A method and system for ultra-low latency coordinated communication

Abstract

This invention relates to the field of wireless communication technology, providing an ultra-low latency collaborative communication method and system. The method includes: constructing an end-to-end network slicing strategy model based on service level protocols, comprising a wireless access network, a bearer network, and a core network; fusing heterogeneous sensing data from vehicle-mounted and roadside systems to construct surrounding environment situational awareness data and generate uplink data packets with service type identifiers; making dynamic decisions based on the service type identifiers in the uplink data packets, triggering a collaborative scheduling logic mechanism, completing end-to-end transmission in each network domain, and outputting multi-access edge computing data packets; performing real-time analysis of the multi-access edge computing data packets, generating final control commands or early warning information, and completing closed-loop distribution via the downlink. This invention collaboratively integrates end-to-end network resource hard isolation with cloud-edge-device integration to construct a full-process deterministic service system from multi-source sensing, guaranteed communication, collaborative decision-making to closed-loop control.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to an ultra-low latency cooperative communication method and system. Background Technology

[0002] With the rapid development of vehicle-to-everything (V2X) and autonomous driving technologies in fields such as smart transportation, logistics, and public safety, end-to-end ultra-low latency and high-reliability communication have become core technologies for ensuring driving safety and traffic efficiency. How to provide deterministic quality of service (QoS) guarantees for latency-sensitive services such as autonomous driving and remote control in complex communication environments spanning multiple network domains has become a crucial issue that urgently needs to be addressed in the deep integration of 5G networks with V2X applications.

[0003] Chinese patent application CN120321793A discloses a VDSL ultra-low latency communication method and system. The method includes: encrypting the clock synchronization channel using a quantum key distribution protocol; dynamically allocating micro-time slot resources within the symbol period of orthogonal frequency division multiplexing (OFDM) to generate dynamically adjusted micro-time slot resource allocation results; calculating the optimal phase offset matrix of the metasurface intelligent reflector using a deep deterministic policy gradient algorithm to obtain an optimized electromagnetic wave propagation path; generating a globally optimized verification matrix by aggregating the gradients of lightweight error correction models trained locally at each node to obtain a compensated data stream; constructing a causal graph model to dynamically prune high-entropy paths to minimize causal entropy; and using multi-agent reinforcement learning with latency-energy efficiency as the game objective to determine the optimal modulation order and subcarrier switching strategy to obtain an optimized stable communication link.

[0004] However, current technologies still face many challenges. Vehicle-to-everything (V2X) communication paths span multiple interconnected network domains, including wireless access networks, bearer networks, and core network processing. However, existing technologies are often limited to localized optimization of a single link, thus lacking a mechanism to ensure end-to-end coordination between network domain resources and application layer requirements. This results in latency bottlenecks potentially appearing randomly in unoptimized links such as the bearer network and core network, preventing safety-critical services like remote driving and fleet collaboration from obtaining deterministic latency guarantees throughout the entire process. Consequently, this leads to significant driving safety risks and obstacles to industry implementation. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides an ultra-low latency cooperative communication method, the specific technical solution of which is as follows:

[0006] Based on business level agreement The key fields are used to generate slice configuration data for three network domains: wireless access network, bearer network, and core network. This slice configuration data is then integrated to construct an end-to-end slice strategy model P. E2E Call the end-to-end consistency verification function Verify(ID)i For the end-to-end slicing strategy model P E2E Perform bandwidth continuity and latency compliance checks. If the checks pass, then the end-to-end slicing strategy model P is updated. E2E Activate if enabled, otherwise reject and trigger an alarm mechanism;

[0007] Collect vehicle status data and roadside environmental data Integrating the vehicle status data and roadside environmental data Construct surrounding environment situational awareness data, determine the service payload based on the surrounding environment situational awareness data, and generate a service type identifier based on the service payload. The network protocol header, the service payload, and the accompanying service type identifier are included. The metadata (Matadata) is encapsulated to form the uplink data packet P. Uplink ;

[0008] Based on uplink data packet P Uplink Business type identifier in In the end-to-end slicing strategy model P E2E Retrieve slice identifier ID i Quality of Service Level Identifier (QCI) i Encapsulate and generate data packets with tagged slicing strategies. Data packets according to the labeled slice policy Trigger the preset collaborative scheduling logic mechanism to generate scheduling decision M sched With air interface resource allocation R alloc Based on the aforementioned scheduling decision M sched With air interface resource allocation R alloc The data packets that drive the tagged slice strategy End-to-end transmission is completed sequentially in the wireless access network, bearer network, and core network to generate and output multi-access edge computing data packets.

[0009] Real-time analysis of multi-access edge computing data packets Generate local decision results R local and global analysis data D global Based on the global analysis data D global A digital twin model of traffic is constructed, a large-scale machine learning model is used for prediction, and an operations research optimization algorithm engine is activated for simulation and deduction, which is then encapsulated into a global optimization strategy S. global Based on the preset fusion logic, the local decision result R is fused. local and global optimization strategy Sglobal Generate the final control command or warning information C final And send it to the target vehicle or roadside facility.

[0010] Furthermore, the end-to-end slicing strategy model P E2E The construction methods include:

[0011] Based on business level agreement Maximum delay in and business priorities Determine the Quality of Service Level Identifier (QCI) i In conjunction with the aforementioned service level agreement Minimum guaranteed broadband Calculate the number of physical resource blocks reserved using the modulation and coding scheme. This data is then integrated, generated, and distributed to the base station as radio access network layer slice configuration data C. RAN ;

[0012] Based on wireless access network layer slice configuration data C RAN Physical resource block reservation quantity Calculate the actual bandwidth configuration in the wireless access network And based on the actual configured broadband The number of flexible Ethernet slots required for broadband transmission is calculated. This further confirms the guaranteed broadband in the bearer network. After completing the aforementioned broadband guarantee The sum of the physical link's carrying capacity BW PHY After the validity verification, the bearer network slice configuration data C for implementing resource hard isolation is generated. TN ;

[0013] Based on bearer network slice configuration data C TN Business Level Agreement Association resolution, construction of path service quality strategy set Bound virtual forwarding instance The path service quality policy set is configured through the session management function in the control plane. and virtual forwarding instances Encapsulated as N4 session rules, and distributed to the edge user plane function via the N4 interface, integrating and generating core network slice configuration data C. CORE ;

[0014] Integrating Wireless Access Network Layer Slicing Configuration Data C RAN , Bearer network slice configuration data C TN and core network slice configuration data C CORE To construct an end-to-end slicing strategy model P E2E.

[0015] Furthermore, the uplink data packet P Uplink The methods of constructing include:

[0016] Utilizing the multimodal sensor system mounted on the vehicle terminal, raw sensor data streams are acquired in real time. The raw sensor data stream is processed by the on-board unit. Real-time processing is performed to generate structured vehicle status data.

[0017] Raw environmental data streams are collected using roadside sensing devices that are permanently deployed on road infrastructure. Combine deep learning models to analyze the original environmental data stream Real-time deep analysis is performed to generate roadside environmental data.

[0018] Integrating vehicle status data and roadside environmental data Construct surrounding environment situational awareness data, generate a service payload based on the surrounding environment situational awareness data, identify the service type of the service payload, and determine the service type identifier. The network protocol header, the service payload, and the identifier carrying the service type are included. The metadata (Matadata) is encapsulated to form the uplink data packet P to be sent. Uplink .

[0019] Furthermore, the multi-access edge computing data packet The generation methods include:

[0020] Parse the received uplink data packet P Uplink Extract business type identifier and the business type identifier As a query index, P in the end-to-end slicing strategy model E2E Retrieve slice identifier ID i and the slice identifier ID i Related Service Quality Level Identifier (QCI) i The slice identifier ID i With Service Quality Level Identifier (QCI) i Appended to uplink data packet P Uplink The above encapsulates and generates data packets with tagged slicing strategies.

[0021] Data packets based on the tagged slice policy Quality of Service Level Identifier (QCI) i Triggered according to the preset collaborative scheduling logic mechanism, a scheduling decision M is generated. sched With air interface resource allocation R alloc If the Quality of Service Level Identifier (QCI) i For the highest service priority, the Grant-Free scheduling mechanism is activated; if the Quality of Service Level Identifier (QCI) is... i When the service is a latency-sensitive service and not the highest priority service, the intelligent uplink pre-scheduling mechanism is activated; if the Quality of Service (QCI) identifier is... i When the service is not time-sensitive and is not the highest priority service, the standard scheduling mechanism is activated.

[0022] Based on scheduling decision M sched and air interface resource allocation R alloc The vehicle-mounted terminal accesses the reserved physical resource blocks in the wireless network and transmits data packets marked with the slicing policy. Transmitted to the base station; the data packet In the bearer network, based on the slice identifier ID it carries... i Routed to a dedicated, flexible Ethernet hard pipe, and directed to a virtual forwarding instance on the user plane of the core network. Generate and output multi-access edge computing data packets

[0023] Furthermore, the final control command or warning information C final The generation methods include:

[0024] Data packets based on multi-access edge computing The local real-time analysis engine is invoked to analyze the vehicle status data in the data packet payload. and roadside environmental data Perform in-depth analysis and output local decision results R. local and global analysis data D global ;

[0025] Based on global analysis data D global A traffic digital twin model is constructed, and a large-scale machine learning model is used to analyze the traffic digital twin model to predict potential macro-level traffic problems. Based on the traffic problems, an operations research optimization algorithm engine is activated to solve for the optimal strategy and encapsulate it into a global optimization strategy S. global ;

[0026] Integrating local decision results R local and global optimization strategy S global Generate the final control command or warning information C finalAnd the final control command or warning information C final The data is sent to the target vehicle or roadside facility. The fusion process follows a preset fusion logic. If the local decision result R... local When the priority is critical, the local decision result R is adopted first. local Conversely, using the global optimization strategy S... global This serves as a basis for decision-making.

