of the first, second, and third components, wherein the first component characterizes the importance of the state information, the second component characterizes the rate of change of the state information, and the third component characterizes the timeliness of the state information, and each of the first, second, and third components has an adjustable weight coefficient; determining the available bandwidth size for state synchronization; and based on the available bandwidth size... A synchronization strategy for state synchronization is determined, wherein: in response to determining that the available bandwidth indicates that bandwidth is allowed, a first synchronization strategy is adopted, in which multiple state information are periodically synchronized according to a preset synchronization period, and at least one target state information among the multiple state information is synchronized by event triggering, wherein the synchronization period for periodically synchronizing the first and second state information among the multiple state information is smaller than the synchronization period for synchronizing the third state information, and the priority score of at least one target state information among the multiple state information is higher than the priority score of the other state information; in response to determining that the available bandwidth indicates that bandwidth is limited, a second synchronization strategy is adopted, in which only at least one target state information is synchronized by event triggering.

[0005] In some embodiments, determining the available bandwidth size for state synchronization includes: calculating an initial available bandwidth size based on network parameters; determining whether the calculated initial available bandwidth size is greater than or equal to the ratio of the semantic load size to be synchronized to a safety margin coefficient, wherein the first to third state information among a plurality of state information have their respective corresponding semantic load sizes, and the safety margin coefficient takes a value between 0 and 1; in response to determining that the initial available bandwidth size is greater than or equal to the ratio, determining that the available bandwidth size indicates bandwidth allowable; and in response to determining that the initial available bandwidth size is less than the ratio, determining that the available bandwidth size indicates bandwidth limited.

[0006] In some embodiments, in response to determining the available bandwidth size to indicate bandwidth limitation, the method further includes: discarding a portion of target state information from at least one target state information whose semantic load size does not match its semantic value, so that another portion of the retained target state information can be synchronized based on the available bandwidth size.

[0007] In some embodiments, the method further includes: after determining a synchronization strategy, performing at least one of state information encoding or compression, wherein: in response to the agent cluster performing cooperative operation and manipulation tasks, a time difference coding algorithm is used; and in response to the agent cluster performing cooperative perception tasks, a sparse matrix compression and hash redundancy tagging algorithm is used.

[0008] In some embodiments, the method further includes: performing channel selection and scheduling after performing at least one of state information encoding or compression, wherein: in response to the agent cluster performing a formation task, the first state information is processed using User Datagram Protocol (UDP) or a combination of UDP and Transmission Control Protocol (TCP), and the third state information is processed using real-time web communication.

[0009] In some embodiments, when any one of the multiple agent nodes acts as the sending node for state synchronization, the other agent nodes act as the receiving nodes for state synchronization, wherein: the sending node is configured to generate and send multiple data packets with timestamps based on a synchronization strategy; and the receiving node is configured to reassemble the multiple data packets based on the timestamps.

[0010] In some embodiments, the receiving node is further configured to: determine whether there is a target data packet in the plurality of data packets that exceeds the delay window; and in response to determining that there is a target data packet, discard the target data packet to reassemble the remaining data packets in the plurality of data packets.

[0011] In some embodiments, the first to third state information in the plurality of state information have their own corresponding queue layers, such that when the available bandwidth is lower than a preset threshold, the synchronization period of the queue layer corresponding to the third state information is extended, and only the queue layers corresponding to the first and second state information are allowed to transmit, until the available bandwidth recovers to above the preset threshold.

[0012] In some embodiments, event-triggered synchronization includes threshold triggering caused by the data value of state information reaching a preset threshold, so that event-triggered synchronization satisfies immediacy in the synchronization strategy.

[0013] In some embodiments, threshold triggering includes absolute threshold triggering and relative rate of change triggering.

[0014] According to one aspect of this disclosure, a state synchronization device for an intelligent agent cluster is provided. The intelligent agent cluster includes multiple intelligent agent nodes. The device includes: an information acquisition module configured to acquire multiple state information of each intelligent agent node, including first state information associated with behavior control, second state information associated with task state, and third state information associated with environmental perception; a priority determination module configured to determine a priority score for each state information based on a weighted sum of a first component, a second component, and a third component, wherein the first component characterizes the importance of the state information, the second component characterizes the rate of change of the state information, and the third component characterizes the timeliness of the state information, and each of the first, second, and third components has an adjustable weight coefficient; a bandwidth determination module configured to determine the available bandwidth size for state synchronization; and a policy determination module. The system is configured to determine a synchronization strategy for state synchronization based on available bandwidth, wherein: a first strategy module is configured to determine the first synchronization strategy in response to determining that the available bandwidth indicates that bandwidth is allowed. In the first synchronization strategy, multiple state information are periodically synchronized according to a preset synchronization period, and at least one target state information among the multiple state information is synchronized by an event. The synchronization period for periodically synchronizing the first and second state information among the multiple state information is shorter than the synchronization period for synchronizing the third state information, and the priority score of at least one target state information among the multiple state information is higher than the priority score of the other state information; a second strategy module is configured to determine the second synchronization strategy in response to determining that the available bandwidth indicates that bandwidth is limited. In the second synchronization strategy, only at least one target state information is synchronized by an event.

[0015] According to another aspect of this disclosure, a computer device is provided, comprising: at least one processor; and a memory having a computer program stored thereon, wherein when executed by the at least one processor, the computer program causes the at least one processor to perform the methods provided above in this disclosure.

[0016] According to another aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, causes the processor to perform the methods provided above in this disclosure.

[0017] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, causes the processor to perform the methods provided above in this disclosure.

[0018] According to one or more embodiments of this disclosure, an adaptive bandwidth-aware mechanism can be implemented, thereby improving robustness, communication efficiency, and cooperative performance in multi-agent system scenarios.

