A dynamic data packet scheduling method and system

Abstract

The application provides a dynamic data packet scheduling method and system, and belongs to the field of data processing. The application introduces wireless state perception, real-time flow perception and TSN time constraint analysis mechanism on the DS-TT side, and forms a closed-loop control through dynamic scheduling decision and scheduling execution feedback, so as to realize time perception collaborative scheduling of cross 5G network and time sensitive network TSN. The application can reduce the influence of wireless fluctuation on the time delay of high priority data packet, reduce scheduling conflict and improve resource utilization efficiency, guarantee that the key time sensitive network TSN service is completed within the specified time window, and is suitable for deterministic communication scenes such as intelligent manufacturing, rail transit and power automation.

Description

Technical Field

[0001] This invention belongs to the field of data processing, and in particular relates to a dynamic data packet scheduling method and system. Background Technology

[0002] With the increasing demands for high reliability, low latency, and deterministic communication in fields such as the Industrial Internet, intelligent manufacturing, rail transit, and power automation, Time-Sensitive Networking (TSN) technology has gradually become a core supporting technology for industrial Ethernet. TSN achieves deterministic protection of critical business traffic through time-aware scheduling, traffic shaping, and precise time synchronization mechanisms.

[0003] At the same time, the fifth-generation mobile communication system, with its advantages of large bandwidth, low latency and flexible deployment, has been widely introduced into industrial scenarios to achieve wireless access and cross-domain interconnection.

[0004] To achieve deep integration of 5G networks and Time-Sensitive Network (TSN), 3GPP introduced the 5G-TSN converged architecture, in which the Distributed TSN Translator (DS-TT) serves as a key network node, responsible for mapping, forwarding, and scheduling data packets between the 5G network and the TSN.

[0005] In 5G TSN converged networks, DS-TT needs to handle multiple types of data streams from both the radio and wired sides. It must meet the strict latency upper bound and deterministic forwarding requirements of Time-Sensitive Networking (TSN), while also adapting to the dynamic latency and jitter characteristics introduced by 5G wireless networks due to changes in channel conditions and scheduling strategy adjustments. Therefore, DS-TT's packet scheduling capability directly impacts the overall deterministic communication performance of the 5G-TSN converged network.

[0006] The disadvantages of existing technologies are as follows: Static scheduling algorithms are insufficiently adaptable to changes in wireless channels: Existing DS-TT systems mostly employ static or fixed-priority scheduling algorithms, which are difficult to dynamically adjust based on changes in 5G wireless channel quality, load fluctuations, and scheduling strategies. When the wireless environment changes, it may lead to increased queuing time for high-priority data packets, resulting in end-to-end latency exceeding limits.

[0007] Lack of a cross-network collaborative scheduling mechanism between 5G and TSN: Existing scheduling schemes typically optimize only a single network side, failing to establish a collaborative scheduling mechanism between the 5G network and the time-sensitive network (TSN), resulting in inconsistent scheduling rhythms between the two sides. When multiple types of services run concurrently, scheduling conflicts are likely to occur, reducing the overall utilization efficiency of network resources.

[0008] The scheduling mechanism is simple and lacks real-time traffic awareness: The existing DS-TT scheduling mechanism has a relatively simple structure and usually relies on static configuration parameters for forwarding control. It does not perceive or provide feedback on real-time traffic status, queue congestion, and packet arrival characteristics, making it difficult to perform fine-grained scheduling based on dynamic changes in services.

[0009] The time-constrained characteristics of Time-Sensitive Network (TSN) services have not been fully considered: services typically have clear transmission cycles, time windows, and deadlines, but the existing DS-TT scheduling mechanism does not fully integrate TSN time-aware scheduling information and cannot differentiate data packets according to time constraints, which can easily affect the deterministic transmission performance of critical services. Summary of the Invention

[0010] To address the aforementioned shortcomings in existing technologies, this invention provides a dynamic data packet scheduling method and system that solves problems such as the existing DS-TT static scheduling algorithm's inability to adapt to changes in 5G wireless channels, lack of cross-network collaborative scheduling between 5G and TSN, simple scheduling mechanism that does not support real-time traffic awareness, and insufficient consideration of time constraints of time-sensitive network TSN services.

[0011] To achieve the above objectives, the technical solution adopted by this invention is: a dynamic data packet scheduling method, comprising the following steps: S1. After the data packets from the 5G network side and the time-sensitive network (TSN) side are uniformly accessed, a traffic status snapshot is formed by classifying the service flow, counting the queue length and queuing delay. S2. Collect dynamic information of the wireless link on the 5G network side, transform the dynamic information into a wireless risk factor that directly affects scheduling, bind the wireless risk factor with the traffic status snapshot to form the wireless part of the joint input frame, and perform time alignment processing; wherein, the wireless risk factor carries both the sampling timestamp and the validity period. S3. After completing the binding and time alignment, obtain the periodicity and window boundary of the Time Sensitive Network (TSN) service, and output conflict warning information by calculating the remaining deadline and constraint level. S4 integrates traffic status snapshots, wireless risk factors, and time-sensitive network (TSN) time constraints to generate a complete policy of weights, quotas, dequeue order, in-window transmission plan, and conflict avoidance. Based on the complete policy, parameter adaptation and threshold hysteresis control are performed. Among them, the TSN time constraints include period, window, deadline, and conflict warning. S5. Based on the threshold hysteresis control results and according to the currently effective executable policies, optimize and protect the critical time-sensitive network TSN services within the window. S6. Based on the optimized protection processing results, perform data forwarding and gated transmission; S7. Statistically analyze the data forwarding and gating transmission results, generate the feedback report required for closed-loop scheduling, and perform parameter adaptation or trigger re-decision to complete the scheduling of dynamic data packets.

[0012] The beneficial effects of this invention are as follows: This system architecture is designed for DS-TT nodes in 5G-TSN converged networks, constructing a dynamic data packet scheduling closed-loop system of "perception-decision-execution-feedback". This system uses a configuration management interface module to uniformly access service data and status information from both the Time-Sensitive Network (TSN) and the 5G network, converting inputs from different network sides into uniformly processable scheduling objects within the DS-TT, thus avoiding conflicts and resource waste caused by inconsistent scheduling rhythms between the two sides.

[0013] Further, S1 includes the following steps: Acquire data packets from the 5G network side and the Time-Sensitive Network (TSN) side; Data packets are classified and identified according to preset service identification rules, and the data packets are mapped to queue groups. Statistics are updated at the granularity of queue / service flow, and the statistical results are saved to the state sharing and buffer area. The statistics include queue length, average and percentile queuing delay estimates, arrival rate, burstiness, drop count, reorder count, and critical flow percentage. The identification fields include 5-tuple, VLAN PCP / DSCP, Stream ID of Time Sensitive Network (TSN), service tag, and flow mapping table issued by CNC / local policy. Based on statistics, when the scheduling cycle boundary or the trigger event threshold is reached, the current statistical value is solidified into a traffic status snapshot. The traffic status snapshot includes the congestion level of each queue, queuing delay estimate, number and proportion of critical flows, burstiness index and snapshot timestamp T_snap. After the traffic status snapshot is generated, write the snapshot sequence number Snapshot_ID and the snapshot timestamp T_snap, and set the traffic snapshot valid flag. After reading the valid flag of the traffic snapshot using the scheduling controller SC, the snapshot sequence number Snapshot_ID is locked as the decision input for this round. At the same time, the read and write versions of the corresponding snapshot are frozen, thus completing the formation of the traffic status snapshot. If S1 generates a new traffic status snapshot before S2 is completed, it will only enter the pending queue and will not replace the currently locked traffic status snapshot, ensuring the consistency of the decision cycle.

[0014] The beneficial effects of the above-mentioned further solutions are: to solidify traffic state snapshots at the boundary of the scheduling cycle or under event-triggered conditions, and to adopt a snapshot sequence number and version freezing mechanism to ensure that the entire scheduling decision process is based on the same consistent input view, avoiding decision mismatch caused by state drift in a high-dynamic traffic environment; at the same time, through the snapshot validity flag and locking mechanism, to prevent new snapshots from interfering with the current decision cycle, thereby improving the certainty, repeatability and overall stability of scheduling decisions.

[0015] Furthermore, S2 includes the following steps: The system collects dynamic information about wireless links on the 5G network side. This dynamic information includes link quality level, scheduling rhythm information, wireless side buffer / queue backlog, and end-to-end wireless transmission latency statistics for the past N cycles. Sliding window filtering and threshold hysteresis determination are applied to dynamic information indicators; Based on the judgment result, the dynamic information is transformed into a wireless risk factor R that directly affects the scheduling. The wireless risk factor R carries both the sampling timestamp T_wireless and the validity period. Bind the wireless risk factor R with the snapshot sequence number Snapshot_ID to form the wireless part of the joint input frame: Snapshot_ID, T_snap, R, T_wireless; The scheduling controller (SC) collects the wireless status of the same period and performs time alignment verification based on the snapshot sequence number Snapshot_ID. If |T_wireless If T_snap| exceeds the threshold, a supplementary sampling is triggered: if the supplementary sampling fails, the wireless risk factor R from the previous cycle is used and the R_use status is marked; if the wireless side interface is abnormal, the scheduling controller SC enters the degraded state and generates an available risk factor R; after completing the binding and time alignment, it enters S3.