[0027] An ultra-low latency cooperative communication system is provided for implementing the aforementioned ultra-low latency cooperative communication method, comprising a network domain slicing configuration module, a vehicle and road test data fusion module, an uplink data transmission module, and a decision and closed-loop control module.

[0028] The network domain slicing configuration module is based on the service level protocol. The key fields are used to generate slice configuration data for three network domains: wireless access network, bearer network, and core network. This slice configuration data is then integrated to construct an end-to-end slice strategy model P. E2E Call the end-to-end consistency verification function Verify(ID) i For the end-to-end slicing strategy model P E2E Perform bandwidth continuity and latency compliance checks. If the checks pass, then the end-to-end slicing strategy model P is updated. E2E Activate if enabled, otherwise reject and trigger an alarm mechanism;

[0029] The vehicle-mounted and roadside data fusion module is used to collect vehicle status data. and roadside environmental data Integrating the vehicle status data and roadside environmental data Construct surrounding environment situational awareness data, determine the service payload based on the surrounding environment situational awareness data, and generate a service type identifier based on the service payload. The network protocol header, the service payload, and the accompanying service type identifier are included. The metadata (Matadata) is encapsulated to form the uplink data packet P. Uplink ;

[0030] The uplink data transmission module is based on the uplink data packet P. Uplink Business type identifier in In the end-to-end slicing strategy model P E2E Retrieve slice identifier ID i Quality of Service Level Identifier (QCI) i Encapsulate and generate data packets with tagged slicing strategies. Data packets according to the labeled slice policy Trigger the preset collaborative scheduling logic mechanism to generate scheduling decision M sched With air interface resource allocation R alloc Based on the aforementioned scheduling decision M sched With air interface resource allocation R alloc The data packets that drive the tagged slice strategy End-to-end transmission is completed sequentially in the wireless access network, bearer network, and core network to generate and output multi-access edge computing data packets.

[0031] The decision-making and closed-loop control module analyzes multi-access edge computing data packets in real time. Generate local decision results R local and global analysis data D global Based on the global analysis data D global A digital twin model of traffic is constructed, a large-scale machine learning model is used for prediction, and an operations research optimization algorithm engine is activated for simulation and deduction, which is then encapsulated into a global optimization strategy S. global Based on the preset fusion logic, the local decision result R is fused. local and global optimization strategy S global Generate the final control command or warning information C final And send it to the target vehicle or roadside facility.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0033] This application uses the Service Class Protocol (SCP) as a unified planning basis throughout the wireless access network, bearer network, and core network, and coordinates resource reservation and policy solidification for end-to-end paths. This avoids the problem of random performance bottlenecks caused by the lack of coordinated guarantees in intermediate network links, such as the bearer network and core network.

[0034] This application uses heterogeneous sensing data from vehicles and roadsides as the basis for decision-making, and completes the binding of service intent determination and network slice identifier at the data packet generation stage, thereby avoiding the problem of transmission resource mismatch or scheduling decision delay caused by network nodes being unable to perceive the true intent of uplink services.

[0035] This application uses the Quality of Service (QoS) level identifier of data packets as the core influencing factor to make dynamic and adaptive decisions and selections for the uplink air interface scheduling mechanism, thereby avoiding the contradiction between ensuring ultra-low latency and wireless resource utilization caused by adopting a single fixed scheduling strategy.

[0036] This application uses local decision-making results representing immediate safety and global optimization strategies representing macro-efficiency as inputs for collaborative decision-making. It then integrates and arbitrates the final control commands based on the principle of safety priority, thus avoiding the problem of accidentally overriding or delaying critical local commands that ensure immediate driving safety in pursuit of global strategies for macro-efficiency of the system. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating the principle of an ultra-low latency cooperative communication method according to the present invention.

[0039] Figure 2 This is a functional block diagram of an ultra-low latency collaborative communication system according to the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Example 1

[0042] Please see Figure 1 As shown, this embodiment provides an ultra-low latency cooperative communication method, including:

[0043] S1000, based on service level agreement The key fields are used to generate slice configuration data for three network domains: wireless access network, bearer network, and core network. This slice configuration data is then integrated to construct an end-to-end slice strategy model P. E2E Call the end-to-end consistency verification function Verify(ID) i For the end-to-end slicing strategy model P E2E Perform bandwidth continuity and latency compliance checks. If the checks pass, then the end-to-end slicing strategy model P is updated. E2E If activated, it will be rejected and an alarm mechanism will be triggered.

[0044] Specifically, this step aims to pre-plan global network resources and configure policies based on the Service Level Agreements (SLAs) of different vehicle-to-everything (V2X) services, thereby building a stable and reliable end-to-end isolated network architecture for subsequent real-time data transmission.

[0045] Further, step S1000 includes:

[0046] Step S1100, based on the service level agreement Maximum delay in and business priorities Determine the Quality of Service Level Identifier (QCI) i In conjunction with the aforementioned service level agreement Minimum guaranteed broadband And the Modulation and Coding Scheme (MCS) calculates the Physical Resource Block (PRB) reservation quantity. This data is then integrated, generated, and distributed to the base station gNodeB. RAN .

[0047] Specifically, this step aims to transform abstract Service Level Agreement (SLA) requirements into concrete and executable radio resource configurations at the Wireless Access Network (RAN) level. By reserving independent Physical Resource Blocks (PRBs) for different service slices and binding them with differentiated Quality of Service (QoS) policies, it fundamentally avoids latency jitter and data packet loss caused by radio interface resource contention, providing a resource foundation for subsequent low-latency transmission.

[0048] In the specific implementation process, firstly, the business level protocol from the actual application scenario is obtained and parsed. The key fields. Where i∈{1,2,...,N} represents the index of the network slice, and a total of N network slices are planned; SLA i Indicates the Service Level Protocol (SLP) of the i-th network slice; ID i A unique identifier representing the i-th network slice; Represents the category label of the i-th network slice; Represents the minimum guaranteed bandwidth of the i-th network slice, in bits per second (bps). This represents the maximum latency of the i-th network slice. It is a hard indicator strictly limited by physical and security boundaries and is a key threshold for ensuring driving safety. This value is generally set to 5ms through extensive testing and optimization. The unit is milliseconds (ms). This represents the reliability requirement of the i-th network slice, which can be quantified as the maximum acceptable packet loss rate to ensure the integrity of data transmission. This represents the service priority of the i-th network slice, which is classified according to the importance and urgency of the service functions and is used for scheduling decisions when resources are scarce.

[0049] The business level agreement It is a technology contract jointly negotiated and developed by vehicle-to-everything (V2X) application service providers (such as autonomous driving solution companies and smart traffic management departments) and 5G network operators. It aims to accurately describe the specific requirements of target applications (such as remote-controlled sweepers and fleet collaborative driving) for communication services under different working conditions.

[0050] Next, based on the service level agreement Maximum delay and business priorities Construct a Quality of Service (QoS) policy. This QoS policy is used to translate the business logic of upper-layer applications into quantifiable metrics that the underlying network can understand and execute, serving as a key bridge connecting upper-layer business logic and underlying network resource configuration. The QoS policy is typically constructed in the form of a pre-defined decision matrix or lookup table, and the maximum latency is considered. and business priorities As input parameters, the output is the best-matching 5G Quality of Service Level Identifier (QCI). i (QoS classidentifier, QCI) and the corresponding air interface scheduling policy.

[0051] Subsequently, the Physical Resource Block (PRB) requirement calculation model was introduced, and the business level agreement was incorporated. Minimum guaranteed broadband Mapping to physical layer resource reservation. This process considers not only the bandwidth requirements of the service, but also the system configuration capabilities of the base station (next Generation Node B, gNodeB) and the dynamic characteristics of the wireless channel. The specific process formula is as follows:

[0052]

[0053] in, This represents the number of Physical Resource Blocks (PRBs) reserved for the i-th network slice, used to build a dedicated resource pool with capacity isolation at the physical layer of the 5G base station, in units of PRBs. BW represents the minimum guaranteed bandwidth requirement for the i-th network slice, used to ensure stable operation of services with minimal communication capacity; PRBThis represents the frequency domain bandwidth of a single Physical Resource Block (PRB), used to decompose the total bandwidth requirement into the number of standard-sized frequency units required. This value is strictly defined by the 5G technical standards developed by the 3rd Generation Partnership Project (3GPP), and the unit is Hertz (Hz). sym This represents the number of Orthogonal Frequency Division Multiplexing (OFDM) symbols used for data transmission within a single scheduling period. This value is a parameter configured at the base station side and depends on the 5G frame structure and time slot format adopted by the operator; the unit is symbols. mod This represents the spectral efficiency corresponding to the modulation and coding scheme (MCS). It is used to ensure the realism of resource allocation and prevent insufficient resource allocation and communication performance degradation caused by overly idealistic evaluations. This value is based on the service level agreement. Reliability requirements Dynamically selected from a standardized MCS table, in bits per second (bps / Hz); This represents the floor function, used to ensure that the allocated physical resources can meet the lower bandwidth requirement under any circumstances.

[0054] Finally, the QoS policy modeling results and the Physical Resource Block (PRB) quantization results are integrated to generate a structured RAN layer slice configuration data for the Radio Access Network. Among them, QCI i This represents the Quality of Service (QoS) level identifier for the i-th network slice. Slice configuration data C RAN The control interface sends the data to the base station gNodeB and activates it. The base station gNodeB strictly follows the received slice configuration data C. RAN By reserving resources, different network slices can be accurately mapped to corresponding, isolated wireless resource areas during actual data transmission, thereby providing technical support and physical foundation for latency control and service assurance of differentiated services.