[0019] These and other aspects of this disclosure will be apparent from the embodiments described below, and will be elucidated with reference to the embodiments described below. Attached Figure Description

[0020] The accompanying drawings exemplify embodiments and form part of the specification, serving together with the textual description to explain exemplary implementations of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of this disclosure. Throughout the drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0021] Figure 1 This is a flowchart illustrating a state synchronization method for an intelligent agent cluster according to an exemplary embodiment.

[0022] Figure 2 This is a flowchart illustrating the process of determining the available bandwidth size according to an exemplary embodiment.

[0023] Figure 3 This is a schematic diagram illustrating a semantic degradation mechanism according to an exemplary embodiment.

[0024] Figure 4 This is a flowchart illustrating a state synchronization method for an agent cluster according to another exemplary embodiment.

[0025] Figure 5 This is a schematic diagram illustrating the queue hierarchy of state information according to an exemplary embodiment.

[0026] Figure 6This is a schematic block diagram illustrating a state synchronization device for an intelligent agent cluster according to an exemplary embodiment.

[0027] Figure 7 This is a block diagram illustrating an exemplary computer device that can be applied to an exemplary embodiment. Detailed Implementation

[0028] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0029] In this disclosure, unless otherwise stated, the use of terms such as "first," "second," etc., to describe various elements is not intended to limit the positional, temporal, or importance relationships of these elements; such terms are merely used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of that element, while in other cases, based on the context, they may refer to different instances.

[0030] The terminology used in the description of the various examples described in this disclosure is for the purpose of describing particular examples only and is not intended to be limiting. Unless the context explicitly indicates otherwise, an element may be one or more unless the number of elements is specifically limited. As used herein, the term "multiple" means two or more, and the term "based on" should be interpreted as "at least partially based on". Furthermore, the terms "and / or" and "at least one of..." cover any one of the listed items and all possible combinations thereof.

[0031] The embodiments of this disclosure provide a state synchronization method for an intelligent agent cluster, which can dynamically adjust the content, frequency and method of state synchronization according to network conditions, thereby realizing an adaptive bandwidth awareness mechanism and improving robustness, communication efficiency and collaborative performance in multi-agent system scenarios.

[0032] Figure 1 This is a flowchart illustrating a state synchronization method for an intelligent agent cluster according to an exemplary embodiment.

[0033] In the embodiments of this disclosure, an agent cluster refers to a multi-agent system comprising multiple agent nodes. In examples, an agent cluster can be a robot cluster, a drone cluster, an unmanned vehicle cluster, etc. Furthermore, agent clusters can be used to perform various tasks, such as collaborative operation and control tasks, collaborative perception tasks, and formation tasks.

[0034] like Figure 1 As shown, the state synchronization method for an intelligent agent cluster includes steps S101, S102, S103, and S104.

[0035] In step S101, multiple state information of each agent node is acquired. This multiple state information includes first state information associated with behavior control, second state information associated with task state, and third state information associated with environmental perception.

[0036] In this example, the multi-agent system focuses on the immediate impact of data on behavioral control, particularly related to the control loop of the agent nodes. The first state information associated with behavioral control can include pose (such as a six-dimensional pose with x-axis, y-axis, z-axis, roll, pitch, and yaw), velocity, acceleration, and angular velocity. This first state information directly affects the motion control of the agent nodes and, in some scenarios, also involves collision avoidance.

[0037] In the example, the multi-agent system also focuses on the immediate impact of data on task execution. Secondary state information associated with the task state can include task objectives, completion rate, task switching signals, and cooperation strategies. This secondary state information ensures consistent behavior among multiple agent nodes in tasks related to cooperation. Furthermore, tertiary state information associated with environmental perception can include point clouds, depth maps, and RGB images. This tertiary state information is used for mapping, obstacle detection, and dynamic programming. In some scenarios, the amount of data in the tertiary state information may be substantial.

[0038] In the example, the first state information, the second state information, and the third state information have different functions and effects, so they can have different levels of transmission. The level of the third state information can be lower than that of the first state information and the second state information.

[0039] Compared to the focus on data integrity and latency in traditional communication, the communication of multi-agent systems is more concerned with the immediate impact of data on behavior control and task execution. Therefore, by dividing the multiple state information of each agent node into fine-grained parts, it can help to quantify the priority of state information in the future.

[0040] In step S102, a priority score for each piece of state information is determined based on a weighted sum of the first, second, and third components. The first component characterizes the importance of the state information, the second component characterizes the rate of change of the state information, and the third component characterizes the timeliness of the state information. Each of the first, second, and third components has an adjustable weight coefficient.

[0041] In the example, the priority score can be calculated by the following expression: Pi = w1 × Ii + w2 × Ci + w3 × Ti, where: Pi represents the priority score, Ii represents the first component corresponding to importance, Ci represents the second component corresponding to the rate of change, Ti represents the third component corresponding to timeliness, and w1, w2 and w3 represent the weight coefficients corresponding to the three components, and each weight coefficient can be dynamically adjusted according to the task scenario of the agent cluster.

[0042] In this way, the priority of each state information can be quantified by scores (e.g., values ​​between 0 and 1) across three dimensions: the first component, the second component, and the third component, to calculate the priority score corresponding to that state information. For example, the first component can be divided into different score ranges, such as high scores (e.g., 0.8~1), medium scores (e.g., 0.5~0.8), and low scores (e.g., 0.2~0.5). For example, pose and velocity can be preset to high scores, environmental information detected by sensors can be preset to medium scores, and the user's own health status or battery level can be preset to low scores. Similarly, the score of the second component can be such that the greater the change, the higher the score, such as the rate of change of velocity or acceleration, the rate of change of sensor data, etc., thereby reflecting the degree of change in the state through the score, thus enabling timely synchronization of dynamic information and reducing lag. At the same time, the score of the third component can be determined based on whether the state information has expired or whether the task is urgent (e.g., obstacle avoidance, task switching, etc.), thereby ensuring that time-sensitive state information is transmitted in a timely manner through the score of this dimension.