[0016] Furthermore, step S3 includes the following steps: Read the key parameter set for each Time-Sensitive Network (TSN) stream. The key parameter set includes the period T, the gating control list (GCL) or equivalent window boundaries W_start to W_end, the maximum available time slot / bandwidth per period, and the service level and priority constraint rules. Among them, W_start represents the start time of the window in which the TSN stream is allowed to start sending data within a scheduling period, and W_end represents the end time of the window in which the TSN stream must complete sending within a scheduling period. Construct a mapping relationship between the Stream ID of the Time-Sensitive Network (TSN) and the local queue / service flow identifier, so that the TSN packets identified in S1 are associated with the corresponding time constraint entries; Based on the mapping results, the effective window boundary within the current scheduling period is calculated using the local unified time base, and the remaining deadline (deadline_remain) is calculated for each time-sensitive network (TSN) data packet to be scheduled. At the same time, the constraint level is output for the queue level. Calculate the window pressure index: Given the remaining capacity of the current window, can the total bytes of the critical packets to be sent be completed within the window? If insufficient or multiple critical flow windows overlap and compete, a conflict warning message is output. After outputting (T, W_start / W_end, deadline_remain, constraint level, conflict warning), the scheduling controller (SC) encapsulates it along with the traffic status snapshot and the radio risk factor R into a decision input frame, and binds them uniformly to the same snapshot sequence number Snapshot_ID and decision version number Input_Ver. Based on the encapsulation result, a consistency check is performed: if a Time-Sensitive Network (TSN) configuration is missing, the window cannot be calculated, the time base is abnormal, or the mapping is missing, a constraint anomaly flag is set and a security policy is adopted. After completing the consistency check, the scheduling controller (SC) freezes the decision version number Input_Ver and generates a policy that can be traced back to the specific input version.

[0017] The beneficial effects of the above-mentioned further solutions are as follows: the above solutions enable the scheduling system to accurately perceive time constraint pressure before scheduling; by combining window capacity assessment and conflict warning output, the window competition risk between critical flows can be effectively identified in advance; and by uniformly encapsulating and freezing time constraints, wireless risk factors and traffic snapshots into the same decision input version, the temporal consistency and traceability of the policy generation process are guaranteed, which significantly improves the reliability and security of cross-5G and TSN collaborative scheduling.

[0018] Furthermore, step S4 includes the following steps: The traffic status, radio risk factor R, resource status, and time constraints of the time-sensitive network (TSN) are used to aggregate the decision version number Input_Ver of the scheduling controller (SC). Based on the convergence results, the overall scheduling priority P is calculated at the granularity of queue / business flow; Based on the comprehensive scheduling priority P, output a complete strategy including weight, quota, dequeue order, in-window sending plan, and conflict avoidance. For the complete policy, apply threshold and hysteresis rules to generate an executable policy Policy_Ver; Write the executable policy Policy_Ver to the policy double buffer. First, write it to the spare policy area along with (Input_Ver, Policy_Ver, effective boundary time). Then, publish the new policy availability flag through an atomic switch. Based on the write results, the system switches to the new executable policy Policy_Ver only when the effective boundary is detected, thus completing threshold hysteresis control.

[0019] The beneficial effects of the above-mentioned further solutions are as follows: This solution integrates traffic status, wireless risk, resource status, and time constraints of Time Sensitive Network (TSN) to form a unified comprehensive scheduling priority model, enabling fine-grained scheduling decisions across queues and service flows; at the same time, it introduces threshold and hysteresis control mechanisms and adopts a policy double-buffering and atomic switching method to publish executable policies, effectively avoiding frequent policy jitter and intermediate state failures, and improving the stability and real-time response capability of the scheduling system in dynamic network environments.

[0020] Furthermore, S5 specifically includes: Based on the current executable policy_Ver, scheduling is differentiated between windowed and non-windowed states. Specifically, when within a Time-Sensitive Network (TSN) window, priority is given to critical TSN queues that are allowed to send within the window, and these queues are sorted from smallest to largest based on their remaining deadline_remain. When outside the window or when the minimum reservation share for a critical TSN queue has been met, services are allocated using a round-robin approach with dynamic weights. For bursty flows, token bucket / leaky bucket constraints are implemented in conjunction with the integer parameters in the current executable policy_Ver. When the latency of a critical TSN queue approaches the threshold or a conflict warning has not been resolved, the service ratio of the critical TSN queue is increased or the strongest preemption is triggered. At the end of each scheduling slot / cycle, the selected data packet or transmission descriptor, along with the current executable policy Policy_Ver, queue service statistics, and exception flags, is entered into S6.

[0021] The beneficial effects of the above-mentioned further solutions are as follows: by distinguishing between time window states and non-window states for scheduling control, critical time-sensitive network (TSN) flows are given strict priority within the window, while resources are reasonably released to non-critical services outside the window; by combining remaining deadline sorting, dynamic weighted round-robin, and burst flow shaping mechanisms, a balance between deterministic guarantees and resource utilization efficiency is achieved; when latency approaches the threshold or conflicts persist, the system supports increasing the service ratio or triggering preemption, enhancing the system's robustness to extreme load scenarios.

[0022] Furthermore, step S6 includes the following steps: Get the send descriptor and the current executable policy, Policy_Ver; At the start of each transmission cycle, based on the window state and non-window state determined by the equivalent window boundaries W_start to W_end and the current time base, a window consistency check is performed on each item in the queue to be transmitted: if it belongs to a critical time-sensitive network TSN flow that must be transmitted within the window, it is only allowed to be transmitted within the window; if it is predicted that the transmission will cross the end of the window, it will be transmitted early, temporarily stored in the next window, or trigger a degradation process according to the current executable policy Policy_Ver; for ordinary service flows, transmission is carried out according to the quota and shaping results of the current executable policy Policy_Ver; when the port is busy or the buffer usage increases, rate limiting, batch transmission, or temporary storage is performed according to the current executable policy Policy_Ver, and forwarding is continuously updated. If necessary, the remaining transmission descriptors are iteratively processed within the same window cycle until the window ends or the current batch is completed. Record execution evidence during the process of deterministic forwarding to the Time-Sensitive Network (TSN) and forwarding / backhauling to the 5G side; The execution evidence is summarized to form a set of key performance indicators (KPIs) for this cycle. The set of key performance indicators (KPIs) for this cycle is then bound to the decision version number Input_Ver, the current executable policy Policy_Ver, and the window cycle number Cycle_ID to generate a feedback report package that can be used for closed-loop optimization. Based on the generated feedback report package, when the packaging boundary is reached, the execution records within this boundary are encapsulated into a record package and written to the buffer in an atomic manner, and the record package ready flag is set; wherein, the record package includes the package identifier record Record_ID, the window cycle number Cycle_ID, the decision version number Input_Ver, the current executable policy Policy_Ver, and the key state summary; Once the record packet is detected as ready, a consistency check is performed. If the check passes, proceed to S7.

[0023] The beneficial effects of the above-mentioned further solutions are: complete execution evidence is recorded during deterministic forwarding and wireless-side forwarding, and it is summarized into a set of key performance indicators (KPIs) that are strongly bound to the policy version, decision version, and window period, so as to realize the whole process of scheduling behavior is observable and traceable; the atomic writing and consistency verification mechanism of record packets is adopted to ensure the integrity and reliability of feedback data, providing a highly reliable data foundation for subsequent closed-loop optimization, policy verification, and problem localization.

[0024] Furthermore, S7 includes the following steps: Collect key performance indicators (KPIs) and system resource status at the granularity of queue / business flow / port / window period; The collected key performance indicators (KPIs) are bound to the executable policy (Policy_Ver), decision version number (Input_Ver), and window period number (Cycle_ID) of this round to form a feedback report package; Based on the feedback report packet, if the critical time-sensitive network TSN flow exceeds the window / expiration period or the window utilization rate is abnormally high and conflicts occur frequently, it is determined to be a policy mismatch or insufficient resources, and a reordering needs to be triggered; if the critical business stably hits the window and the congestion is controllable, it is determined to be a fine-tuning optimization, and the weight / quota / integrated rate increment and hysteresis threshold adjustment strategy are output to complete the scheduling of dynamic data packets.

[0025] The beneficial effects of the above-mentioned further solutions are: continuous evaluation of key performance indicators (KPIs) and system resource status enables online judgment of policy execution effectiveness; when key TSN flow out-of-window, overdue, or frequent window conflicts are detected, policy reordering or resource adjustment can be triggered in a timely manner; by outputting fine-tuning parameters and hysteresis thresholds, smooth optimization is achieved, avoiding unnecessary full recalculation, thereby improving the overall scheduling efficiency, adaptability, and long-term operational stability of the system.