[0055] Step S1200, based on the Radio Access Network RAN ​​layer slice configuration data C RAN Physical Resource Block (PRB) reservation quantity Calculate the actual bandwidth configuration in the Radio Access Network (RAN) And based on the actual configured broadband Calculate the number of Flexible Ethernet (FlexE) slots required for broadband transmission. This further confirms the guaranteed broadband in the bearer network TN. After completing the aforementioned broadband guarantee The sum of the physical link PHY's carrying capacity BW PHY After the validity verification, the bearer network TN slice configuration data C for implementing resource hard isolation is generated. TN .

[0056] Specifically, this step aims to configure the radio access network RAN ​​layer slice based on the data in step S1100. RAN By leveraging Flexible Ethernet (FlexE) technology at the Physical Layer (PHY) of the Transport Network (TN), a hard-pipe model with deterministic latency and bandwidth guarantees is created for each network slice to achieve strict isolation of services between slices. This hard-pipe model divides high-speed physical interfaces into multiple standardized time slot units and combines these time slots as needed according to service requirements to construct logical clients for different network slices, achieving fine-grained scheduling and physical isolation of bandwidth resources.

[0057] In the specific implementation process, firstly, based on the RAN layer slice configuration data C of the radio access network... RAN Physical Resource Block (PRB) reservation quantity for the i-th network slice Calculate the actual configured bandwidth corresponding to these physical resource blocks in the Radio Access Network (RAN). Instead of directly using the business level agreement The original minimum guaranteed bandwidth The specific process is as follows:

[0058]

[0059] in, This indicates that the RAN layer of the radio access network actually configures the bandwidth for the i-th network slice to ensure that the capacity of the bearer network TN accurately matches the actual egress capacity of the radio side, thus avoiding the creation of new network bottlenecks. The unit is bits per second (bps).

[0060] Next, based on the actual broadband allocated by the RAN layer of the wireless access network. The specific process for calculating the minimum integer number of FlexE slots required to satisfy this bandwidth for transmission is as follows:

[0061]

[0062] in, BW represents the number of FlexE Ethernet slots allocated to the i-th network slice. TSThe standard bandwidth representing a single FlexE timeslot is the basic bandwidth unit in FlexE technology, defined by the international standard for FlexE technology, and is measured in bits per second (bps). This represents the floor function, used to ensure that the total allocated time slot resources are sufficient to carry all traffic transmitted from the upper layer, thereby preventing the bearer network (TN) from becoming a new performance bottleneck.

[0063] Due to the number of flexible Ethernet FlexE slots It is the integer obtained by rounding up, which provides the total capacity. It is usually slightly larger than the bandwidth actually allocated by the RAN layer. Therefore, it is necessary to base the number of FlexE time slots on the flexible Ethernet network. Calculate the final guaranteed bandwidth for FlexE clients formed by these Flexible Ethernet FlexE time slots in the bearer network TN. The specific process formula is as follows:

[0064]

[0065] in, This represents the guaranteed bandwidth allocated to the i-th network slice and actually effective in the bearer network TN, in bits per second (bps).

[0066] After planning the guaranteed bandwidth for all slices, a validity check is performed to ensure that the total planned bandwidth does not exceed the physical link PHY's capacity limit (BW). PHY The specific verification process satisfies the following constraints:

[0067]

[0068] Among them, BW PHY This represents the total bandwidth of the PHY in the TN physical link of the bearer network. It is a hardware specification parameter of a device and is obtained from the asset management information of the network device.

[0069] Finally, after passing the legality verification, the calculated and planned information is integrated to generate a structured TN slice configuration data for the bearer network. So that it can be distributed to relevant network devices for execution.

[0070] Step S1300, based on the TN slice configuration data C of the bearer network TN Business Level Agreement Association resolution, construction of path service quality strategy set Bound virtual forwarding instance The path service quality policy set is implemented through the Session Management Function (SMF) in the Control Plane Function (CPF). and virtual forwarding instances Encapsulated as N4 session rules, and distributed to the edge user plane function UPF via the N4 interface, integrating to generate core network CN slice configuration data C. CORE .

[0071] Specifically, this step aims to physically deploy the User Plane Function (UPF) at the network edge, configuring dedicated data forwarding paths and processing resources for each network slice. The UPF, as an independent functional entity responsible for user plane data processing within the 5G core network (CN), is deployed at the edge, which is a crucial technical means to achieve end-to-end low latency.

[0072] In the specific implementation process, firstly, the configuration data output from the preceding steps and the original service requirements are strategically correlated and parsed to lay the foundation for the subsequent construction of the user plane forwarding model. The system receives the bearer network TN slice configuration data C output from step S1200. TN This data provides the core basis for traffic identification and link anchoring, enabling the User Plane Function (UPF) to accurately identify data streams transmitted through specific Flexible Ethernet (FlexE) hard pipes and anchor them to their respective slice identity IDs. i Simultaneously, the system associates this identification result with the Service Level Agreement obtained in step S1100. The service level agreement provides a basis for formulating differentiated strategies. When the User Plane Function (UPF) configures data C through the bearer network TN slice... TN Identify the slice identity ID of slice i i Then, the service level agreement of the slice can be combined. To determine the specific processing strategy. Therefore, the TN slice configuration data C of the bearer network. TN With Business Class Agreement The synergy between these two elements is crucial for ensuring that the User Plane Function (UPF) processes traffic correctly and accurately. Together, they constitute a prerequisite for the execution of differentiated user plane strategies and are a key prerequisite for achieving end-to-end Quality of Service (QoS) assurance.

[0073] Secondly, a QoS policy model based on virtual forwarding instances is constructed. To achieve logical isolation between different slices of the core network (CN) user plane, a logically independent virtual forwarding instance is created for each slice i on the deployed user plane function (UPF). (Virtual Forwarding Instance, VFI). The virtual forwarding instance can be considered a virtual router dedicated to slice i. Based on each virtual forwarding instance... Configure differentiated processing and forwarding rule sets The specific formula is as follows:

[0074]

[0075] in, This represents the path service quality policy set for the i-th network slice, used to encapsulate the complete rules that the User Plane Function (UPF) must follow when processing the data flow of slice i. This represents the service priority of the i-th network slice, and this value is directly inherited from the service level protocol. Business priorities defined in This represents the target processing latency for the i-th network slice in the user plane function UPF, based on the maximum end-to-end latency. The decomposition is performed across multiple network segments, including the Radio Access Network (RAN), the Bearer Network (TN), and the Core Network (CN), with the unit being milliseconds (ms). This represents the specific forwarding rule for the i-th network slice, used to determine whether service traffic is offloaded locally or sent to the central cloud. This value is based on the service level agreement. Chinese category tags To be determined.

[0076] Furthermore, after the QoS policy model is constructed, the abstract QoS policy model is transformed into underlying forwarding rules executable by the User Plane Function (UPF) through the interaction process between the control plane and the user plane, thus completing the actual deployment of the final policy. The Session Management Function (SMF) in the Control Plane Function (CPF) obtains the QoS policy model (i.e., the virtual forwarding instance for each network slice i). and the associated path service quality policy set These N4 session rules are encapsulated into a set of standard N4 session rules. These N4 session rules are then transmitted to the User Plane Function (UPF) via the N4 interface. The N4 interface is a standardized interface in the 5G core network (CN) specifically used for session interaction between the Session Management Function (SMF) and the User Plane UPF, and it carries the Packet Forwarding Control Protocol (PFCP) communication protocol. After receiving and parsing the N4 session rules, the UPF creates virtual forwarding instances on its data plane. The corresponding logical resources are loaded, and the path service quality policy set is also loaded. Defined QoS processing and forwarding strategies. At this point, the User Plane Function (UPF) possesses the ability to automate and differentiate the processing of data streams from different slices.

[0077] Finally, the above information is integrated to generate structured core network (CN) slice configuration data. This lays the foundation for subsequent steps.

[0078] Step S1400: Integrate the Radio Access Network (RAN) layer slice configuration data C RAN TN slice configuration data C TN and core network CN slicing configuration data C CORE To construct an end-to-end slicing strategy model P E2E and call the end-to-end consistency verification function Verify(ID) i From the two dimensions of bandwidth continuity and latency compliance, the end-to-end slicing strategy model P is analyzed. E2E Perform a verification; if the end-to-end slicing strategy model P... E2E If the verification is successful, the configuration baseline is persisted and activated; otherwise, the end-to-end slicing strategy model P is rejected. E2E And trigger the alarm mechanism.

[0079] Specifically, this step aims to systematically integrate, policy-bind, and verify the discrete slice configuration data generated for the Radio Access Network (RAN), Bearer Network (TN), and Core Network (CN) in the previous steps, ensuring the integrity and coordination of end-to-end slice resource configuration.

[0080] In the specific implementation process, firstly, based on the Radio Access Network RAN ​​layer slice configuration data C generated in step S1100... RAN Step S1200 generates the TN slice configuration data C. TN And the core network CN slice configuration data C generated in step S1300 CORE Integrate.