[0043] In the example, the scores for the first, second, and third components can vary depending on the task scenario faced by the multi-agent system. For instance, in formation flying missions, the score and / or weighting coefficient of state information on the first component, corresponding to importance, can be set higher. Similarly, in emergency obstacle avoidance, the score and / or weighting coefficient of state information on the second component, corresponding to the rate of change, and the score and / or weighting coefficient on the third component, corresponding to timeliness, can be set higher.

[0044] In the example, high-priority status information may include control safety-related information such as pose and velocity; mission-critical commands such as emergency stop, collision warning, and mission switching instructions; and safety broadcasts such as the location of sudden obstacles.

[0045] In step S103, the available bandwidth size for state synchronization is determined.

[0046] In the example, basic network conditions can be recorded during the initialization of the multi-agent system, followed by real-time detection of network parameters to determine the available bandwidth. To this end, a bandwidth detection module can be configured to evaluate the real-time status of the communication link, providing a network status basis for subsequent synchronization strategy selection. For example, the bandwidth detection module can continuously measure the latency, packet loss rate, and throughput performance of the communication link to estimate the available bandwidth and infer whether the current communication link status is ample (referred to as bandwidth-allowed) or limited (referred to as bandwidth-limited) in this paper. To this end, the bandwidth detection module can include a round-trip time (RTT) measurement unit, a packet loss statistics unit, a throughput estimation unit, and a link state classifier. This link state classifier can classify the current communication link status as bandwidth-allowed or bandwidth-limited based on preset threshold rules or machine learning prediction models.

[0047] In step S104, a synchronization strategy for state synchronization is determined based on the available bandwidth. Step S104 further includes steps S1041 and S1042.

[0048] In step S1041, in response to determining that the available bandwidth indicates bandwidth is allowed, a first synchronization strategy is adopted. In the first synchronization strategy, multiple state information items are periodically synchronized according to a preset synchronization period, and at least one target state information item among the multiple state information items is synchronized via an event. The synchronization period for periodically synchronizing the first and second state information items among the multiple state information items is shorter than the synchronization period for the third state information item. The priority score of the at least one target state information item among the multiple state information items is higher than the priority scores of the other state information items.

[0049] In the example, when bandwidth allows, the first synchronization strategy employed is a parallel strategy of periodic synchronization and event-triggered synchronization. In this periodic synchronization, the synchronization periods for the first, second, and third state information can differ, meaning the frequency of state synchronization can vary. The third state information, typically with a larger data volume, can be synchronized at a lower frequency compared to the first and second state information. Periodic synchronization ensures the integrity of state information synchronization. Simultaneously, event-triggered synchronization ensures the immediacy of state information synchronization, thus allowing higher-priority target state information to be synchronized in parallel using this method.

[0050] In the example, the priority scores of multiple status information can be sorted from largest to smallest, thereby enabling event-triggered synchronization of a predetermined number of target status information (such as the top three target status information) that are sorted first.

[0051] In the example, practical application scenarios where bandwidth is available may include unmanned vehicle platooning in indoor or near-field mesh networks, local area networks, Wi-Fi 6 or 5G uplink conditions, as well as outdoor drone performances or platooning with good base station coverage and low packet loss.

[0052] In step S1042, in response to determining that the available bandwidth indicates bandwidth limitation, a second synchronization strategy is adopted. In the second synchronization strategy, event-triggered synchronization is performed only on the at least one target status information.

[0053] In the example, when bandwidth is limited, the synchronization strategy differs from when bandwidth is available. In this case, only target state information with higher priority scores is synchronized using event-triggered synchronization. For example, only the target state information with the highest priority score can be synchronized.

[0054] In the example, real-world application scenarios with limited bandwidth may include remote inspection (weak cellular signal, low backhaul link bandwidth) or large-scale disaster area grid damage, as well as large-scale outdoor robot swarms working in congested wireless environments (Wi-Fi overload) or through walls / at long distances.

[0055] Therefore, in the state synchronization method for intelligent agent clusters according to embodiments of this disclosure, by classifying state information and accurately quantifying its priority, and flexibly adapting different synchronization strategies to different network states, the content, frequency, and method of state synchronization are dynamically adjusted according to network conditions. This achieves an adaptive bandwidth-aware mechanism, thereby improving robustness, communication efficiency, and collaborative performance in multi-agent system scenarios. Even in environments with high packet loss and low bandwidth, the collaborative stability of the intelligent agent cluster can still be maintained.

[0056] In some embodiments, special scenarios may arise in practical applications where tasks of equal priority compete for bandwidth or lower priority tasks are unable to obtain bandwidth. To address these special scenarios, different robust scheduling mechanisms can be introduced, such as scheduling-level anti-collision mechanisms, packet loss tolerance mechanisms, network QoS guarantee mechanisms, and security fallback mechanisms. In scheduling-level anti-collision mechanisms, time-slice priority switching can be used to allow tasks of equal priority to rotate through micro-time slices, preventing a single task from monopolizing bandwidth. In packet loss tolerance mechanisms, elastic degradation can be used, allowing low-risk tasks to downsample or compress to release bandwidth. In network QoS guarantee mechanisms, hierarchical bandwidth allocation can be used, setting bandwidth limits or reservation ratios for different task categories. In security fallback mechanisms, control channel reservation can be used, for example, reserving 5–10% of the link for control or security emergency signals.

[0057] Figure 2 This is a flowchart illustrating the process of determining the available bandwidth size according to an exemplary embodiment.

[0058] like Figure 2 As shown, in some embodiments, the process of determining the available bandwidth size for state synchronization (such as in conjunction with...) Figure 1 The step S103 may include steps S201, S202, S203A and S203B.

[0059] In step S201, the initial available bandwidth is calculated based on network parameters.

[0060] In the example, the bandwidth detection module described above can be used to detect network parameters used to calculate bandwidth size (bps), such as round-trip delay (also known as latency), packet loss rate, and throughput performance.