[0026] The present invention also provides a dynamic data packet scheduling system, comprising: The real-time traffic awareness module is used to identify and count the traffic flow of data packets from the 5G network side and the time-sensitive network (TSN) side, and form a traffic status snapshot. The wireless state awareness module is used to collect dynamic information of wireless links on the 5G network side, transform the dynamic information into wireless risk factors that directly affect scheduling, bind the wireless risk factors with traffic state snapshots to form the wireless part of the joint input frame, and perform time alignment processing. The system status monitoring module is used to continuously monitor the resource and congestion status of DS-TT itself; The TSN time constraint module is used to obtain the periodicity and window boundary of Time-Sensitive Network (TSN) services, and output conflict warning information by calculating the remaining deadline and constraint level. The scheduling controller (SC) is used to integrate traffic state snapshots, radio risk factors, and time constraints of time-sensitive networks (TSN) to generate a complete policy of weight, quota, dequeue order, in-window transmission plan, and conflict avoidance. The strategy adjustment module is used to perform parameter adaptation and threshold hysteresis control on the complete strategy; The queue scheduling module is used to optimize and protect critical time-sensitive network (TSN) services within a window based on the threshold hysteresis control results and the currently effective executable policies. The data forwarding module is used to perform data forwarding and gated transmission based on the optimized protection processing results; The scheduling execution and feedback module is used to statistically analyze the data forwarding and gating transmission results, generate the feedback report required for closed-loop scheduling, and perform parameter adaptation or trigger re-decision to complete the scheduling of dynamic data packets.

[0027] The beneficial effects of this invention are: This invention provides a dynamic packet scheduling method system based on DS-TT (Distributed TSN Translator) nodes in a 5G TSN converged network. Addressing the shortcomings of existing DS-TT static scheduling algorithms, such as difficulty adapting to 5G wireless channel variations, lack of cross-network collaborative scheduling between 5G and TSN, simple scheduling mechanisms that do not support real-time traffic awareness, and insufficient consideration of time constraints for time-sensitive network (TSN) services, this invention introduces wireless state awareness, real-time traffic awareness, and TSN time constraint resolution mechanisms on the DS-TT side. Through dynamic scheduling decision-making and scheduling execution feedback, a closed-loop control is formed, achieving time-aware collaborative scheduling across 5G and TSN networks. This invention can reduce the impact of wireless fluctuations on the latency of high-priority packets, reduce scheduling conflicts, improve resource utilization efficiency, and ensure that critical TSN services are forwarded within a specified time window. It is applicable to deterministic communication scenarios such as smart manufacturing, rail transit, and power automation. This invention has at least the following beneficial effects: By using dynamic scheduling algorithms to adapt to changes in 5G wireless channels, the latency risk of high-priority data packets in wireless fluctuation scenarios can be reduced. To achieve coordinated scheduling between 5G networks and Time-Sensitive Networks (TSNs), reduce scheduling conflicts, and improve network resource utilization efficiency; Introducing a real-time traffic awareness mechanism enables DS-TT to dynamically adjust its scheduling strategy based on current service load and queue status; By leveraging the time-constrained characteristics of Time-Sensitive Network (TSN) services, time-aware scheduling is implemented to ensure that critical service data packets are forwarded within the specified time window. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the system structure of the present invention.

[0029] Figure 2 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0030] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0031] Example 1 like Figure 1 As shown, the present invention provides a dynamic data packet scheduling system, comprising: The real-time traffic awareness module is used to identify and count the traffic flow of data packets from the 5G network side and the time-sensitive network (TSN) side, and form a traffic status snapshot. The wireless state awareness module is used to collect dynamic information of wireless links on the 5G network side, transform the dynamic information into wireless risk factors that directly affect scheduling, bind the wireless risk factors with traffic state snapshots to form the wireless part of the joint input frame, and perform time alignment processing. The system status monitoring module is used to continuously monitor the resource and congestion status of DS-TT itself; The TSN time constraint module is used to obtain the periodicity and window boundary of Time-Sensitive Network (TSN) services, and output conflict warning information by calculating the remaining deadline and constraint level. The scheduling controller (SC) is used to integrate traffic state snapshots, radio risk factors, and time constraints of time-sensitive networks (TSN) to generate a complete policy of weight, quota, dequeue order, in-window transmission plan, and conflict avoidance. The strategy adjustment module is used to perform parameter adaptation and threshold hysteresis control on the complete strategy; The queue scheduling module is used to optimize and protect critical time-sensitive network (TSN) services within a window based on the threshold hysteresis control results and the currently effective executable policies. The data forwarding module is used to perform data forwarding and gated transmission based on the optimized protection processing results; The scheduling execution and feedback module is used to statistically analyze the data forwarding and gating transmission results, generate the feedback report required for closed-loop scheduling, and perform parameter adaptation or trigger re-decision to complete the scheduling of dynamic data packets.

[0032] In this embodiment, the present invention is a dynamic packet scheduling system architecture based on 5G TSN DS-TT, deployed on the DS-TT node side, including: The wireless state awareness module is used to acquire wireless link status and scheduling information from the 5G protocol stack / baseband side and generate wireless jitter and latency evaluation parameters; the real-time traffic awareness module is used to monitor the arrival characteristics, queue length, queuing latency, congestion level and other status information of the service flow entering DS-TT; the TSN time constraint parsing module is used to parse the period parameters, time windows, deadlines and other constraints of time-sensitive network (TSN) services and form a time constraint model that can be used for scheduling; the dynamic scheduling decision module is used to integrate wireless state, traffic state and time-sensitive network (TSN) data.

[0033] In this embodiment, the system architecture used in this invention is geared towards the DS-TT node of a 5G-TSN converged network, constructing a dynamic data packet scheduling closed-loop system of "perception-decision-execution-feedback". The system uniformly accesses service data and status information from the time-sensitive network (TSN) and the 5G network through a configuration management interface module, converting inputs from different network sides into scheduling objects that can be uniformly processed within the DS-TT, avoiding conflicts and resource waste caused by inconsistent scheduling rhythms between the two sides.

[0034] like Figure 1 As shown, the real-time traffic awareness module (TM) identifies and counts incoming data packets to form traffic characteristics such as queue length, queuing delay, arrival rate and burstiness, and provides a stable and consistent "traffic status snapshot" to the upper level through state sharing and caching mechanisms. The Wireless Status Aware Module (WM) collects link quality and scheduling rhythm on the 5G wireless side, assesses latency jitter and quantifies it into wireless risk information, enabling scheduling to dynamically adapt to changes in the wireless channel. The System Status Monitoring Module (SM) continuously monitors the DS-TT's own resource and congestion status, such as queue congestion trends, CPU and port load, thereby preventing the strategy from failing when the calculation is correct but the execution resources are insufficient. The TSN Time Constraint Module (TCM) parses the period T, window boundaries (W_start~W_end), deadline, and priority constraints of the Time-Sensitive Network (TSN) flow, and outputs window pressure and conflict warnings, providing a time base and constraints for deterministic assurance. The Scheduling Controller (SC) is responsible for fusing, aligning, and orchestrating multi-source information from modules TM / WM / SM / TCM, and driving the Dynamic Scheduling Decision Module (DPM) to generate cross-network collaborative scheduling strategies. The DPM integrates wireless risks, queue congestion, and time constraints, calculates queue weights and quotas, forms an in-window transmission plan, and provides conflict avoidance strategies (such as critical flow reservation, peak shifting, suppressing non-critical flows, and shaping), thereby simultaneously satisfying the determinism of Time-Sensitive Network (TSN) and the dynamism of 5G. The strategy adjustment module (ADJ) performs parameter adaptation and threshold / hysteresis control on the output of the dynamic scheduling decision module (DPM) to suppress frequent strategy jitter caused by short-term fluctuations, so that the strategy can remain stable and converged in the dynamic environment. The adjusted strategy is then sent back to the scheduling controller (SC) for unified distribution and version management. The Queue Scheduling (QSM) module performs dequeue order selection, weighted round-robin, and preemption control according to the policy to ensure that critical time-sensitive network TSN services within the window have priority to obtain forwarding opportunities; The data forwarding module (FWD) performs gated transmission and deterministic forwarding according to the transmission plan, enabling the output of deterministic traffic to the time-sensitive network (TSN) side, while also supporting forwarding and backhaul to the 5G side; The scheduling execution and feedback module (EXFB) performs statistics and evaluation on the actual execution results (such as actual latency, windowing, drop, number of conflicts, window utilization, etc.) and sends the feedback loop back to the scheduling controller (SC) to trigger the policy adjustment module (ADJ) to fine-tune or drive the dynamic scheduling decision module (DPM) to re-plan, thereby forming a traceable and verifiable closed-loop optimization mechanism to continuously improve the determinism and resource utilization efficiency of the end-side 5G-TSN converged network.