[0081] Secondly, after completing the cross-domain configuration data integration, based on the slice identifier ID i By binding the configuration data of the same service slice in different network domains, an end-to-end slicing strategy model P with a globally consistent perspective is constructed. E2E Its data structure is as follows:

[0082] P E2E ={(ID) i C RAN C TN C CORE |i = 1, 2, ..., N};

[0083] Next, in the strategy model P E2E Before being finally solidified, the system needs to call the end-to-end consistency verification function Verify(ID). iThe verification function aims to examine the synergy and rationality of cross-domain resource allocation from two key dimensions: bandwidth continuity and latency compliance. Its verification logic is expressed as follows:

[0084]

[0085] Among them, Verify(ID) i ) represents a Boolean logic function for end-to-end consistency verification of the i-th network slice, used to determine whether the configuration of the i-th network slice meets the end-to-end consistency requirements; BW(·) represents the bandwidth continuity check, used to check that the path capacity of the data stream will not encounter a bottleneck when it is forwarded in the network; BW(·) represents the bandwidth extraction function, used to read and return the guaranteed bandwidth value actually configured for the network domain as a slice, in bits per second (bps). and These represent the guaranteed bandwidth values ​​read and returned for the Radio Access Network (RAN) and the Transport Network (TN), respectively, in bits per second (bps); ∧ represents the logical AND operator, used to connect the two independent logical judgments of bandwidth verification and latency verification. This indicates latency compliance verification, used to ensure that the end-to-end performance commitments of the slice are fulfilled; T(·) represents the estimated end-to-end cumulative latency in milliseconds (ms); T(·) represents the segmented latency estimation function, used to estimate the latency values ​​of different network domains in milliseconds (ms). as well as These represent the estimated latency values ​​for the Radio Access Network (RAN), the Transport Network (TN), and the Core Network (CN), respectively, in milliseconds (ms). This represents the maximum latency of the i-th network slice.

[0086] Finally, after passing end-to-end consistency verification, the system will perform corresponding operations based on the verification results. If the end-to-end slicing strategy model P... E2E If the configuration data in the configuration file is verified, the system recognizes the slicing strategy model P. E2E For a policy to be legitimate and deployable, it must be submitted and persistently stored in the network controller or policy repository as the authoritative configuration baseline guiding network operation and officially enter the active state; conversely, if the end-to-end slicing policy model P... E2E If the configured data fails validation, the system rejects the slice strategy model P. E2E Simultaneously triggering an alarm mechanism to prompt network planners to make necessary adjustments and replanning.

[0087] S2000, collects vehicle status data and roadside environmental data Integrating the vehicle status data and roadside environmental data Construct surrounding environment situational awareness data, determine the service payload based on the surrounding environment situational awareness data, and generate a service type identifier based on the service payload. The network protocol header, the service payload, and the accompanying service type identifier are included. The metadata (Matadata) is encapsulated to form the uplink data packet P. Uplink .

[0088] Specifically, this step aims to transform the heterogeneous raw sensing data from vehicle-mounted terminals and road infrastructure (roadside) into uplink data packets with clearly defined service type identifiers through a standardized collection, formatting, and packaging process. This step serves as a bridge connecting the physical sensing world and the digital communication network.

[0089] Further, step S2000 includes:

[0090] Step S2100: Utilize the multimodal sensor system mounted on the vehicle terminal VT to collect raw sensor data streams in real time. The raw sensor data stream is processed via the on-board unit (OBU). Real-time processing is performed to generate structured vehicle status data.

[0091] Specifically, this step aims to perform real-time preliminary processing and structured convergence of the vehicle's own perception information, transforming high-dimensional, unstructured raw sensor data into low-dimensional, high-value structured vehicle state data with clear physical meaning.

[0092] In the specific implementation process, firstly, the multimodal sensor system built into the vehicle terminal (VT) is used to continuously scan and measure the vehicle's surrounding environment at a high frequency, thereby generating a massive amount of multimodal raw sensor data stream. The raw sensor data stream It is characterized by massive data volume and diverse formats (such as pixel matrices of images and point cloud data of radar), and it is unprocessed at the semantic level, so it cannot be directly used for communication and decision-making.

[0093] Secondly, the on-board unit (OBU) deployed inside the vehicle is used as an on-board edge computing node to process the input raw sensor data stream. Real-time processing is performed to achieve environmental perception and state recognition. The core processing flow mainly includes the following three stages:

[0094] First, feature extraction. This involves analyzing the input raw sensor data stream. Preliminary analysis of various types of data is performed to extract key semantic elements, such as identifying lane boundaries from camera images and extracting obstacle contours from LiDAR point clouds.

[0095] Second, sensor data fusion. This involves fusing the raw sensor data streams... Align and fuse various types of data in time and space to obtain more accurate and robust perception results than a single sensor.

[0096] Third, state estimation. Based on the fused data, the core dynamic state of the vehicle is estimated using algorithms such as Kalman filtering.

[0097] Finally, after the above processing, the unstructured raw sensor data stream... The data is refined and converged into structured vehicle state data. The vehicle status data is typically formatted according to industry standard protocols for Vehicle-to-Everything (V2X) to ensure good interoperability in subsequent cross-device and cross-platform interactions. The specific expression formula is as follows:

[0098]

[0099] in, Represents structured vehicle status data; T s Indicates vehicle status data A precise timestamp at the time of generation, used to prevent the system from making incorrect decisions based on outdated information; the unit is milliseconds (ms). V ID This represents a unique vehicle identifier used to distinguish different communication entities in multi-vehicle collaborative scenarios; P pos The location information of a vehicle is typically represented by a coordinate data structure containing longitude, latitude, and altitude information; V vec The velocity vector represents the vehicle's speed and is used to comprehensively describe the vehicle's operating state, including the magnitude of its speed and direction of travel; H end The heading angle of a vehicle is used to describe the angle between the vehicle's longitudinal axis and geographic true north, and is measured in degrees (°). nt It represents the vehicle's driving intention and is semantic information extracted from the underlying data, used to predict the vehicle's expected trajectory in the near future.

[0100] Step S2200: Collect raw environmental data streams using roadside sensing devices fixedly deployed on road infrastructure. Combine deep learning models to analyze the original environmental data stream Real-time deep analysis is performed to generate roadside environmental data.

[0101] Specifically, this step aims to utilize roadside sensing devices deployed on road infrastructure to conduct comprehensive and seamless monitoring of the local traffic environment, and distribute the processed structured environmental information to vehicles within the coverage area. This compensates for the limitations of single-vehicle sensing caused by physical constraints such as limited field of view and occlusion interference, providing vehicles with crucial beyond-line-of-sight sensing capabilities.

[0102] In practice, firstly, roadside sensing devices fixedly deployed along the road are used to continuously monitor and collect traffic data within their jurisdiction, thereby generating massive, multimodal raw environmental data streams.

[0103] Subsequently, the raw environmental data stream is processed via a roadside unit (RSU) or a roadside sensing platform with enhanced computing power that is wired to it. Real-time deep analysis processing is performed. This deep analysis process combines multiple advanced intelligent algorithms to provide fundamental support for understanding traffic scenarios by transforming low-level raw signals into high-level semantic information. Specifically, the intelligent algorithms employed include:

[0104] First, computer vision-based target detection and multi-target tracking algorithms. These utilize deep learning models (such as YOLO, Faster R-CNN, etc.) to analyze raw environmental data streams. The video stream is parsed at the frame level to detect and identify traffic participants (such as motor vehicles, non-motor vehicles, pedestrians, etc.) in real time. Combined with multi-object tracking (MOT) technology, a unique identifier ID is assigned to each independent target object, and its motion trajectory in time series is continuously tracked.

[0105] Second, a traffic event recognition algorithm based on spatiotemporal behavior analysis. Building upon the aforementioned target monitoring and multi-target tracking algorithms, this algorithm further analyzes the behavioral patterns and interrelationships of each target object in the temporal and spatial dimensions. By comprehensively judging multi-dimensional parameters such as target trajectory, instantaneous speed, and acceleration, it automatically identifies and infers traffic events with clear semantics. For example, a sudden decrease in the distance between two vehicles is identified as a "rear-end collision risk," or a sustained drop in the average speed of multiple vehicles in an area below a preset threshold is identified as "road congestion."

[0106] Finally, after the above processing, the original environmental data stream Transformed into structured roadside environmental data The roadside environmental data The data is distributed to vehicle terminals (VTs) within the coverage area by the roadside unit (RSU). Its data model is a structured data structure containing multiple list objects to facilitate efficient parsing and utilization. The specific expression formula is as follows:

[0107]

[0108] in, Represents structured roadside environmental data; T' s Represents roadside environmental data The generated precise timestamps, measured in milliseconds (ms), ensure data timeliness; RUS ID This represents a unique Roadside Unit (RSU) identifier, used to identify the data source and provide the corresponding geographic location; {O list} represents the set of dynamic target objects, containing all independent dynamic entities detected and continuously tracked within the field of view of the Roadside Unit (RSU); {E list} represents a specific set of traffic semantic events, used to perceive typical events that cause interference or risk to traffic flow.

[0109] The dynamic target object set {O list In the context of this structure, each element `object` represents a identified dynamic target object, whose key attributes include a unique tracking identifier `obj` dynamically assigned by the roadside equipment. ID ; The category label of the target object, obj Type Enumerated values ​​such as 'CAR' (car), 'PEDESTRIAN' (pedestrian), etc.; the set of location coordinates of the target object, obj. pos and the velocity vector obj of the target object vec .

[0110] The specific traffic semantic event set {E list In this context, each element `event` represents a recognized traffic semantic event, and its key attributes include a unique event identifier `event`. ID Event category tags Type Examples of enumerated values ​​include 'ACCIDENT' (traffic accidents) and 'CONGESTION' (traffic congestion); the geographical scope of the event's impact is the location. Boundary .