[0061] In step S202, it is determined whether the calculated initial available bandwidth (bps) is greater than or equal to the ratio of the semantic load size (bps) to the security margin coefficient α. The first to third state information among the multiple state information can each have their own corresponding semantic load size. The security margin coefficient α can take a value between 0 and 1.

[0062] In the example, the semantic payload size to be synchronized can refer to the amount of semantic data that the agent node needs to send. Different state information can have different semantic payload sizes. For example, the semantic payload size corresponding to the state information of a six-dimensional pose is smaller than that corresponding to the state information of a point cloud or video. Therefore, whether the bandwidth status is bandwidth-allowed or bandwidth-limited can be determined based on both the calculated initial available bandwidth size and the semantic payload size to be synchronized. For example, the same available bandwidth of 2 Mbps is sufficient for a system that only transmits six-dimensional poses (i.e., bandwidth-allowed), but limited for a system that needs to transmit point clouds or video (i.e., bandwidth-limited).

[0063] In step S203A, in response to determining that the initial available bandwidth size is greater than or equal to the ratio, the available bandwidth size indicator bandwidth is determined to be allowed.

[0064] In step S203B, in response to determining that the initial available bandwidth size is less than the ratio, the available bandwidth size indicates that the bandwidth is limited.

[0065] By introducing semantic load size to measure bandwidth status, the above method avoids the one-sidedness of judging bandwidth status by looking at a single network indicator, and thus provides higher accuracy for adaptive bandwidth awareness.

[0066] In some embodiments, in response to determining the available bandwidth size to indicate bandwidth limitation, the state synchronization method may further include: discarding a portion of target state information from at least one target state information whose semantic load size does not match its semantic value, so that another portion of the retained target state information can be synchronized based on the available bandwidth size.

[0067] In the example, when bandwidth is limited, the synchronization content has been reduced to include only higher-priority target state information according to the second synchronization strategy. In this case, further content can be discarded from this target state information to better suit scenarios with excessively limited bandwidth. Accordingly, a semantic degradation communication mechanism can be employed, retaining only the more critical state information while discarding some low-semantic-value state information, such as those whose semantic load size does not match their semantic value. For example, in video data, large areas of sky and ground texture have a large semantic load but low semantic value.

[0068] Therefore, semantic degradation communication mechanisms can further optimize the content to be synchronized in bandwidth-constrained scenarios, thereby providing more flexible bandwidth adaptability.

[0069] Figure 3 This is a schematic diagram illustrating a semantic degradation mechanism according to an exemplary embodiment.

[0070] like Figure 3 As shown, the set of state information to be synchronized 301 may include one or more state information sets. The subset of state information 302 contains a large amount of data but has low semantic value. Therefore, in bandwidth-constrained scenarios, the subset of state information 302 can be discarded from the set of state information 301, retaining only the remaining state information 303. This allows for further optimization of the content to be synchronized in bandwidth-constrained scenarios, thus providing greater flexibility to adapt to bandwidth conditions.

[0071] In some embodiments, in addition to the semantic degradation communication mechanism described above, the multi-agent system may also have certain unique mechanisms, including an intent prediction mechanism to replace missing data, an edge-autonomous isolation mode mechanism, a cognitive consistency repair mechanism, and a cooperative compression mechanism. In the intent prediction mechanism to replace missing data, action-based reasoning can be used to continue collaboration on disconnected agent nodes in the multi-agent system. In the edge-autonomous isolation mode mechanism, agent nodes can automatically switch to independent control mode when the network degrades. In the cognitive consistency repair mechanism, agent nodes can perform state re-fusion when communication is restored to avoid instruction conflicts. In the cooperative compression mechanism, only task differences can be transmitted, relying on a shared world model to complete the data, where the world model can refer to the understanding of the external world jointly maintained and shared by all agent nodes in the multi-agent system.

[0072] Figure 4 This is a flowchart illustrating a state synchronization method for an agent cluster according to another exemplary embodiment.

[0073] like Figure 4 As shown, after determining the synchronization strategy in step S401, the state synchronization method may further include steps S402 of encoding and compression, and steps S403 of channel selection and scheduling. Here, step S401 of determining the synchronization strategy can, for example, be a combination of... Figure 1 Step S104.

[0074] In some embodiments, in step S402, at least one of state information encoding or compression may be performed. A temporal differential encoding algorithm may be used in response to the agent cluster performing cooperative operation and manipulation tasks. A sparse matrix compression and hash redundancy tagging algorithm may be used in response to the agent cluster performing cooperative perception tasks.

[0075] In the examples, compression algorithms such as temporal differential coding, hash redundancy tagging, ZSTD, and LZ4 can be used to reduce the data volume. Additionally, sparse matrix coding can be used for sparse sensing data (such as sparse point clouds and semantic masks). Taking multi-robot cooperative handling as an example, the data type of state information is continuous pose trajectory and velocity. However, because the continuous change in pose is small and differential coding is efficient, temporal differential coding can be used. Taking distributed sensing SLAM as another example, the data type of state information is point cloud maps and feature maps. Because point clouds are sparse, repetitive, and highly redundant, sparse matrix compression and hash redundancy tagging can be used.

[0076] In the example, to perform the encoding and compression step S402, a state encoding module can be set up to encode and / or compress the state information generated by the agent nodes, thereby reducing the communication load. This state encoding module may include a time-difference encoder, a sparse matrix encoder, a redundancy detector, and a hash filter, etc. In addition, the state encoding module can also consider hierarchical filtering of state information to ensure that critical information is transmitted preferentially. For this purpose, the state encoding module may also include a state priority evaluation unit, which can calculate priority scores for the input state information based on importance, rate of change, and timeliness, achieving dynamic sorting.

[0077] Therefore, the state synchronization method according to the embodiments of this disclosure can support a variety of encoding and compression mechanisms, thereby helping to reduce communication load.