[0035] In summary, this invention introduces wireless state awareness, real-time traffic awareness, and TSN time constraint resolution mechanisms on the DS-TT side, and forms a closed-loop control through dynamic scheduling decision-making and scheduling execution feedback to achieve time-aware collaborative scheduling across 5G networks and time-sensitive network TSNs. This invention can reduce the impact of wireless fluctuations on the latency of high-priority data packets, reduce scheduling conflicts and improve resource utilization efficiency, and ensure that critical time-sensitive network TSN services are forwarded within the specified time window. It is suitable for deterministic communication scenarios such as smart manufacturing, rail transit, and power automation.

[0036] Example 2 Based on the module division of labor and data flow shown in the above system block diagram, such as Figure 2 As shown, the method flow of this invention can be further summarized into the following seven steps: S1, real-time traffic awareness and queue state construction; S2, wireless state awareness, jitter risk quantification, and wireless risk factor R generation; S3, time reduction and conflict warning generation for Time Sensitive Network (TSN); S4, dynamic scheduling decision generation and policy issuance; S5, queue scheduling execution and window priority protection; S6, data forwarding and gating transmission / scheduling; S7, scheduling execution and feedback update and triggering policy adjustment, specifically as follows: S1. After unifying the access of data packets from the 5G network side and the Time-Sensitive Network (TSN) side, a traffic status snapshot is formed by classifying service flows, calculating queue length and queuing latency. The implementation method is as follows: Acquire data packets from the 5G network side and the Time-Sensitive Network (TSN) side; Data packets are classified and identified according to preset service identification rules, and the data packets are mapped to queue groups. Statistics are updated at the granularity of queue / service flow, and the statistical results are saved to the state sharing and buffer area. The statistics include queue length, average and percentile queuing delay estimates, arrival rate, burstiness, drop count, reorder count, and critical flow percentage. The identification fields include 5-tuple, VLAN PCP / DSCP, Stream ID of Time Sensitive Network (TSN), service tag, and flow mapping table issued by CNC / local policy. Based on statistics, when the scheduling cycle boundary or the trigger event threshold is reached, the current statistical value is solidified into a traffic status snapshot. The traffic status snapshot includes the congestion level of each queue, queuing delay estimate, number and proportion of critical flows, burstiness index and snapshot timestamp T_snap. After the traffic status snapshot is generated, write the snapshot sequence number Snapshot_ID and the snapshot timestamp T_snap, and set the traffic snapshot valid flag. After reading the valid flag of the traffic snapshot using the scheduling controller SC, the snapshot sequence number Snapshot_ID is locked as the decision input for this round. At the same time, the read and write versions of the corresponding snapshot are frozen, thus completing the formation of the traffic status snapshot. If S1 generates a new traffic status snapshot before S2 is completed, it will only enter the pending queue and will not replace the currently locked traffic status snapshot, ensuring the consistency of the decision cycle.

[0037] In this embodiment, the Interface Management Module (IFM) uniformly accesses inbound data packets from the 5G network side and the Time-Sensitive Network (TSN) side, and then hands them over to the Real-Time Traffic Awareness Module (TM) to complete service flow classification, statistics, and status caching, providing a quantifiable "traffic profile" for subsequent scheduling and control. The Real-Time Traffic Awareness Module (TM) first classifies and identifies data packets according to preset service identification rules. Identification fields may include a five-tuple, VLAN PCP / DSCP, TSN Stream ID, service tag, and the flow mapping table issued by CNC / local policies. The TM then maps the data packets to corresponding logical queues or queue groups (critical TSN queues, real-time queues, ordinary queues, etc.). For each inbound data packet, the Real-Time Traffic Awareness Module (TM) writes high-precision inbound timestamp, packet length, flow ID, basic priority, and queue index, etc., to support subsequent queuing latency and urgency calculations. Meanwhile, the Real-Time Traffic Awareness Module (TM) continuously updates statistics at the "queue / service flow" granularity: queue length (number of packets / bytes), average and percentile queuing delay estimates, arrival rate, burst rate, drop count, reordering count, and critical flow percentage, etc., and saves these statistical results in the state sharing and buffer area for low-overhead reading by the scheduling controller (SC). To ensure data consistency within the same decision cycle, S1 introduces a "snapshot freezing mechanism": when the scheduling cycle boundary is reached (such as 1ms or the configuration cycle) or an event threshold is triggered (queue depth exceeds the threshold, queuing delay approaches the upper limit, burst rate is abnormal), the Real-Time Traffic Awareness Module (TM) freezes the current statistical values ​​into a "traffic status snapshot". The snapshot includes at least the congestion level of each queue, queuing delay estimate, number and percentage of critical flows, burst rate index, and snapshot timestamp.

[0038] S1→S2 Transition Process: After a traffic snapshot is generated, the Real-Time Traffic Aware Module (TM) writes the snapshot sequence number Snapshot_ID and the snapshot timestamp T_snap, and sets the "Traffic Snapshot Valid" flag. The Scheduling Controller (SC) reads this flag and locks the snapshot sequence number Snapshot_ID as the input for this round of decision-making. Simultaneously, it freezes the read / write versions of the corresponding snapshot (double buffering / multi-version circular cache) to prevent it from being overwritten by new statistics during S2. Subsequently, the Scheduling Controller (SC) triggers the Wireless Status Aware Module (WM) to collect wireless status data for the same period and perform time alignment based on the snapshot sequence number Snapshot_ID. If a new traffic snapshot is generated before S2 is completed, it only enters the "pending queue" and does not replace the currently locked snapshot, ensuring consistency in the decision-making cycle.

[0039] S2. Collect dynamic information of the wireless link from the 5G network side, transform the dynamic information into a wireless risk factor that directly affects scheduling, and bind the wireless risk factor with the traffic state snapshot to form the wireless part of the joint input frame, and perform time alignment processing; wherein, the wireless risk factor carries both a sampling timestamp and a validity period, and the implementation method is as follows: The system collects dynamic information about wireless links on the 5G network side. This dynamic information includes link quality level, scheduling rhythm information, wireless side buffer / queue backlog, and end-to-end wireless transmission latency statistics for the past N cycles. Sliding window filtering and threshold hysteresis determination are applied to dynamic information indicators; Based on the judgment result, the dynamic information is transformed into a wireless risk factor R that directly affects the scheduling. The wireless risk factor R carries both the sampling timestamp T_wireless and the validity period. Bind the wireless risk factor R with the snapshot sequence number Snapshot_ID to form the wireless part of the joint input frame: Snapshot_ID, T_snap, R, T_wireless; The scheduling controller (SC) collects the wireless status of the same period and performs time alignment verification based on the snapshot sequence number Snapshot_ID. If |T_wireless If T_snap| exceeds the threshold, a supplementary sampling is triggered: if the supplementary sampling fails, the wireless risk factor R from the previous cycle is used and the R_use status is marked; if the wireless side interface is abnormal, the scheduling controller SC enters the degraded state and generates an available risk factor R; after completing the binding and time alignment, it enters S3.

[0040] In this embodiment, in S2, the Wireless State Aware Module (WM) collects dynamic information of the wireless link from the 5G network side and converts it into a "wireless risk factor R" that can be directly applied to scheduling to solve the problem that static scheduling cannot adapt to changes in the wireless channel. The Wireless State Aware Module (WM) reads wireless state input through an interface with the 5G protocol stack / baseband module, including but not limited to: link quality level (which can be abstracted from CQI / BLER trends), scheduling rhythm information (e.g., resource allocation cycle, TTI beat changes), wireless side buffer / queue backlog, and end-to-end wireless transmission delay statistics (mean, variance, quantiles, range, etc.) for the past N cycles. To avoid frequent fluctuations in the scheduling strategy due to short-term jitter, the Wireless State Aware Module (WM) performs sliding window filtering and threshold hysteresis judgment on the above indicators: the risk level is increased only when it exceeds the threshold K times consecutively, and the risk level is decreased only when it falls back K times consecutively; amplitude limiting or median filtering can be used to suppress abnormal spikes. The Wireless State Aware Module (WM) then quantifies the wireless state into a risk factor R, which can be broken down into a set of executable parameters, such as: critical service protection coefficient (increasing the weight of critical queues and allowing a higher probability of preemption), low priority suppression coefficient (reducing the quota of ordinary queues and strengthening shaping), and queuing threshold tightening coefficient (shortening the allowed waiting time for critical flows), etc., which are easy for SC / DPM to directly reference. The risk factor R also carries a sampling timestamp T_wireless and an expiration date to avoid expired states misleading decisions.