[0111] Step S2300: Merge vehicle status data and roadside environmental data Construct surrounding environment situational awareness data, generate a service payload based on the surrounding environment situational awareness data, identify the service type of the service payload, and determine the service type identifier. The network protocol header, the service payload, and the identifier carrying the service type are included. The metadata (Matadata) is encapsulated to form the uplink data packet P to be sent. Uplink .

[0112] Specifically, this step is a crucial juncture between the perception and decision-making layer and the network transport layer. Its core task is to encapsulate the structured information generated in the preceding steps into uplink data packets (P) that conform to network transmission specifications, based on real-time business requirements. Uplink .

[0113] In the specific implementation process, firstly, the application layer of the vehicle terminal VT receives the vehicle status data from step S2100. and roadside environmental data in step S2200 Based on the timestamps and location information in each data set, the data is aligned and fused in both time and space dimensions to form a more comprehensive situational awareness dataset. This situational awareness dataset surpasses the perception capabilities of a single vehicle, providing it with a broader understanding of its environment. Based on this situational awareness dataset, the application layer identifies potential hazards that the vehicle's own sensors cannot detect, generates a payload accordingly, and executes corresponding business decisions.

[0114] For example, an unmanned cleaning vehicle is traveling along a road, and its multimodal sensor system collects onboard status data. It indicates that the road ahead is clear. However, it also receives roadside environmental data from the roadside unit (RSU). The system displays a pedestrian preparing to cross the road in the driver's blind spot. After aligning and fusing these two data sources in time and space, the application layer of the vehicle terminal (VT) displays this previously invisible collision risk in its internal surrounding environment situational awareness map. Based on this, the decision-making module immediately makes an "emergency braking warning" business decision.

[0115] Furthermore, after the application layer of the vehicle terminal (VT) makes business decisions based on the surrounding environment situational awareness data, it sequentially performs business type identification and data packet encapsulation operations on the service payload to construct the final uplink data packet P. Uplink First, the application layer of the vehicle terminal (VT) determines the service type identifier for the service payload to be sent. Subsequently, the business type identifier Encapsulated within metadata; finally, the standard network protocol header, service payload, and metadata with identity tags are integrated and encapsulated to form the final uplink data packet P to be sent. UplinkThe specific formula is as follows:

[0116]

[0117] Among them, P Uplink This represents the uplink data packet to be sent, a fully encapsulated data unit ready to be delivered to the network protocol stack for transmission; the Header represents the standard network protocol header, containing control information necessary for the packet's routing and transmission in the network, such as source / destination IP addresses and port numbers; the Payload(...) represents the data packet's service payload, and (·) indicates content encapsulation, the encapsulated content of which is dynamically determined by the application layer's real-time decisions; Metadata(Meta svc This represents metadata attached to a data packet, which internally encapsulates a metadata object (Meta). svc The metadata object Meta svc The core content is the business type identifier. Where i represents the network slice index, which indicates that this type of service is associated with the i-th network slice.

[0118] S3000, based on parsing of uplink data packets P Uplink Business type identifier in In the end-to-end slicing strategy model P E2E Retrieve slice identifier ID i Quality of Service Level Identifier (QCI) i Encapsulate and generate data packets with tagged slicing strategies. Data packets according to the labeled slice policy Trigger the preset collaborative scheduling logic mechanism to generate scheduling decision M sched With air interface resource allocation R alloc Based on the aforementioned scheduling decision M sched With air interface resource allocation R alloc The data packets that drive the tagged slice strategy End-to-end transmission is completed sequentially in the Radio Access Network (RAN), the Bearer Network (TN), and the Core Network (CN) to generate and output Multi-Access Edge Computing (MEC) data packets.

[0119] Specifically, this step aims to build upon the end-to-end slicing strategy model P constructed in step S1400. E2E The uplink data packet P constructed in step S2300 Uplink The system dynamically selects and executes the optimal low-latency transmission mechanism. This step achieves ultra-low latency communication by intelligently matching application layer service requirements with network layer resource availability in real time.

[0120] Further, step S3000 includes:

[0121] Step S3100: Parse the received uplink data packet P Uplink Extract business type identifier and the business type identifier As a query index, P in the end-to-end slicing strategy model E2E Retrieve slice identifier ID i and the slice identifier ID i Associated Quality of Service (QoS) Level Identifier (QCI) i The slice identifier ID i With Quality of Service (QoS) Level Identifier (QCI) i Appended to uplink data packet P Uplink The above encapsulates and generates data packets with tagged slicing strategies.

[0122] Specifically, this step aims to process the uplink data packet P Uplink Application layer business type identifier carried It translates and maps in real time into standardized network layer policy instructions that the network layer can directly understand and execute.

[0123] In the specific implementation process, firstly, the base station gNodeB, which serves as the network entry point, parses the received uplink data packet P. Uplink And extract and parse the metadata object Meta from it. svc Obtain the data packet P Uplink Business type identifier

[0124] Subsequently, the base station gNodeB will assign the service type identifier. As a query index, in the end-to-end slicing strategy model P E2E Fast matching is performed. The end-to-end slicing strategy model P is described. E2E It serves as a policy routing table, defining the business type identifier. Mapping rules to specific network behaviors. After a successful match, the base station gNodeB implements the end-to-end slicing strategy model P. E2E The two core network layer parameters corresponding to this service type are retrieved from the database: Slice Identifier ID. i and its associated Quality of Service (QoS) Level Identifier (QCI) i .

[0125] Ultimately, based on the uplink data packet P Uplink Add critical scheduling instructions required for subsequent network transmission (i.e., Quality of Service (QoS) level identifiers, QCI) i ) and routing instructions (i.e., slice identifier ID)i Encapsulate them into packets with tagged slicing strategies. The specific formula is as follows:

[0126]

[0127] For example, an unmanned cleaning vehicle is driving along the road, and the situational awareness map around it has been integrated with roadside environmental data. It identifies a potential collision risk, and based on this, the decision module makes a business decision to issue an "emergency braking warning," which is then encapsulated into an uplink data packet P. Uplink When the uplink data packet P Uplink When the 5G module of the unmanned cleaning vehicle sends a data packet and it arrives at the gNodeB base station, the gNodeB base station utilizes its efficient and intelligent network scheduling function. The gNodeB base station does not concern itself with the uplink data packet P. Uplink Instead of analyzing the service payload content (i.e., the specific details of the warning), it directly parses the uplink data packet P. Uplink Meta data object svc And identify the business type identifier. It is designated as 'V2X-Safety'. Subsequently, the base station gNodeB utilizes the service type identifier. In the end-to-end slicing strategy model P E2E A quick query is performed to find the predefined mapping rule: all 'V2X-Safety' service types must be carried by network slice 1 and assigned a QoS level of 9. Therefore, the base station gNodeB will transmit the uplink data packet P... Uplink Packets encapsulated with a tagged slicing strategy At this point, the transmission path of the warning data packet has been clarified: since the QoS level identifier DCI1=9 has high priority, it will immediately trigger the Grant-Free scheduling mechanism, while the slice identifier ID1 ensures that the data packet is routed without error to the specific Flexible Ethernet FlexE hard pipe and User Plane Function UPF virtual forwarding instance reserved for slice 1 with the highest priority.

[0128] Step S3200: Based on the data packets with the marked slice policy Quality of Service (QoS) Level Identifier (QCI) i Triggered according to the preset collaborative scheduling logic mechanism, a scheduling decision M is generated. sched With air interface resource allocation R alloc If the Quality of Service (QoS) Level Identifier (QCI) i For the highest service priority, the Grant-Free scheduling mechanism is activated; if the Quality of Service (QoS) level identifier (QCI) is... iWhen the service is a latency-sensitive service and not the highest priority service, the intelligent uplink pre-scheduling mechanism is activated; if the Quality of Service (QoS) level identifier (QCI) is... i When the service is not time-sensitive and is not the highest priority service, the standard scheduling mechanism is activated.

[0129] Specifically, this step aims to utilize the data packets with the tagged slice strategy encapsulated in step S3100. This is a key collaborative step for dynamically deciding the optimal wireless air interface scheduling scheme for data packets, thereby ensuring ultra-low latency.

[0130] In the specific implementation process, the base station gNodeB, as the unified manager and scheduling center of air interface resources, receives data packets marked with slice policies. Subsequently, the focus was on analyzing the Quality of Service (QoS) Class Identifier (QCI) carried within it. i The base station gNodeB internally pre-configures a set of Quality of Service (QoS) level identifiers (QCI). i The corresponding collaborative scheduling logic. This collaborative scheduling logic functions like a policy-based decision engine, which determines the QoS (Quality of Service) level identifier (QCI) based on different criteria. i This triggers the corresponding collaborative scheduling logic mechanism, which specifically includes the following collaborative scheduling logic mechanisms:

[0131] First, the grant-free scheduling mechanism. This applies when the Quality of Service (QoS) Level Identifier (QCI) is used. i When the value corresponds to the highest service priority in the V2X emergency safety level, the base station gNodeB activates this mechanism. Under this mechanism, the vehicular terminal (VT) does not need to perform the regular scheduling request (SR) and uplink grant (UG) signaling interaction process, but directly assigns the slice ID... i Data transmission is performed on a reserved, dedicated Grant-Free resource pool, thereby saving critical air interface signaling interaction latency of milliseconds or even tens of milliseconds.

[0132] Second, the intelligent uplink pre-scheduling mechanism PRE-SCHEDULE. This is implemented when the Quality of Service (QoS) Level Identifier (QCI) is used. i When the value corresponds to a latency-sensitive but not the highest priority service, the base station gNodeB activates this mechanism. This mechanism intelligently allocates a time-sensitive uplink grant UG only when downlink data destined for the vehicle terminal VT is detected. This approach not only shortens scheduling latency but also avoids occupying idle resources when there is no service, thus achieving an efficient balance between ensuring low latency and conserving air interface resources.