[0078] In some embodiments, channel selection and scheduling may be performed in step S403. In response to the agent cluster performing formation tasks, the first and second state information may employ User Datagram Protocol (UDP) or a combination of UDP and Transmission Control Protocol (TCP), and the third state information may employ Web Real-Time Communication (WebRTC).

[0079] In the example, different transmission channels such as UDP, TCP, and WebRTC can be selected based on different levels of state information to ensure low-latency and high-reliability transmission of critical state information. Here, the level of the third state information can be lower than that of the first and second state information. Taking the real-time position and obstacle avoidance of UAV formation as an example, pose or velocity state information requires low latency, collision warning state information requires strong arrival guarantee, while environmental awareness state information can be synchronized in a low-frequency or event-triggered manner. Accordingly, for pose or velocity state information, low latency and minimal header overhead are required, so UDP can be used; for collision warning state information, the main channel guarantees low latency, and the backup channel guarantees arrival confirmation, so UDP primary transmission and TCP backup confirmation can be used; for environmental awareness state information (such as video or point cloud), bandwidth and adaptive bitrate are required, so WebRTC multistream can be used.

[0080] In the example, to perform the channel selection and scheduling step S403, a communication interface module can be configured. This communication interface module can select different link protocols for data transmission. For this purpose, the communication interface module may include a multi-channel selector, which can select the optimal channel among UDP, TCP, WebRTC, etc., based on the level of status information. Furthermore, the communication interface module may also include a real-time transmission queue (UDP channel), a reliable transmission queue (TCP channel), and a streaming interface (WebRTC channel).

[0081] Because the optimal transmission channel can be selected based on different levels of state information, robustness and communication efficiency are further improved.

[0082] In some embodiments, when any one of the multiple agent nodes acts as the sending node for state synchronization, the other agent nodes can act as the receiving nodes for state synchronization. The sending node can be configured to generate and send multiple data packets with timestamps based on a synchronization strategy, and the receiving node can be configured to reassemble the multiple data packets based on the timestamps.

[0083] In the example, the sending node can package and send state information data according to a synchronization strategy, and the receiving node can reassemble the state information based on timestamps, and provide ACK feedback or request retransmission when necessary. Therefore, the aforementioned communication interface module can further implement functions such as ACK feedback management and timing reassembly. Specifically, the communication interface module can also include an ACK and retransmission manager and a state timing reassembly unit. The ACK and retransmission manager can manage the retransmission of out-of-order or lost packets based on the timestamp and number of the data packets, and the state timing reassembly unit can reassemble the state information at the receiving end according to the timestamp to ensure the consistency of the task logic.

[0084] Using timestamps can simplify the transmission and reassembly of data packets, thereby improving the real-time performance and robustness of cluster collaboration.

[0085] In some embodiments, the receiving node may also be configured to: determine whether there is a target data packet in the plurality of data packets that exceeds the delay window; and in response to determining that there is a target data packet, discard the target data packet to reassemble the remaining data packets in the plurality of data packets.

[0086] In the example, taking a drone swarm as an example, the network uses UDP broadcast mode. When moving at high speed, data packets may be out of order or delayed. Therefore, state information can be rearranged according to timestamps, and old packets exceeding the delay window (such as 50ms) are discarded to avoid state rollback. Missing packets are not retransmitted but filled in using trajectory extrapolation. This is because, in group collaboration, the latest information is more important than complete information; out-of-order reordering, timeout discarding, and trajectory extrapolation can prevent global oscillations.

[0087] In some embodiments, the first to third state information among the above multiple state information may each have their own corresponding queue layer, such that when the available bandwidth is lower than a preset threshold, the synchronization period of the queue layer corresponding to the third state information is extended, and only the queue layers corresponding to the first and second state information are allowed to transmit, until the available bandwidth recovers to above the preset threshold.

[0088] In the example, taking the requirement for strong consistency in real-time obstacle avoidance of high-speed UAV formations as an example, the queue layer can include a safety layer, a control layer, and a perception layer corresponding to the first, second, and third state information, respectively. The safety layer guarantees priority, the control layer guarantees near real-time performance, and the perception layer can synchronize periodically at low frequency. For example, a threshold for the rate of change of pose or velocity triggers immediate reporting, and a collision threshold is reached for immediate broadcasting. When the available bandwidth decreases below a preset threshold, if the synchronization strategy is still within the bandwidth limit, the synchronization period of the perception layer can be automatically extended (the perception layer can also enter differential or compression mode), and only synchronization of the safety and control layers is allowed.

[0089] This adaptive cycle length backoff mechanism allows for more flexible adaptation to network conditions and enhances adaptive bandwidth awareness.

[0090] Figure 5 This is a schematic diagram illustrating the queue hierarchy of state information according to an exemplary embodiment.

[0091] like Figure 5 As shown, the queue layering can include a security layer 501, a control layer 502, and a perception layer 503. Based on the assessment of the network condition, if it is necessary to further adjust the transmitted content to adapt to the current network condition, the synchronization period of the state information of the perception layer 503 can be extended, while only the state information of the security layer 501 and the control layer 502 is allowed to be transmitted, until the available bandwidth recovers to a more sufficient state.

[0092] In some embodiments, event-triggered synchronization may include threshold triggering caused by the data value of state information reaching a preset threshold, so that event-triggered synchronization satisfies immediacy in the synchronization strategy.

[0093] In some embodiments, threshold triggering may include absolute threshold triggering and relative rate of change triggering.

[0094] In the example, absolute threshold triggering can mean that it is triggered immediately when a physical quantity (such as a collision, electrical charge, etc.) exceeds a threshold. Relative rate of change triggering can mean that it is triggered immediately when the rate of change of a physical quantity (such as a sudden change in pose) exceeds a threshold.