[0041] S2→S3 transition process: After generating the risk factor R, the Wireless State Aware Module (WM) binds it with the snapshot sequence number Snapshot_ID locked in the previous stage to form the radio part of the "joint input frame" (Snapshot_ID, T_snap, R, T_wireless). The Scheduler Controller (SC) performs a time alignment check after receiving the data: if |T_wireless... If T_snap exceeds the threshold (e.g., exceeding one scheduling cycle), a supplementary snapshot is triggered; if the supplementary snapshot fails, the risk factor R from the previous cycle is retained and marked as "R_Retained". If a radio-side interface anomaly prevents the status from being obtained, the scheduling controller SC enters a conservative degradation phase: temporarily increasing critical flow protection and suppressing non-critical quotas, while still generating an available risk factor R to ensure process continuity. After binding and alignment are completed, the scheduling controller SC enters S3, triggering the TSN Time Constraint Module (TCM) to resolve the window and deadline, and also binding the result to the snapshot sequence number Snapshot_ID, forming a complete decision input.

[0042] S3. Obtain the periodicity and window boundary of Time-Sensitive Network (TSN) services, calculate the remaining deadline and constraint level, and output conflict warning information. The implementation method is as follows: Read the key parameter set for each Time-Sensitive Network (TSN) stream. The key parameter set includes the period T, the gating control list (GCL) or equivalent window boundaries W_start to W_end, the maximum available time slot / bandwidth per period, and service level and priority constraint rules. Among them, W_start represents the start time of the window in which the TSN stream is allowed to start sending data within a scheduling period, corresponding to the start time of the gate opening in the gating control list, or the lower boundary of the sending time window obtained by equivalent conversion from the gating control list. W_end represents the end time of the window in which the TSN stream must complete sending within a scheduling period, corresponding to the end time of the gate closing in the gating control list, or the upper boundary of the sending time window obtained by equivalent conversion from the gating control list. Construct a mapping relationship between the Stream ID of the Time-Sensitive Network (TSN) and the local queue / service flow identifier, so that the TSN packets identified in S1 are associated with the corresponding time constraint entries; Based on the mapping results, the effective window boundary within the current scheduling period is calculated using the local unified time base, and the remaining deadline (deadline_remain) is calculated for each time-sensitive network (TSN) data packet to be scheduled. At the same time, the constraint level is output for the queue level. Calculate the window pressure index: Given the remaining capacity of the current window, can the total bytes of the critical packets to be sent be completed within the window? If insufficient or multiple critical flow windows overlap and compete, a conflict warning message is output. After outputting (T, W_start / W_end, deadline_remain, constraint level, conflict warning), the scheduling controller (SC) encapsulates it along with the traffic status snapshot and the radio risk factor R into a decision input frame, and binds them uniformly to the same snapshot sequence number Snapshot_ID and decision version number Input_Ver. Based on the encapsulation result, a consistency check is performed: if a Time-Sensitive Network (TSN) configuration is missing, the window cannot be calculated, the time base is abnormal, or the mapping is missing, a constraint anomaly flag is set and a security policy is adopted. After completing the consistency check, the scheduling controller (SC) freezes the decision version number Input_Ver and generates a policy that can be traced back to the specific input version.

[0043] In this embodiment, in S3, the TSN Time Constraint Module (TCM) outputs the periodicity, windowing, and deadline requirements of Time-Sensitive Network (TSN) services in a structured manner, enabling DS-TT to have time-aware scheduling capabilities. The TSN Time Constraint Module (TCM) reads the key parameter set for each TSN stream from the CNC-issued configuration or the local configuration table, including: period T, gating control list GCL or equivalent window boundary (W_start to W_end), maximum available time slot / bandwidth per period, and service level and priority constraint rules. The TSN Time Constraint Module (TCM) first establishes a mapping relationship between the TSN Stream ID and the local queue / service stream identifier, enabling the TSN packets identified in S1 to be associated with the corresponding time constraint entries. Subsequently, the TSN Time Constraint Module (TCM) calculates the effective window boundary within the current scheduling period based on a local unified time base (such as gPTP time or a time base mapping provided by 5G), and calculates the remaining deadline_remain (the remaining time from the current moment until the forwarding must be completed) for each data packet to be scheduled by the TSN Time Constraint Module (TCM). Simultaneously, it outputs a "constraint level" (must be sent within the window, allow cross-window delay, or discard / degrade if out of window, etc.) at the queue level. The TSN Time Constraint Module (TCM) also calculates a "window pressure index": whether the total bytes of the critical packets to be sent can be completed within the window given the remaining capacity of the current window; if insufficient or multiple critical flow windows overlap and compete, it outputs a "conflict warning message" to prompt subsequent strategies to enable conflict avoidance measures such as peak shifting, reservation, or suppression of non-critical flows.

[0044] S3→S4 Transition Process: After the TSN Time Constraint Module (TCM) outputs (T, W_start / W_end, deadline_remain, constraint level, conflict warning), the Scheduling Controller (SC) encapsulates it together with the traffic status snapshot from S1 and the radio risk factor R from S2 into a "decision input frame," and binds them uniformly to the same snapshot sequence number Snapshot_ID and decision version number Input_Ver. Before entering S4, the SC performs a consistency check: if it finds that the Time Sensitive Network (TSN) configuration is missing, the window cannot be calculated, the time base is abnormal, or the mapping is missing, it sets the "constraint abnormality" flag and adopts a safety policy (temporarily promoting the relevant flow to a conservative critical flow and restricting the dequeueing of ordinary flows near the window) to ensure that critical services are not mis-scheduled. After completing the check, the SC freezes the decision version number Input_Ver and calls the Dynamic Scheduling Decision Module (DPM) to ensure that the policy generated by the Dynamic Scheduling Decision Module (DPM) is traceable to the specific input version.

[0045] S4. Integrating traffic status snapshots, wireless risk factors, and Time-Sensitive Network (TSN) time constraints, a complete policy is generated, comprising weights, quotas, dequeue order, in-window transmission plans, and conflict avoidance. Based on this complete policy, parameter adaptation and threshold hysteresis control are performed. The TSN time constraints include period, window, deadline, and conflict warning, implemented as follows: The traffic status, radio risk factor R, resource status, and time constraints of the time-sensitive network (TSN) are used to aggregate the decision version number Input_Ver of the scheduling controller (SC). Based on the convergence results, the overall scheduling priority P is calculated at the granularity of queue / business flow; Based on the comprehensive scheduling priority P, output a complete strategy including weight, quota, dequeue order, in-window sending plan, and conflict avoidance. For the complete policy, apply threshold and hysteresis rules to generate an executable policy Policy_Ver; Write the executable policy Policy_Ver to the policy double buffer. First, write it to the spare policy area along with (Input_Ver, Policy_Ver, effective boundary time). Then, publish the new policy availability flag through an atomic switch. Based on the write results, the system switches to the new executable policy Policy_Ver only when the effective boundary is detected, thus completing threshold hysteresis control.

[0046] In this embodiment, in S4, the scheduling controller SC fuses multi-source inputs from the real-time traffic awareness module (TM), the wireless status awareness module (WM), the system status monitoring module (SM), and the TSN time constraint module (TCM), and drives the dynamic scheduling decision module (DPM) to generate a complete policy of "weight + quota + dequeue order + in-window transmission plan + conflict avoidance". At the same time, the policy adjustment module (ADJ) completes parameter adaptation and threshold hysteresis control to avoid frequent policy jitter. First, the scheduling controller SC aggregates the traffic status (queue congestion, queuing delay, burstiness) corresponding to the decision version number Input_Ver, the wireless risk factor R, the resource status (queue congestion trend, CPU and port load, processing bottleneck prompts) provided by the system status monitoring module (SM), and the TSN time constraints (period, window, deadline, conflict warning). Subsequently, the Dynamic Scheduling Decision Module (DPM) calculates the overall scheduling priority P at the queue / service flow granularity. The overall scheduling priority P is composed of basic priority, deadline urgency (a smaller deadline_remain value indicates greater urgency), queue congestion (queue latency / queue depth), and wireless risk correction terms (increasing R strengthens critical protection and suppresses non-critical traffic). Based on this, the DPM outputs queue weights and service quotas, in-window transmission plans (the transmission order and minimum reserved share of critical time-sensitive network (TSN) flows within the window), and conflict avoidance strategies (off-peak serialization, prohibiting ordinary flows from leaving the queue at window edges, reserving time slots for critical flows, or strengthening the shaping of bursty flows). The Policy Adjustment Module (ADJ) outputs application thresholds and hysteresis rules to the DPM: for example, the wireless risk level must change continuously K times before triggering a significant adjustment; after window pressure is relieved, the suppression coefficient is gradually rolled back to avoid "sudden tightening and loosening." Finally, an executable policy Policy_Ver is generated (including validity period, gating parameters, shaping parameters, preemption switch, etc.).