[0133] Third, the standard scheduling mechanism STANDARD-SR. When the Quality of Service (QoS) level identifier (QCI) is used... i When the value corresponds to other ordinary, non-latency-sensitive services, the base station gNodeB activates this mechanism and adopts the standard scheduling request (SR) procedure. This procedure follows the conventional model of "on-demand application and sequential allocation" to ensure the fair and effective allocation of network resources.

[0134] Finally, after triggering the corresponding collaborative scheduling logic mechanism, the system generates a specific scheduling decision M. sched With air interface resource allocation R alloc The scheduling decision M sched It is an enumeration value representing the scheduling strategy dynamically selected by the base station gNodeB based on service priority. Its value comes from a predefined enumeration set M. sched ∈{'Grant-Free','PRE-SCHEDULE','STANDARD-SR'}. The air interface resource allocation R... alloc It encapsulates detailed parameters of the physical resources used for uplink transmission, and these parameters are related to the scheduling decision M. sched The values ​​are closely related, specifically as follows: when M sched When 'Grant-Free' is enabled, the air interface resource allocation R is pre-configured. alloc This can be an index or pointer to the pre-configured resource. The vehicle terminal (VT) uses this index to find the specific time-frequency location, which has already been obtained through Radio Resource Control (RRC) signaling; when M... sched When ='PRE-SCHEDULE' or 'STANDARD-SR', air interface resource allocation R alloc This represents a dynamic uplink authorization UG, the content of which is consistent with the parameters carried in the downlink control information (DCI), and can be represented as a tuple: R alloc =(PRB) map MCS index TPC command ), of which PRB map This represents a bitmap or "start position / length" indicator that specifies the exact location of the allocated Physical Resource Block (PRB) in the frequency domain; MCS index The index of the modulation and coding scheme (MCS) informs the vehicle terminal (VT) what combination of efficiency and robustness should be used to transmit data; TPC command This indicates a Transmit Power Control (TPC) command, used to adjust the transmit power of the vehicle terminal VT.

[0135] Step S3300, based on scheduling decision M sched and air interface resource allocation R alloc The vehicle-mounted terminal (VT) uses the reserved physical resource block (PRB) in the radio access network (RAN) to transmit data packets marked with the slicing policy. Transmitted to base station gNodeB; the data packet In the bearer network TN, based on the slice identifier ID it carries... i It is routed to a dedicated Flexible Ethernet (FlexE) hard pipe and directed to a virtual forwarding instance on the user plane of the core network (CN). Generate and output multi-access edge computing (MEC) data packets

[0136] Specifically, this step aims to base the scheduling decision M generated in step S3200 on... sched and air interface resource allocation R alloc The data packet with the tagged slice strategy encapsulated in the driving step S3100 Efficient and reliable physical transmission is completed in the end-to-end isolated channel from the Radio Access Network (RAN) to the Core Network (CN) constructed in step S1000.

[0137] In the specific implementation process, data packets Following the physical path, the following three key network domains are traversed sequentially:

[0138] The first phase involves efficient transmission within the Radio Access Network (RAN) domain. The onboard terminal (VT) relies on scheduling decisions (M)... sched and air interface resource allocation R alloc Packets that have been tagged with the slicing policy Transmitted to the base station gNodeB via the wireless air interface. Since the Physical Resource Blocks (PRBs) for high-priority services have been reserved, the transmission enjoys the highest priority. This effectively avoids air interface congestion and conflicts caused by public network services, laying a solid foundation for low latency throughout the entire process.

[0139] The second phase involves deterministic transmission within the TN (Transport Network) domain. After being received by the base station gNodeB, the data packet enters the TN. At this point, the data packet carries the slice identifier ID. i It begins to play a crucial role. Network devices in the TN (Transport Network Node) will recognize this slice identifier ID. i This allows for precise routing of data packets to the specific Flexible Ethernet FlexE hard pipe allocated to them in step S1200. Within this physically isolated channel, packet transmission is unaffected by traffic fluctuations from other network slices, ensuring that transmission latency and jitter are strictly controlled and maintained at a deterministic, extremely low level.

[0140] The third stage involves precise reception of the user plane data in the core network (CN) domain. The data packets ultimately arrive at the UPF / MEC node in the core network (CN). The user plane function (UPF) uses the packet's slice identifier (ID) as a reference. i This directs it to the virtual forwarding instance constructed in step S1300. The data packets are then processed. This process ensures that once a data packet arrives, it enters a dedicated processing queue without waiting in line, and is directly handed over to the application on the Multi-access Edge Computing (MEC) platform for response and processing.

[0141] Ultimately, the data packet After completing the end-to-end transmission from the Radio Access Network (RAN) to the Core Network (CN), the Multi-Access Edge Computing (MEC) data packets are generated and output. And ensure that it is delivered completely and without errors to the target application on the edge computing MEC platform. The total latency T of the entire end-to-end uplink transmission. trans This can be modeled by the sum of the delays of each segment, and the specific process formula is as follows:

[0142] T trans =T air +T transport +T core ;

[0143] Among them, T trans T represents the total latency of end-to-end uplink transmission, which is the complete time span from the time the transmission is initiated by the vehicle terminal (VT) to the time the application in the multi-access edge computing (MEC) receives the data. It is a core performance metric. air This represents the RAN (Radio Network Area Transmission) air interface latency, which is mainly composed of the propagation delay of the radio channel and signaling scheduling overhead; T transport This represents the transmission delay of the bearer network (TN), which is mainly caused by the data exchange and transmission delay in the wired network; T core This indicates the processing latency of the core network (CN) user plane. This latency is mainly generated by the user plane function (UPF) for data packet lookup, forwarding, and other processing.

[0144] S4000, real-time analysis of multi-access edge computing (MEC) data packets. Generate local decision results R local and global analysis data D global Based on the global analysis data D global A digital twin model of traffic is constructed, a large-scale machine learning model is used for prediction, and an operations research optimization algorithm engine is activated for simulation and deduction, which is then encapsulated into a global optimization strategy S. global Based on the preset fusion logic, the local decision result R is fused. local and global optimization strategy Sglobal Generate the final control command or warning information C final And send it to the target vehicle or roadside facility.

[0145] Specifically, this step aims to leverage the collaborative computing power of the multi-access edge computing (MEC) platform deployed at the network edge and the central cloud to process the multi-access edge computing (MEC) data packets successfully transmitted by the S3300. Deep processing is performed to achieve millisecond-level rapid response to traffic events and global macro-level optimization of the traffic system. This step is a key closed loop connecting communication capabilities with intelligent applications.

[0146] Further, step S4000 includes:

[0147] Step S4100, MEC data packets based on multi-access edge computing The local real-time analysis engine is invoked to analyze the vehicle status data in the data packet payload. and roadside environmental data Perform in-depth analysis and output local decision results R. local and global analysis data D global .

[0148] Specifically, this step aims to push computing tasks closer to the network edge where the data source is located, and complete the most time-sensitive calculations through a multi-access edge computing (MEC) platform to achieve ultra-low latency response for services, thereby meeting the high real-time requirements of intelligent transportation safety applications.

[0149] In the specific implementation process, the vehicle-road cooperative application deployed on the multi-access edge computing (MEC) platform receives the multi-access edge computing (MEC) data packet. Then, the data packets will be parsed immediately. Next, the vehicle-to-everything (V2X) application will invoke the local real-time analysis engine to analyze the data packets. Vehicle status data and roadside environmental data A deep secondary analysis is performed. This process typically involves computationally intensive algorithms such as collision detection, trajectory prediction, and beyond-line-of-sight hazard perception, aiming to infer immediate and dangerous emergencies from the collected data. Based on the analysis results, two different types of output are generated simultaneously, as follows:

[0150] First, output the local decision result R. local The local decision result R localIt encapsulates immediate, executable decision instructions for local, high-time-sensitivity events. It is the core output of multi-access edge computing (MEC) platforms designed to achieve millisecond-level business response. It avoids redundant analysis processes, directly generating the most critical decision instructions to trigger downstream instruction generation and security responses in the shortest possible time. The specific local decision result R... local The formula for expressing this is as follows:

[0151] R local =(Decision) Type Priority, Target ID ,Content);

[0152] Among them, Decision Type The `LocalDecisionType` is an enumeration value used to define the type of action to be performed, such as `COLLISION_WARNING` or `EMERGENCY_BRAKE_COMMAND`. `Priority` is an enumeration value used to indicate the urgency of the decision, such as `CRITICAL` or `HIGH`. `Target` is the target value. ID Represents one or more target entity identifiers; it is a list used to specify the recipient of the decision, such as [VID]. A VID B Content represents key information on which local decisions are based, such as the precise location of the hazard source and its velocity vector.

[0153] Second, output global analysis data D global The global analysis data D global It is a non-real-time, strategic-level data aggregation method, primarily generated by aggregating, de-identifying, and extracting features from the local real-time processing results of multi-access edge computing (MEC). It removes details requiring immediate response while retaining statistical and pattern information of significant value for macro-level traffic situation analysis, providing data support for large-scale traffic flow analysis, model training, and global strategy optimization in the cloud.

[0154] Step S4200, based on global analysis data D global A traffic digital twin model is constructed, and a large-scale machine learning model is used to analyze the traffic digital twin model to predict potential macro-level traffic problems. Based on the traffic problems, an operations research optimization algorithm engine is activated to solve for the optimal strategy and encapsulate it into a global optimization strategy S. global .