[0095] Since the state synchronization method of this disclosure provides a state change threshold triggering strategy, it can enable the intelligent agent cluster to maintain strong consistency and collision avoidance in high-speed maneuvering or sudden obstacle scenarios, thereby improving the robustness of the system.

[0096] Embodiments of this disclosure also provide a state synchronization device for an intelligent agent cluster.

[0097] Figure 6 This is a schematic block diagram illustrating a state synchronization device 600 for an intelligent agent cluster according to an exemplary embodiment.

[0098] like Figure 6 As shown, in some embodiments, the state synchronization device 600 includes at least an information acquisition module 601, a priority determination module 602, a bandwidth determination module 603, and a strategy determination module 604.

[0099] The information acquisition module 601 is configured to acquire multiple state information of each intelligent agent node, including first state information associated with behavior control, second state information associated with task state, and third state information associated with environmental perception.

[0100] The priority determination module 602 is configured to determine the priority score of each state information based on a weighted sum of a first component, a second component, and a third component. The first component is used to characterize the importance of the state information, the second component is used to characterize the rate of change of the state information, and the third component is used to characterize the timeliness of the state information. The first component, the second component, and the third component all have adjustable weight coefficients.

[0101] The bandwidth determination module 603 is configured to determine the amount of available bandwidth for state synchronization.

[0102] The strategy determination module 604 is configured to determine a synchronization strategy for state synchronization based on the available bandwidth.

[0103] The strategy determination module 604 includes a first strategy module 6041 and a second strategy module 6042.

[0104] The first strategy module 6041 is configured to determine a first synchronization strategy in response to determining the available bandwidth size and indicating that bandwidth is allowed. In the first synchronization strategy, multiple status information are periodically synchronized according to a preset synchronization period, and at least one target status information among the multiple status information is event-triggered synchronized. The synchronization period for periodically synchronizing the first and second status information among the multiple status information is smaller than the synchronization period for the third status information, and the priority score of at least one target status information among the multiple status information is higher than the priority score of the other status information.

[0105] The second strategy module 6042 is configured to determine a second synchronization strategy in response to determining that the available bandwidth size indicates that the bandwidth is limited, in which event-triggered synchronization is performed only for at least one target status information.

[0106] The operations of the aforementioned information acquisition module 601, priority determination module 602, bandwidth determination module 603, and strategy determination module 604 can be combined with... Figure 1 The operations of steps S101, S102, S103, and S104 are the same, and the operations of the first strategy module 6041 and the second strategy module 6042 can be combined. Figure 1 The operations of steps S1041 and S1042 are the same, so the details of each aspect will not be repeated here.

[0107] In some embodiments, the bandwidth determination module 603 may include: a bandwidth calculation module 6031, configured to calculate an initial available bandwidth size based on network parameters; a bandwidth comparison module 6032, configured to determine whether the calculated initial available bandwidth size is greater than or equal to the ratio of the semantic load size to be synchronized to the safety margin coefficient, wherein the first to third state information among the multiple state information have their respective corresponding semantic load sizes, and the safety margin coefficient is between 0 and 1; a first determination module 6033A, configured to determine an available bandwidth size indicating bandwidth allowance in response to determining that the initial available bandwidth size is greater than or equal to the ratio; and a second determination module 6033B, configured to determine an available bandwidth size indicating bandwidth limitation in response to determining that the initial available bandwidth size is less than the ratio.

[0108] In some embodiments, the state synchronization device 600 may further include: a semantic degradation module 605, configured to discard a portion of target state information from at least one target state information that does not match the semantic load size and semantic value in response to determining that the available bandwidth size indicates bandwidth limitation, so that the remaining portion of target state information can be synchronized based on the available bandwidth size.

[0109] In some embodiments, the state synchronization device 600 may further include: an encoding and compression module 606, configured to perform at least one of state information encoding or compression, wherein: in response to the agent cluster performing cooperative operation and manipulation tasks, a time difference encoding algorithm is used; and in response to the agent cluster performing cooperative perception tasks, a sparse matrix compression and hash redundancy tagging algorithm is used.

[0110] In some embodiments, the state synchronization device 600 may further include: a channel selection and scheduling module 607, configured to respond to an intelligent agent cluster performing a formation task, using UDP or a combination of UDP and TCP for the first state information and the second state information, and using real-time web communication for the third state information.

[0111] While specific functions have been discussed above with reference to specific modules, it should be noted that the functions of the modules discussed herein can be divided into multiple modules, and / or at least some functions of multiple modules can be combined into a single module. The specific actions performed by the modules discussed herein include the specific module itself performing the action, or alternatively, the specific module calling or otherwise accessing another component or module that performs the action (or performs the action in conjunction with the specific module). Therefore, a specific module performing an action can include the specific module performing the action itself and / or another module that performs the action, called or otherwise accessed by the specific module.

[0112] It should also be understood that the above regarding Figure 6The described modules can be implemented in hardware or in hardware in combination with software and / or firmware. For example, these modules can be implemented as computer program code / instructions configured to execute in one or more processors and stored in a computer-readable storage medium. Alternatively, these modules can be implemented as hardware logic / circuit. For example, these modules can be implemented together in a System on Chip (SoC). An SoC may include an integrated circuit chip (which includes one or more components in a processor (e.g., a Central Processing Unit (CPU), microcontroller, microprocessor, digital signal processor (DSP), etc.), memory, one or more communication interfaces, and / or other circuitry) and may optionally execute received program code and / or include embedded firmware to perform functions.

[0113] According to one aspect of this disclosure, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory. The processor is configured to execute the computer program to implement the steps of any of the method embodiments described above.

[0114] According to one aspect of this disclosure, a non-transitory computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of any of the method embodiments described above.

[0115] According to one aspect of this disclosure, a computer program product is also provided, which includes a computer program that, when executed by a processor, implements the steps of any of the method embodiments described above.