[0047] S4→S5 transition process: The scheduling controller (SC) writes the policy into the "policy double buffer," first writing it to the backup policy area along with (Input_Ver, Policy_Ver, effective boundary time), and then publishing the "new policy available" flag via atomic switching. The queue scheduling module (QSM) only switches to the new executable policy Policy_Ver when it detects the effective boundary (such as the next 1ms cycle or window boundary), avoiding out-of-order and jitter caused by partial updates. If the system status monitoring module (SM) reports a sudden increase in CPU / port load or a large deviation between the queue status and the decision version number Input_Ver, the scheduling controller (SC) can trigger a rapid re-decision: either fine-tune the quota and integer parameters (still using the same executable policy Policy_Ver sub-version), or return to the dynamic scheduling decision module (DPM) to recalculate and republish the policy before entering S5 for execution.

[0048] S5. Based on the threshold hysteresis control results and according to the currently effective executable policies, optimize and protect critical time-sensitive network (TSN) services within the window, specifically as follows: Based on the current executable policy_Ver, scheduling is differentiated between windowed and non-windowed states. Specifically, when within a Time-Sensitive Network (TSN) window, priority is given to critical TSN queues that are allowed to send within the window, and these queues are sorted from smallest to largest based on their remaining deadline_remain. When outside the window or when the minimum reservation share for a critical TSN queue has been met, services are allocated using a round-robin approach with dynamic weights. For bursty flows, token bucket / leaky bucket constraints are implemented in conjunction with the integer parameters in the current executable policy_Ver. When the latency of a critical TSN queue approaches the threshold or a conflict warning has not been resolved, the service ratio of the critical TSN queue is increased or the strongest preemption is triggered. At the end of each scheduling slot / cycle, the selected data packet or transmission descriptor, along with the current executable policy Policy_Ver, queue service statistics, and exception flags, is entered into S6.

[0049] In this embodiment, in S5, the Queue Scheduling Module (QSM) performs dequeue order selection, weighted round-robin scheduling, and preemption control based on the currently effective scheduling policy Policy_Ver, and implements the policy as a sequence of transmittable data packets or transmission descriptors. The Queue Scheduling Module (QSM) distinguishes between "window state" and "non-window state" scheduling: when within the Time Sensitive Network (TSN) window (based on W_start / W_end calculated by the TSN Time Constraint Module (TCM) and the gating parameters), the Queue Scheduling Module (QSM) prioritizes the selection of critical TSN queues that are allowed to be transmitted within the window, and sorts them in ascending order according to the remaining deadline_remain, combined with the transmission plan output by the Dynamic Scheduling Decision Module (DPM). If necessary, the preemption mechanism is activated to prioritize the dequeueing of critical packets; when outside the window or after the critical queue has met the minimum reservation quota, the Queue Scheduling Module (QSM) allocates services among ordinary queues according to dynamic weighted round-robin scheduling, while adhering to the quota limit and suppression coefficient to prevent non-critical flows from crowding out critical windows. For bursty flows, the Queue Scheduling Module (QSM) can use the shaping parameters in the policy to implement token bucket / leaky bucket constraints, smoothing the dequeue rate and reducing the risk of instantaneous congestion and collisions. When the queuing delay of a critical queue approaches the threshold or the collision warning has not been lifted, the Queue Scheduling Module (QSM) can temporarily increase the service ratio of the critical queue or trigger stronger preemption (limited by the executable policy Policy_Ver). The output of the Queue Scheduling Module (QSM) includes not only the dequeued data packet body but also a "transmission descriptor" (destination port, gating time, priority flag, whether cross-window is allowed, whether temporary storage is required, etc.) for the next forwarding module to execute precisely.

[0050] S5→S6 Transition Process: At the end of each scheduling slot / cycle, the Queue Scheduling Module (QSM) submits the selected data packets or transmission descriptors in batches to the Data Forwarding Module (FWD), carrying the current executable policy (Policy_Ver), queue service statistics (number of bytes / packets served by each queue in this cycle), and exception flags (whether preemption occurred, whether reshaping was triggered, and whether there is a risk of cross-window transmission). Upon receiving the data, the Data Forwarding Module (FWD) performs gating verification and transmission scheduling. If the Data Forwarding Module (FWD) reports "insufficient window / cross-window risk" or port busy, the Queue Scheduling Module (QSM) will adjust subsequent selections according to policy instructions (e.g., reducing dequeueing from ordinary queues and increasing temporary storage), but will not directly change the executable policy (Policy_Ver) to avoid spontaneous drift on the execution side. After the handover is completed, the module enters S6 to execute gating forwarding, and completes scheduling execution and feedback closed-loop updates at the end of S6.

[0051] S6. Based on the optimized protection processing results, perform data forwarding and gated transmission. The implementation method is as follows: Get the send descriptor and the current executable policy, Policy_Ver; At the start of each transmission cycle, based on the window state and non-window state determined by the equivalent window boundaries W_start to W_end and the current time base, a window consistency check is performed on each item in the queue to be transmitted: if it belongs to a critical time-sensitive network TSN flow that must be transmitted within the window, it is only allowed to be transmitted within the window; if it is predicted that the transmission will cross the end of the window, it will be transmitted early, temporarily stored in the next window, or trigger a degradation process according to the current executable policy Policy_Ver; for ordinary service flows, transmission is carried out according to the quota and shaping results of the current executable policy Policy_Ver; when the port is busy or the buffer usage increases, rate limiting, batch transmission, or temporary storage is performed according to the current executable policy Policy_Ver, and forwarding is continuously updated. If necessary, the remaining transmission descriptors are iteratively processed within the same window cycle until the window ends or the current batch is completed. Record execution evidence during the process of deterministic forwarding to the Time-Sensitive Network (TSN) and forwarding / backhauling to the 5G side; The execution evidence is summarized to form a set of key performance indicators (KPIs) for this cycle. The set of key performance indicators (KPIs) for this cycle is then bound to the decision version number Input_Ver, the current executable policy Policy_Ver, and the window cycle number Cycle_ID to generate a feedback report package that can be used for closed-loop optimization. Based on the generated feedback report package, when the packaging boundary is reached, the execution records within this boundary are encapsulated into a record package and written to the buffer in an atomic manner, and the record package ready flag is set; wherein, the record package includes the package identifier record Record_ID, the window cycle number Cycle_ID, the decision version number Input_Ver, the current executable policy Policy_Ver, and the key state summary; Once the record packet is detected as ready, a consistency check is performed. If the check passes, proceed to S7.

[0052] In this embodiment, in S6, the data forwarding module (FWD) receives the "transmission descriptor" (containing information such as target port, queue source, packet priority, allowed transmission window, policy quota, whether cross-window is allowed, and whether reshaping / temporary storage is required) output by the queue scheduling module (QSM) and the gating parameters of the currently effective policy, performs planned gating transmission and deterministic forwarding, and submits the execution result to the scheduling execution and feedback module (EXFB) for statistical analysis, providing a traceable basis for subsequent policy adjustments. At the start of each transmission micro-cycle, the Data Forwarding Module (FWD) determines the "inside / outside window" status based on the time window boundaries (W_start~W_end) provided by the TSN Time Constraint Module (TCM) and the current time base. It then performs a window consistency check on each packet in the queue to be transmitted: if it belongs to a critical time-sensitive network TSN flow that must be transmitted within the window, it is only allowed to be transmitted within the window; if transmission is predicted to exceed the end of the window (e.g., insufficient remaining window margin or port busy causing queuing), it executes a strategy to send early, temporarily store in the next window, or trigger degradation processing (e.g., marking risks, discarding and counting explicitly expired packets). For ordinary service flows, the Data Forwarding Module (FWD) transmits according to policy quotas and shaping results to avoid conflicts and jitter at the window edge; when port busy or buffer usage increases, the Data Forwarding Module (FWD) executes rate limiting, batch transmission, or temporary storage according to policy, and continuously updates the forwarding context such as "remaining window time, remaining quota, port busy / idle, and buffer usage," iteratively processing the remaining transmission descriptors within the same window cycle when necessary, until the window ends or the current batch is completed. During the deterministic forwarding to the Time-Sensitive Network (TSN) side and the forwarding / backhaul to the 5G side, the Data Forwarding Module (FWD) records key execution evidence, including dequeue time, actual transmission time, port waiting time, gating judgment results (in-window transmission / cross-window temporary storage / degradation / dropping), number of temporary storage and reordering operations, number of bytes sent within the window, and window utilization. Subsequently, the Schedule Execution and Feedback Module (EXFB) merges and summarizes the execution records from the Data Forwarding Module (FWD) to form a set of KPIs for the current cycle, such as actual end-to-end latency and jitter, number of window crossings, number of drops, number of collisions, window utilization, queue congestion changes, and CPU / port load changes. The KPIs are then bound to the input decision version number Input_Ver, policy version Policy_Ver, and window cycle number Cycle_ID to generate a "feedback report package" that can be used for closed-loop optimization.