[0155] Specifically, this step aims to leverage the powerful computing capabilities and global data advantages of the central cloud as a macro-optimization step. It handles complex computational tasks that are not real-time or urgent but are crucial to the overall operational efficiency of the transportation system, thereby providing the transportation system with a global optimization strategy. global .

[0156] In the specific implementation process, firstly, the central cloud platform, as a data aggregation and analysis center, continuously receives the discrete global analysis data D reported in step S4100. global and the global analysis data D global Integration is performed across time and space dimensions. This is based on the global analysis data D of the wide-area traffic information. global A regional, dynamically updated digital twin model of transportation is built in the cloud.

[0157] Subsequently, large-scale machine learning models perform in-depth analysis of the aforementioned traffic digital twin model to identify or predict potential macro-level traffic problems. For example, time series prediction models predict the probability of congestion on a corridor within the next 30 minutes based on the changing trends of traffic flow data. Once a specific problem is diagnosed, such as "imminent congestion," the cloud-based operations research optimization algorithm engine is activated. This engine utilizes the traffic digital twin model to perform multiple "what-if" simulations. For example, regarding the aforementioned impending congestion, the optimization algorithm simulates various traffic light timing schemes or convoy routing strategies and evaluates the impact of each scheme on future traffic flow (such as average travel time and queue length). The operations research optimization algorithm then solves for the optimal strategy based on maximizing a preset objective (such as "maximizing regional traffic efficiency").

[0158] Finally, after optimizing and solving for the optimal strategy, the system encapsulates it into a standardized, deployable global optimization strategy S. global The global optimization strategy S global Typically, this is distributed to relevant edge nodes in the form of configuration updates or macro-level behavioral guidelines, as shown in the following formula:

[0159] S global =(Policy) Type ,Parameters,Effective Scope );

[0160] Among them, S global The global optimization strategy is represented by a tuple; Policy Type This represents the type of global optimization strategy, and is an enumeration value used to define the global optimization strategy S. globalProperties, such as signal timing update 'SIGNAL_TIMING_UPDATE' or fleet routing suggestion 'FLEET_ROUTING_ADVISORY'; Parameters represent the global optimization strategy S. global The specific parameters are related to the global optimization policy type. Type Closely related data structures, such as when the global optimization policy type is Policy Type When set to 'SIGNAL_TIMING_UPDATE', the specific parameters of the global optimization strategy may include a completely new scheme for signal light phases and durations specific to a particular intersection; Effective Scope Indicates the scope of the global optimization strategy; it is a list or geographical description, such as ['RSU']. A ','RSU B '] represents the global optimization strategy S global It is only effective for the areas covered by the two roadside units A and B RSU.

[0161] For example, in a city center area, three adjacent intersections A, B, and C are each covered by three Multi-access Edge Computing (MEC) nodes. Over the past hour, the central cloud platform has continuously received global analysis data D reported by these three MEC nodes. global The specific steps are as follows:

[0162] First, global situational analysis and problem diagnosis. The cloud-based analytics engine will process this global analysis data... global Aggregate analysis revealed that along the corridor from A to B to C, the frequency of "emergency braking" and "close-range collision" events increased by 50% compared to the previous period, while traffic flow data showed that the average speed of traffic in this corridor decreased by 30%. Based on this information, a machine learning model predicted that the corridor would experience severe congestion in 20 minutes.

[0163] Subsequently, strategy generation and simulation were performed. The cloud-based traffic simulation model, based on the aforementioned early warning information, initiated an optimization algorithm to simulate three new traffic light timing schemes. After evaluation, the "green wave" scheme was deemed the optimal solution, as it improved traffic efficiency. The "green wave" scheme is a technical term in intelligent traffic signal control, referring to the coordinated and linked control of traffic lights at multiple consecutive intersections on main roads. By setting a precise phase difference for the green light start time at each intersection, it ensures vehicles pass smoothly through each intersection at the recommended speed, avoiding or reducing stopping time, and achieving continuous, unobstructed traffic flow.

[0164] Furthermore, the system handles output encapsulation and policy distribution. It parses the aforementioned "green wave" scheme and incorporates it into the global optimization strategy S.global The following global optimization strategy S is ultimately generated. global =('SIGNAL_TIMING_UPDATE',{GreenWave_Plan_v2},['RSU A ',RSU B ','RSU C ']). Among them, 'SIGNAL_TIMING_UPDATE' is used to inform the receiver that more edge computing MEC nodes are connected to update the traffic light timing; {GreenWave_Plan_v2} contains all the technical details of the "green wave" scheme, specifically including the precise status (e.g., red, green, yellow), green light duration, and phase sequence of the three traffic lights A, B, and C at intersections for the next hour, as well as other complex parameter information; ['RSU A ',RSU B ','RSU C The above "green wave" scheme is specified to indicate that this strategy is only effective for the areas covered by A, B and C in the road test unit (RSU).

[0165] Finally, the follow-up actions. This global optimization strategy S global The event is asynchronously pushed to the three multilateral access edge computing (MEC) nodes responsible for intersections A, B, and C. In subsequent steps, when these MEC nodes are processing local events, they follow this global optimization strategy S. global As a macro-level guide, it generates smarter and more coordinated downlink instructions, thereby achieving overall efficiency and safety of traffic flow.

[0166] Step S4300: Integrate local decision results R local and global optimization strategy S global Generate the final control command or warning information C final And the final control command or warning information C final The data is sent to the target vehicle or roadside facility. The fusion process follows a preset fusion logic. If the local decision result R... local When the priority is critical, the local decision result R is adopted first. local Conversely, using the global optimization strategy S... global This serves as a basis for decision-making.

[0167] Specifically, this step aims to integrate the local decision results R output from step S4100. local and the global optimization strategy S generated based on step S4200 global Through the collaborative work of edge intelligence and cloud intelligence, a context-aware final control command or early warning message can be generated and issued, capable of simultaneously responding to emergencies and optimizing the overall situation.final .

[0168] In the specific implementation process, the multi-side access edge computing (MEC) node serves as the final fusion point for decision-making. Its internal fusion engine comprehensively considers two inputs with different properties: the local decision result R. local and global optimization strategy S global This fusion engine does not simply overlay information, but follows a priority-based fusion logic, as follows:

[0169] First, the principle of safety first. When the received local decision result R... local When the priority is 'CRITICAL', such as an imminent collision warning, the fusion engine will grant it the highest processing authority. In this case, the local decision result R... local The decisions made are prioritized and immediately packaged into final instructions, while the global optimization strategy in the cloud is... global This information is only optional and should never delay or cover critical security decisions.

[0170] Second, the principle of efficiency optimization. When the local decision outcome R... local When the priority is "normal" or "high" and it is not a critical security incident, the fusion engine will apply the cloud-based global optimization strategy S. global Integrating these elements as key decision-making criteria will further enhance the intelligence and efficiency of decision-making processes.

[0171] After completing the above fusion decision, the system will encapsulate it into the final control command or early warning information C. final And through the low-latency downlink ensured by network slicing, it accurately sends the data to the target vehicle or relevant roadside facilities, ultimately completing the entire closed loop of "sensing-communication-computing-control". The specific formula is as follows:

[0172] C final =(Timestamp,Command) Type ,Target ID ,Content);

[0173] Among them, Timestamp represents the final control command or warning information C. final The generated timestamps are used to assess the timeliness of commands and provide an accurate time reference for subsequent system logs and post-event analysis; Command Type C indicates control commands or warning information final The instruction type is an enumeration value used to explicitly inform the target receiver of the type of action to be performed, such as collision warning or emergency braking. IDThis represents a list of final target receiver identifiers used to ensure accurate delivery of instructions. Elements in this list can be V. ID A single or multiple vehicle IDs can be used to issue personalized commands to a specific vehicle; this can also be an RSU. ID One or more roadside unit IDs are used to control or update specific roadside facilities; Content represents control commands or warning information. final The parameter load, whose value is related to the control command or warning information C final Command type Type Tightly coupled, encapsulating the execution of the control command or warning information C final All the required detailed parameters.

[0174] Example 2

[0175] This embodiment, based on Embodiment 1, provides an ultra-low latency cooperative communication system, such as... Figure 2 As shown, it includes a network domain slicing configuration module, an on-board and road test data fusion module, an uplink data transmission module, and a decision and closed-loop control module;

[0176] The network domain slicing configuration module is based on the service level protocol. The key fields are used to generate slice configuration data for three network domains: wireless access network, bearer network, and core network. This slice configuration data is then integrated to construct an end-to-end slice strategy model P. E2E Call the end-to-end consistency verification function Verify(ID) i For the end-to-end slicing strategy model P E2E Perform bandwidth continuity and latency compliance checks. If the checks pass, then the end-to-end slicing strategy model P is updated. E2E Activate if enabled, otherwise reject and trigger an alarm mechanism;

[0177] The vehicle-mounted and roadside data fusion module is used to collect vehicle status data. and roadside environmental data Integrating the vehicle status data and roadside environmental data Construct surrounding environment situational awareness data, determine the service payload based on the surrounding environment situational awareness data, and generate a service type identifier based on the service payload. The network protocol header, the service payload, and the accompanying service type identifier are included. The metadata (Matadata) is encapsulated to form the uplink data packet P. Uplink ;

[0178] The uplink data transmission module is based on the uplink data packet P. Uplink Business type identifier in In the end-to-end slicing strategy model P E2E Retrieve slice identifier ID i Quality of Service Level Identifier (QCI) i Encapsulate and generate data packets with tagged slicing strategies. Data packets according to the labeled slice policy Trigger the preset collaborative scheduling logic mechanism to generate scheduling decision M sched With air interface resource allocation R alloc Based on the aforementioned scheduling decision M sched With air interface resource allocation R alloc The data packets that drive the tagged slice strategy End-to-end transmission is completed sequentially in the wireless access network, bearer network, and core network to generate and output multi-access edge computing data packets.