[0116] In the following text, combined with Figure 7 Illustrative examples describing such computer devices, non-transitory computer-readable storage media, and computer program products.

[0117] Figure 7 An example configuration of a computer device 700 that can be used to implement the methods described herein is shown.

[0118] Computer device 700 can be a variety of different types of devices. Examples of computer device 700 include, but are not limited to: desktop computers, server computers, laptop or netbook computers, mobile devices (e.g., tablet computers, cellular or other wireless phones (e.g., smartphones), notebook computers, mobile stations), wearable devices (e.g., glasses, watches), entertainment devices (e.g., entertainment appliances, set-top boxes communicatively coupled to a display device, game consoles), televisions or other display devices, automotive computers, and so on.

[0119] Computer device 700 may include at least one processor 702, memory 704, multiple communication interfaces 706, display device 708, other input / output (I / O) devices 710, and one or more mass storage devices 712 capable of communicating with each other, such as via system bus 714 or other suitable connections.

[0120] Processor 702 may be a single processing unit or multiple processing units, and all processing units may include single or multiple computing units or multiple cores. Processor 702 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuits, and / or any device that manipulates signals based on operating instructions. Among other capabilities, processor 702 may be configured to acquire and execute computer-readable instructions stored in memory 704, mass storage device 712, or other computer-readable media, such as program code of operating system 716, program code of application program 718, program code of other program 720, etc.

[0121] Memory 704 and mass storage device 712 are examples of computer-readable storage media for storing instructions that are executed by processor 702 to perform the various functions described above. For example, memory 704 can generally include both volatile and non-volatile memory (e.g., RAM, ROM, etc.). Furthermore, mass storage device 712 can generally include hard disk drives, solid-state drives, removable media, including external and removable drives, memory cards, flash memory, floppy disks, optical disks (e.g., CDs, DVDs), storage arrays, network-attached storage, storage area networks, etc. Both memory 704 and mass storage device 712 can be collectively referred to herein as memory or computer-readable storage media, and can be non-transitory media capable of storing computer-readable, processor-executable program instructions as computer program code, which can be executed by processor 702 as a specific machine configured to perform the operations and functions described in the examples herein.

[0122] Multiple programs may be stored on mass storage device 712. These programs include operating system 716, one or more application programs 718, other programs 720, and program data 722, and they may be loaded into memory 704 for execution. Examples of such application programs or program modules may include computer program logic (e.g., computer program code or instructions) of the methods described herein and / or other embodiments described herein.

[0123] Although Figure 7The modules 716, 718, 720, and 722, or portions thereof, are illustrated as being stored in memory 704 of computer device 700; however, modules 716, 718, 720, and 722 may be implemented using any form of computer-readable medium accessible by computer device 700. As used herein, “computer-readable medium” includes at least two types of computer-readable media: computer-readable storage media and communication media.

[0124] Computer-readable storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, DVD, or other optical storage devices, magnetic cassettes, magnetic tapes, disk storage devices or other magnetic storage devices, or any other non-transmission medium that can be used to store information for access by computer devices. In contrast, communication media can embody computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms. Computer-readable storage media as defined herein do not include communication media.

[0125] One or more communication interfaces 706 are used for exchanging data with other devices, such as via a network, direct connection, etc. Such communication interfaces can be one or more of the following: any type of network interface (e.g., a network interface card (NIC)), wired or wireless (such as IEEE 802.11 Wireless LAN (WLAN)) wireless interface, Wi-MAX interface, Ethernet interface, Universal Serial Bus (USB) interface, cellular network interface, Bluetooth. TM Interfaces include near field communication (NFC) interfaces. Communication interface 706 facilitates communication across various network and protocol types, including wired networks (e.g., LAN, cable, etc.) and wireless networks (e.g., WLAN, cellular, satellite, etc.), the Internet, etc. Communication interface 706 can also provide communication with external storage devices (not shown) such as storage arrays, network-attached storage, storage area networks, etc.

[0126] In some examples, a display device 708, such as a monitor, may be included for displaying information and images to the user. Other I / O devices 710 may be devices that receive various inputs from the user and provide various outputs to the user, and may include touch input devices, gesture input devices, cameras, keyboards, remote controls, mice, printers, audio input / output devices, and so on.

[0127] The technologies described herein can be supported by these various configurations of computer device 700, and are not limited to specific examples of the technologies described herein. For example, the functionality can also be implemented wholly or partially on a “cloud” using a distributed system. A cloud includes and / or represents a platform for resources. The platform abstracts the underlying functionality of the cloud’s hardware (e.g., servers) and software resources. Resources may include applications and / or data that can be used when performing computational processing on a server remote from computer device 700. Resources may also include services provided via the Internet and / or via subscriber networks such as cellular or Wi-Fi networks. The platform can abstract resources and functionality to connect computer device 700 to other computer devices. Therefore, the implementation of the functionality described herein can be distributed throughout the cloud. For example, the functionality can be implemented partly on computer device 700 and partly through a platform that abstracts the functionality of the cloud.