[0053] S6→S7 Transition Process: When the data forwarding module (FWD) reaches the packet boundary (window ends, policy validity period reaches, or batch sending descriptor processing is completed / critical exception is triggered), it encapsulates the execution records within this boundary into a sending record packet, including the packet identifier record Record_ID, window cycle number Cycle_ID, decision version number Input_Ver, current executable policy Policy_Ver, and key status summary (port busy rate, cache usage, window pressure flag, etc.), and atomically writes it to the input buffer of the scheduling execution and feedback module (EXFB) and sets the "record packet ready" flag. After the scheduling execution and feedback module (EXFB) detects this flag, it performs a consistency check (version number matching, cycle number monotonicity, record completeness). If the check passes, it enters S7 to carry out the feedback output required for statistics, judgment, and triggering (fine-tuning / re-decision). If the check fails, it reports an exception and outputs the minimum set of degraded feedback to ensure that the closed loop is not interrupted.

[0054] S7. Statistically analyze the data forwarding and gating transmission results to generate the feedback report required for closed-loop scheduling, and perform parameter adaptation or trigger re-decision to complete the scheduling of dynamic data packets. Specifically: Collect key performance indicators (KPIs) and system resource status at the granularity of queue / business flow / port / window period; The collected key performance indicators (KPIs) are bound to the executable policy (Policy_Ver), decision version number (Input_Ver), and window period number (Cycle_ID) of this round to form a feedback report package; Based on the feedback report packet, if the critical time-sensitive network TSN flow exceeds the window / expiration period or the window utilization rate is abnormally high and conflicts occur frequently, it is determined to be a policy mismatch or insufficient resources, and a reordering needs to be triggered; if the critical business stably hits the window and the congestion is controllable, it is determined to be a fine-tuning optimization, and the weight / quota / integrated rate increment and hysteresis threshold adjustment strategy are output to complete the scheduling of dynamic data packets.

[0055] In this embodiment, in S7, the scheduling execution and feedback module (EXFB) statistically analyzes and attributes the gating transmission and forwarding results of S6, forming the feedback report required for closed-loop scheduling, and driving the scheduling controller SC and the policy adjustment module (ADJ) to perform parameter adaptation or trigger re-decision. The scheduling execution and feedback module (EXFB) first collects key performance indicators (KPIs) at the granularity of "queue / service flow / port / window period", including: actual forwarding latency (queueing latency from enqueue to transmission, forwarding latency from transmission to the output port), latency jitter (variance, percentile difference), time-sensitive network (TSN) window compliance rate (whether it falls within W_start~W_end), deadline hit rate, number of window overruns and overdues, number of drops and reorders, number of conflicts (statistics of delays / preemption caused by multiple key flows competing for the same window), window utilization (ratio of actual bytes sent within the window to available capacity), queue congestion trend (queue length growth rate, number of times queuing latency approaches the threshold), and system resource status (CPU usage, port busy rate, cache usage). To ensure traceability, the scheduling execution and feedback module (EXFB) binds the aforementioned KPIs with the current effective policy version (Policy_Ver), the decision input version (Input_Ver), and the window period number to form a "feedback report package," which is then written to the feedback channel for the scheduling controller (SC) to read. The EXFB then executes its judgment logic: if critical time-sensitive network TSN flows exceed the window / expiration time or have abnormally high window utilization and frequent conflicts, it is judged as "policy mismatch or insufficient resources," requiring a reordering; if critical services stably hit the window and congestion is controllable, with only slight jitter or low resource utilization, it is judged as "fine-tunable," and suggested weight / quota / integrated rate increments and hysteresis threshold adjustments are output. When necessary, the EXFB can also attribute anomalies, such as jitter caused by a sudden increase in wireless risk, congestion caused by a sudden surge in ordinary flows, or window congestion caused by port busyness, so that the SC / ADJ can make targeted adjustments rather than blindly weighting.

[0056] S7 → Loop transition process: (1) Parameter fine-tuning back to S1: If the scheduling execution and feedback module (EXFB) determines that only fine-tuning is needed, it sets the "fine-tuning flag" and writes the suggested parameters into the parameter update area of ​​the policy adjustment module (ADJ); the scheduling controller SC atomically publishes the parameter sub-version at the next scheduling boundary (without changing the main structure of the window planning), and then the process returns to S1 to resample and enter the next round of closed loop. (2) Rescheduling and re-decision required back to S4: If the scheduling execution and feedback module (EXFB) determines that rescheduling is needed, it sets the "forced re-decision flag" and sends the reason code, key performance indicator (KPI) summary and conflict warning information to the scheduling controller SC; the scheduling controller SC immediately freezes the feedback report and triggers the dynamic scheduling decision module (DPM) to enter S4 to recalculate the weight, quota, window sending plan and conflict avoidance strategy, generate a new execution policy Policy_Ver, and enter S5 / S6 for execution after double buffer switching. Both paths require the scheduling controller (SC) to record the version link (Input_Ver→Policy_Ver→Feedback Report ID) to ensure closed-loop traceability in auditing, review, and patent description.

[0057] In summary, the 5G TSN dynamic data packet scheduling method and system architecture based on DS-TT in this invention are as follows: a closed-loop system of "monitoring and perception layer - scheduling control and policy layer - execution and forwarding layer" is constructed within the DS-TT node. 5G / TSN data and status are uniformly accessed through the interface management module, and the scheduling controller coordinates dynamic scheduling decisions, policy adjustments, queue scheduling and gating forwarding to achieve deterministic forwarding and dynamic adaptive scheduling on the end side.

[0058] In this invention, the TSN time-constrained windowed deterministic scheduling mechanism generates window pressure and conflict warnings based on the period T, window W_start~W_end and the deadline model, and performs window consistency verification during the scheduling and forwarding phases to ensure that critical TSN services are forwarded within the specified time window.

[0059] In this invention, a cross-network collaborative combined strategy generation and stable update mechanism is implemented: the Dynamic Scheduling Decision Module (DPM) outputs a combined strategy of "weight + quota + in-window transmission plan + conflict avoidance", and the Strategy Adjustment Module (ADJ) uses threshold and hysteresis for adaptive updates to reduce strategy jitter caused by wireless fluctuations and sudden traffic.

[0060] In this invention, the queue scheduling and gating forwarding have a traceable execution feedback mechanism: the queue scheduling module (QSM) performs weighted polling / preemption guarantee according to the policy, the data forwarding module (FWD) sends data according to the plan and supports cross-window processing; the scheduling execution and feedback module (EXFB) binds KPIs with Input_Ver / Policy_Ver / Cycle_ID versions and feeds them back to the controller to achieve verifiable and reviewable closed-loop optimization.

[0061] This invention addresses the core needs of deterministic wireless networks in fields such as industrial automation and intelligent manufacturing. Its deployment location (DS-TT side) is compatible with existing 5G-TSN converged architectures and can be implemented through software and programmable hardware. Its feasibility for implementation is primarily reflected in the following aspects: The software is deployable: the module can run on the DS-TT CPU / embedded Linux, obtain wireless status through the interface with the 5G protocol stack, and achieve scheduling and execution through queue management.

[0062] Hardware enhancements are possible: Strictly real-time processes such as "dequeueing / gating / timestamping" can be accelerated by FPGA / hardware queues to further reduce jitter.

[0063] Mature engineering interfaces: It can connect to CNC configuration distribution and local management interfaces, adapting to the engineering operation and maintenance needs of industrial sites.

[0064] Easy integration into the supply chain: DS-TT is often provided by equipment manufacturers / gateway manufacturers / system integrators. This invention can be integrated into 5G TSN gateways, edge computing boxes, and industrial routers as a "DS-TT software feature / enhancement package" to form a marketable differentiated function.

[0065] The target market is clear and the demand is strong: the demand for "determinism + wireless" continues to grow in scenarios such as industrial internet, intelligent manufacturing, rail transit, power automation, ports and mines, and 5G+TSN is the mainstream direction of integration.

Claims

1. A dynamic data packet scheduling method, characterized in that, Includes the following steps: S1. After the data packets from the 5G network side and the time-sensitive network (TSN) side are uniformly accessed, a traffic status snapshot is formed by classifying the service flow, counting the queue length and queuing delay. S2. Collect dynamic information of the wireless link on the 5G network side, transform the dynamic information into a wireless risk factor that directly affects scheduling, bind the wireless risk factor with the traffic status snapshot to form the wireless part of the joint input frame, and perform time alignment processing; wherein, the wireless risk factor carries both the sampling timestamp and the validity period. S3. After completing the binding and time alignment, obtain the periodicity and window boundary of the Time Sensitive Network (TSN) service, and output conflict warning information by calculating the remaining deadline and constraint level. S4 integrates traffic status snapshots, wireless risk factors, and time-sensitive network (TSN) time constraints to generate a complete policy of weights, quotas, dequeue order, in-window transmission plan, and conflict avoidance. Based on the complete policy, parameter adaptation and threshold hysteresis control are performed. Among them, the TSN time constraints include period, window, deadline, and conflict warning. S5. Based on the threshold hysteresis control results and according to the currently effective executable policies, optimize and protect the critical time-sensitive network TSN services within the window. S6. Based on the optimized protection processing results, perform data forwarding and gated transmission; S7. Statistically analyze the data forwarding and gating transmission results, generate the feedback report required for closed-loop scheduling, and perform parameter adaptation or trigger re-decision to complete the scheduling of dynamic data packets.