[0179] The decision-making and closed-loop control module analyzes multi-access edge computing data packets in real time. Generate local decision results R local and global analysis data D global Based on the global analysis data D global A digital twin model of traffic is constructed, a large-scale machine learning model is used for prediction, and an operations research optimization algorithm engine is activated for simulation and deduction, which is then encapsulated into a global optimization strategy S. global Based on the preset fusion logic, the local decision result R is fused. local and global optimization strategy S global Generate the final control command or warning information C final And send it to the target vehicle or roadside facility.

[0180] The parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0181] 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 method for ultra-low latency cooperative communication, characterized in that, include: Based on the key fields of the service level agreement, slice configuration data for three network domains—wireless access network, bearer network, and core network—is generated. The slice configuration data is integrated and an end-to-end slice policy model is constructed. An end-to-end consistency verification function is called to verify the bandwidth continuity and latency compliance of the end-to-end slice policy model. If the verification passes, the end-to-end slice policy model is activated; otherwise, it is rejected and an alarm mechanism is triggered. Collect vehicle status data and roadside environment data, fuse the vehicle status data and roadside environment data to construct surrounding environment situational awareness data, determine service payload based on the surrounding environment situational awareness data, generate service type identifier based on service payload identification, and encapsulate the network protocol header, the service payload and metadata with attached service type identifier to form uplink data packets. Based on the service type identifier in the uplink data packet, the slice identifier and service quality level identifier are retrieved from the end-to-end slice strategy model, and the data packet with the slice strategy marked is encapsulated and generated. Based on the data packets with the marked slice strategy, a preset collaborative scheduling logic mechanism is triggered to generate scheduling decisions and air interface resource allocation; based on the scheduling decisions and air interface resource allocation, the data packets with the marked slice strategy are driven to complete end-to-end transmission in the wireless access network, bearer network and core network in sequence, generating and outputting multi-access edge computing data packets; The method for generating multi-access edge computing data packets includes: parsing the received uplink data packets, extracting the service type identifier, using the service type identifier as a query index to search in the end-to-end slicing strategy model, obtaining the slicing identifier and the service quality level identifier associated with the slicing identifier, appending the slicing identifier and the service quality level identifier to the uplink data packet, and encapsulating it to generate a data packet marked with the slicing strategy. Based on the Quality of Service (QoS) level identifier in the data packet with the tagged slice policy, a preset collaborative scheduling logic mechanism is triggered to generate scheduling decisions and air interface resource allocation. If the QoS level identifier corresponds to the highest service priority, the Grant-Free scheduling mechanism is activated; if the QoS level identifier corresponds to a latency-sensitive service but is not the highest service priority, the intelligent uplink pre-scheduling mechanism is activated; if the QoS level identifier corresponds to a non-latency-sensitive service and is not the highest service priority, the standard scheduling mechanism is activated. Based on scheduling decisions and air interface resource allocation, the vehicle terminal transmits data packets marked with slicing strategies to the base station through the reserved physical resource blocks in the wireless network access network; in the bearer network, the data packets are routed to dedicated flexible Ethernet hard pipes according to the slice identifier they carry, and are directed to virtual forwarding instances on the user plane of the core network to generate and output multi-access edge computing data packets. The system analyzes multi-access edge computing data packets in real time to generate local decision results and global analysis data. Based on the global analysis data, it constructs a traffic digital twin model, uses a large-scale machine learning model for prediction, and activates an operations research optimization algorithm engine for simulation and deduction. The results are then encapsulated into a global optimization strategy. According to a preset fusion logic, the local decision results and the global optimization strategy are fused to generate the final control command or early warning information and send it to the target vehicle or roadside facility.

2. The ultra-low latency cooperative communication method according to claim 1, characterized in that, The method for constructing the end-to-end slicing strategy model includes: The service quality level identifier is determined based on the maximum latency and service priority in the service level agreement, and the number of physical resource blocks reserved is calculated in combination with the minimum guaranteed bandwidth and modulation and coding scheme in the service level agreement. Then, the wireless access network layer slice configuration data is integrated, generated, and distributed to the base station. Based on the number of physical resource blocks reserved in the wireless access network layer slice configuration data, the actual configured bandwidth in the wireless access network is calculated; and based on the actual configured bandwidth, the number of flexible Ethernet time slots required for broadband transmission is calculated, thereby confirming the guaranteed bandwidth in the bearer network; after completing the legality verification of the total guaranteed bandwidth and the bearer limit of the physical link, bearer network slice configuration data for performing resource hard isolation is generated. Based on the association resolution of bearer network slice configuration data and service level protocol, a virtual forwarding instance bound to the path service quality policy set is constructed; through the session management function in the control plane function, the path service quality policy set and the virtual forwarding instance are encapsulated into N4 session rules, and distributed to the edge user plane function through the N4 interface to integrate and generate core network slice configuration data. Integrate wireless access network slice configuration data, bearer network slice configuration data, and core network slice configuration data to construct an end-to-end slice strategy model.

3. The ultra-low latency cooperative communication method according to claim 2, characterized in that, The aforementioned method for verifying the legality of the total bandwidth and the physical link's carrying capacity refers to ensuring that the total bandwidth will not exceed the physical link's carrying capacity.

4. The ultra-low latency cooperative communication method according to claim 2, characterized in that, The end-to-end consistency verification function verifies bandwidth continuity and latency compliance using the following methods: the guaranteed bandwidth value of the wireless access network is less than or equal to the guaranteed bandwidth value of the bearer network, and the sum of the estimated latency values ​​of all network domains is less than or equal to the maximum latency of the network slice.

5. The ultra-low latency cooperative communication method according to claim 1, characterized in that, The method for constructing the uplink data packet includes: By utilizing the multimodal sensor system mounted on the vehicle terminal, raw sensor data streams are collected in real time, and the raw sensor data streams are processed in real time by the vehicle unit to generate structured vehicle status data. By using roadside sensing devices that are fixedly deployed on road infrastructure, raw environmental data streams are collected, and the raw environmental data streams are processed in real time using a deep learning model to generate roadside environmental data. By integrating vehicle status data and roadside environmental data, a surrounding environment situational awareness data is constructed. A service payload is generated based on the surrounding environment situational awareness data. The service payload is then used to identify the service type and determine the service type identifier. The network protocol header, the service payload, and the metadata carrying the service type identifier are then encapsulated to form an uplink data packet to be sent.

6. The ultra-low latency cooperative communication method according to claim 1, characterized in that, The method for generating the final control command or early warning information includes: Based on multi-access edge computing data packets, the local real-time analysis engine is invoked to perform in-depth analysis on the vehicle status data and roadside environment data in the data packet payload, and output local decision results and global analysis data. A traffic digital twin model is constructed based on global analysis data. The traffic digital twin model is analyzed using a large-scale machine learning model to predict potential macro-traffic problems. Based on the traffic problems, an operations research optimization algorithm engine is activated to solve for the optimal strategy and encapsulate it into a global optimization strategy. The fusion process integrates local decision-making results and global optimization strategies to generate final control commands or early warning information, which are then sent to target vehicles or roadside facilities. The fusion process follows a preset fusion logic: if the priority of the local decision-making result is critical, the local decision-making result is adopted first; otherwise, the global optimization strategy is used as the decision-making basis.

7. An ultra-low latency cooperative communication system, used to implement the ultra-low latency cooperative communication method according to any one of claims 1-6, characterized in that, include: The module includes a network domain slicing configuration module, an on-board and road test data fusion module, an uplink data transmission module, and a decision-making and closed-loop control module. The network domain slicing configuration module generates slicing configuration data for three network domains—wireless access network, bearer network, and core network—based on key fields of the service level protocol. It integrates the slicing configuration data and constructs an end-to-end slicing policy model. It then calls an end-to-end consistency verification function to verify the bandwidth continuity and latency compliance of the end-to-end slicing policy model. If the verification passes, the end-to-end slicing policy model is activated; otherwise, it is rejected and an alarm mechanism is triggered. The vehicle-mounted and roadside data fusion module is used to collect vehicle-mounted status data and roadside environmental data, fuse the vehicle-mounted status data and roadside environmental data to construct surrounding environmental situational awareness data, determine service payload based on the surrounding environmental situational awareness data, generate service type identifiers based on service payloads, and combine the network protocol header, the service payload, and the metadata with the attached service type identifiers. according to Encapsulation is performed to form uplink data packets; The uplink data transmission module retrieves the slice identifier and service quality level identifier from the end-to-end slice strategy model based on the service type identifier in the uplink data packet, and encapsulates and generates a data packet marked with the slice strategy. Based on the data packets with the marked slice strategy, a preset collaborative scheduling logic mechanism is triggered to generate scheduling decisions and air interface resource allocation; based on the scheduling decisions and air interface resource allocation, the data packets with the marked slice strategy are driven to complete end-to-end transmission in the wireless access network, bearer network and core network in sequence, generating and outputting multi-access edge computing data packets; The decision-making and closed-loop control module analyzes multi-access edge computing data packets in real time, generates local decision results and global analysis data, constructs a traffic digital twin model based on the global analysis data, uses a large-scale machine learning model for prediction, activates an operations research optimization algorithm engine for simulation and deduction, encapsulates it into a global optimization strategy, and according to the preset fusion logic, merges the local decision results and the global optimization strategy to generate the final control command or early warning information and send it to the target vehicle or roadside facilities.

Images