Claims

1. A state synchronization method for a cluster of agents, characterized in that, The agent cluster includes a plurality of agent nodes, and the method includes: obtaining a plurality of state information of each agent node, including first state information associated with behavior control, second state information associated with task state, and third state information associated with environment perception; determining a priority score of each state information based on a weighted sum of a first component, a second component and a third component, wherein the first component is used to represent the importance of the state information, the second component is used to represent the change rate of the state information, and the third component is used to represent the timeliness of the state information, and the first component, the second component and the third component all have adjustable weight coefficients; determining the available bandwidth size for state synchronization, including: calculating an initial available bandwidth size based on network parameters; determining whether the calculated initial available bandwidth size is greater than or equal to the ratio of the semantic load size to be synchronized and the safety margin coefficient, wherein the first state information to the third state information in the plurality of state information has a respective corresponding semantic load size, and the safety margin coefficient has a value between 0 and 1; in response to determining that the initial available bandwidth size is greater than or equal to the ratio, determining that the available bandwidth size indicates bandwidth allowed; and in response to determining that the initial available bandwidth size is less than the ratio, determining that the available bandwidth size indicates bandwidth limited; and determining a synchronization strategy for state synchronization based on the available bandwidth size, wherein: in response to determining that the available bandwidth size indicates bandwidth allowed, determining to adopt a first synchronization strategy, in which the plurality of state information is periodically synchronized according to a preset synchronization period, and at least one target state information in the plurality of state information is event-triggered synchronized, wherein the synchronization period for periodically synchronizing the first state information and the second state information in the plurality of state information is smaller than the synchronization period for the third state information, and the priority score of the at least one target state information in the plurality of state information is higher than the priority score of other state information; in response to determining that the available bandwidth size indicates bandwidth limited, determining to adopt a second synchronization strategy, in which only the at least one target state information is event-triggered synchronized.

2. The method of claim 1, wherein, in response to determining that the available bandwidth size indicates bandwidth limited, the method further includes: discarding a part of target state information from the at least one target state information, which does not match the semantic load size and semantic value, so that another part of target state information retained can be synchronized based on the available bandwidth size.

3. The method of claim 1, wherein, The method further includes: after determining the synchronization strategy, performing at least one of state information encoding or compression, wherein: in response to the agent cluster being used to perform cooperative operation and manipulation task, a time difference encoding algorithm is adopted; and in response to the agent cluster being used to perform cooperative perception task, a sparse matrix compression and hash redundancy marking algorithm is adopted.

4. The method of claim 3, wherein, The method further comprises: after performing at least one of state information encoding or compression, performing channel selection and scheduling, wherein: in response to the agent cluster being used to perform formation tasks, adopting a user datagram protocol or a combination of a user datagram protocol and a transmission control protocol for the first state information and the second state information, and adopting a web real-time communication for the third state information.

5. The method of claim 1, wherein, When any agent node in the plurality of agent nodes is a sending node for state synchronization, other agent nodes are receiving nodes for state synchronization, wherein: the sending node is configured to generate and send a plurality of data packets with timestamps based on the synchronization strategy; and the receiving node is configured to reorganize the plurality of data packets based on the timestamps.

6. The method of claim 5, wherein, The receiving node is further configured to: determine whether there is a target data packet in the plurality of data packets that exceeds a delay window; and in response to determining that there is the target data packet, discard the target data packet to reorganize the remaining data packets in the plurality of data packets.

7. The method of claim 1, wherein, The first state information to the third state information in the plurality of state information each have a corresponding queue hierarchy, such that when the available bandwidth size is below a preset threshold, the synchronization period of the queue hierarchy corresponding to the third state information is extended, and only the queue hierarchies corresponding to the first state information and the second state information are allowed to be transmitted until the available bandwidth size recovers above the preset threshold.

8. The method of claim 1, wherein, The event-triggered synchronization includes a threshold trigger caused by the data value of the state information reaching a preset threshold, such that in the synchronization strategy, the event-triggered synchronization satisfies instantaneity.

9. The method of claim 8, wherein, The threshold trigger includes an absolute threshold trigger and a relative change rate trigger.

10. A state synchronizing apparatus for a cluster of intelligent agents, characterized by, The agent cluster includes a plurality of agent nodes, and the device includes: an information acquisition module configured to acquire a plurality of state information of each agent node, including first state information associated with behavior control, second state information associated with task state, and third state information associated with environment perception; a priority determination module configured to determine a priority score of each state information based on a weighted sum of a first component, a second component, and a third component, wherein the first component is used to represent the importance of the state information, the second component is used to represent the change rate of the state information, and the third component is used to represent the timeliness of the state information, and the first component, the second component, and the third component each have an adjustable weight coefficient; a bandwidth determination module configured to determine an available bandwidth size for state synchronization, including: a bandwidth calculation module configured to calculate an initial available bandwidth size based on network parameters; a bandwidth comparison module configured to determine whether the calculated initial available bandwidth size is greater than or equal to the ratio of a semantic load size and a safety margin coefficient, wherein the first state information to the third state information in the plurality of state information each have a corresponding semantic load size, and the safety margin coefficient has a value between 0 and 1; and a bandwidth adjustment module configured to adjust the available bandwidth size based on the determination result. The first determining module is configured to determine that the available bandwidth size indicates bandwidth permission in response to judging that the initial available bandwidth size is greater than or equal to the ratio; and The second determining module is configured to determine that the available bandwidth size indicates bandwidth limitation in response to judging that the initial available bandwidth size is less than the ratio; and The policy determining module is configured to determine a synchronization policy for state synchronization based on the available bandwidth size, wherein: The first policy module is configured to determine to adopt a first synchronization policy in response to determining that the available bandwidth size indicates bandwidth permission, in which the first synchronization policy is to periodically synchronize the plurality of state information according to a preset synchronization period and to event-triggered synchronize at least one target state information in the plurality of state information, wherein the synchronization period for periodically synchronizing the first state information and the second state information in the plurality of state information is smaller than the synchronization period for the third state information, and the priority score of the at least one target state information in the plurality of state information is higher than the priority score of other state information; The second policy module is configured to determine to adopt a second synchronization policy in response to determining that the available bandwidth size indicates bandwidth limitation, in which the second synchronization policy is to only event-triggered synchronize the at least one target state information.

11. A computer device, comprising: The computer device comprises: at least one processor; a memory having a computer program stored thereon, wherein the computer program, when executed by the at least one processor, causes the at least one processor to perform the method of any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer readable storage medium has a computer program stored thereon, and the computer program, when executed by a processor, causes the processor to perform the method of any one of claims 1-9.

13. A computer program product, characterised in that, The computer program product comprises a computer program, and the computer program, when executed by a processor, causes the processor to perform the method of any one of claims 1-9.

Images