2. The dynamic data packet scheduling method according to claim 1, characterized in that, S1 includes the following steps: Acquire data packets from the 5G network side and the Time-Sensitive Network (TSN) side; Data packets are classified and identified according to preset service identification rules, and the data packets are mapped to queue groups. Statistics are updated at the granularity of queue / service flow, and the statistical results are saved to the state sharing and buffer. The statistics include queue length, average and percentile queuing delay estimates, arrival rate, burstiness, drop count, reorder count, and critical flow percentage. The identification fields include 5-tuple, VLAN PCP / DSCP, Stream ID of Time Sensitive Network (TSN), service tag, and flow mapping table issued by CNC / local policy. Based on statistics, when the scheduling cycle boundary or the trigger event threshold is reached, the current statistical value is solidified into a traffic status snapshot. The traffic status snapshot includes the congestion level of each queue, queuing delay estimate, number and proportion of critical flows, burstiness index and snapshot timestamp T_snap. After the traffic status snapshot is generated, write the snapshot sequence number Snapshot_ID and the snapshot timestamp T_snap, and set the traffic snapshot valid flag. After reading the valid flag of the traffic snapshot using the scheduling controller SC, the snapshot sequence number Snapshot_ID is locked as the decision input for this round. At the same time, the read and write versions of the corresponding snapshot are frozen, thus completing the formation of the traffic status snapshot. If S1 generates a new traffic status snapshot before S2 is completed, it will only enter the pending queue and will not replace the currently locked traffic status snapshot, ensuring the consistency of the decision cycle.

3. The dynamic data packet scheduling method according to claim 2, characterized in that, S2 includes the following steps: The system collects dynamic information about wireless links on the 5G network side. This dynamic information includes link quality level, scheduling rhythm information, wireless side buffer / queue backlog, and end-to-end wireless transmission latency statistics for the past N cycles. Sliding window filtering and threshold hysteresis determination are applied to dynamic information indicators; Based on the judgment result, the dynamic information is transformed into a wireless risk factor R that directly affects the scheduling. The wireless risk factor R carries both the sampling timestamp T_wireless and the validity period. Bind the wireless risk factor R with the snapshot sequence number Snapshot_ID to form the wireless part of the joint input frame: Snapshot_ID, T_snap, R, T_wireless; The scheduling controller (SC) collects the wireless status of the same period and performs time alignment verification based on the snapshot sequence number Snapshot_ID. If |T_wireless If T_snap| exceeds the threshold, a supplementary sampling is triggered: if the supplementary sampling fails, the wireless risk factor R from the previous period is used and the R_use status is marked; if the wireless side interface is abnormal, the scheduling controller SC enters the degraded state and generates an available risk factor R; after completing the binding and time alignment, it enters S3.

4. The dynamic data packet scheduling method according to claim 1, characterized in that, S3 includes the following steps: Read the key parameter set for each Time-Sensitive Network (TSN) stream. The key parameter set includes the period T, the gating control list (GCL) or equivalent window boundaries W_start to W_end, the maximum available time slot / bandwidth per period, and the service level and priority constraint rules. Among them, W_start represents the start time of the window in which the TSN stream is allowed to start sending data within a scheduling period, and W_end represents the end time of the window in which the TSN stream must complete sending within a scheduling period. Construct a mapping relationship between the Stream ID of the Time-Sensitive Network (TSN) and the local queue / service flow identifier, so that the TSN packets identified in S1 are associated with the corresponding time constraint entries; Based on the mapping results, the effective window boundary within the current scheduling period is calculated using the local unified time base, and the remaining deadline (deadline_remain) is calculated for each time-sensitive network (TSN) data packet to be scheduled. At the same time, the constraint level is output for the queue level. Calculate the window pressure index: Given the remaining capacity of the current window, can the total bytes of the critical packets to be sent be completed within the window? If insufficient or multiple critical flow windows overlap and compete, a conflict warning message is output. After outputting (T, W_start / W_end, deadline_remain, constraint level, conflict warning), the scheduling controller (SC) encapsulates it along with the traffic status snapshot and the radio risk factor R into a decision input frame, and binds them uniformly to the same snapshot sequence number Snapshot_ID and decision version number Input_Ver. Based on the encapsulation result, a consistency check is performed: if a Time-Sensitive Network (TSN) configuration is missing, the window cannot be calculated, the time base is abnormal, or the mapping is missing, a constraint anomaly flag is set and a security policy is adopted. After completing the consistency check, the scheduling controller (SC) freezes the decision version number Input_Ver and generates a policy that can be traced back to the specific input version.

5. The dynamic data packet scheduling method according to claim 4, characterized in that, S4 includes the following steps: The traffic status, radio risk factor R, resource status, and time constraints of the time-sensitive network (TSN) are used to aggregate the decision version number Input_Ver of the scheduling controller (SC). Based on the convergence results, the overall scheduling priority P is calculated at the granularity of queue / business flow; Based on the comprehensive scheduling priority P, output a complete strategy including weight, quota, dequeue order, in-window sending plan, and conflict avoidance. For the complete policy, apply threshold and hysteresis rules to generate an executable policy Policy_Ver; Write the executable policy Policy_Ver to the policy double buffer. First, write it to the spare policy area along with (Input_Ver, Policy_Ver, effective boundary time). Then, publish the new policy availability flag through an atomic switch. Based on the write results, the system switches to the new executable policy Policy_Ver only when the effective boundary is detected, thus completing threshold hysteresis control.

6. The dynamic data packet scheduling method according to claim 5, characterized in that, Specifically, S5 is: Based on the current executable policy_Ver, scheduling is differentiated between windowed and non-windowed states. Specifically, when within a Time-Sensitive Network (TSN) window, priority is given to critical TSN queues that are allowed to send within the window, and these queues are sorted from smallest to largest based on their remaining deadline_remain. When outside the window or when the minimum reservation share for a critical TSN queue has been met, services are allocated using a round-robin approach with dynamic weights. For bursty flows, token bucket / leaky bucket constraints are implemented in conjunction with the integer parameters in the current executable policy_Ver. When the latency of a critical TSN queue approaches the threshold or a conflict warning has not been resolved, the service ratio of the critical TSN queue is increased or the strongest preemption is triggered. At the end of each scheduling slot / cycle, the selected data packet or transmission descriptor, along with the current executable policy Policy_Ver, queue service statistics, and exception flags, is entered into S6.

7. The dynamic data packet scheduling method according to claim 6, characterized in that, S6 includes the following steps: Get the send descriptor and the current executable policy, Policy_Ver; At the start of each transmission cycle, based on the window state and non-window state determined by the equivalent window boundaries W_start to W_end and the current time base, a window consistency check is performed on each item in the queue to be transmitted: if it belongs to a critical time-sensitive network TSN flow that must be transmitted within the window, it is only allowed to be transmitted within the window; if it is predicted that the transmission will cross the end of the window, it will be transmitted early, temporarily stored in the next window, or trigger degradation processing according to the current executable policy Policy_Ver; for ordinary service flows, transmission is carried out according to the quota and shaping results of the current executable policy Policy_Ver; when the port is busy or the buffer usage increases, rate limiting, batch transmission, or temporary storage is performed according to the current executable policy Policy_Ver, and forwarding is continuously updated. If necessary, the remaining transmission descriptors are iteratively processed within the same window cycle until the window ends or the current batch is processed. Record execution evidence during the process of deterministic forwarding to the Time-Sensitive Network (TSN) and forwarding / backhauling to the 5G side; The execution evidence is summarized to form a set of key performance indicators (KPIs) for this cycle. The set of key performance indicators (KPIs) for this cycle is then bound to the decision version number Input_Ver, the current executable policy Policy_Ver, and the window cycle number Cycle_ID to generate a feedback report package that can be used for closed-loop optimization. Based on the generated feedback report package, when the packaging boundary is reached, the execution records within this boundary are encapsulated into a record package and written to the buffer in an atomic manner, and the record package ready flag is set; wherein, the record package includes the package identifier record Record_ID, the window cycle number Cycle_ID, the decision version number Input_Ver, the current executable policy Policy_Ver, and the key state summary; Once the record packet is detected as ready, a consistency check is performed. If the check passes, proceed to S7.

8. The dynamic data packet scheduling method according to claim 7, characterized in that, S7 includes the following steps: Collect key performance indicators (KPIs) and system resource status at the granularity of queue / business flow / port / window period; The collected key performance indicators (KPIs) are bound to the executable policy (Policy_Ver), decision version number (Input_Ver), and window period number (Cycle_ID) of this round to form a feedback report package; Based on the feedback report packet, if the critical time-sensitive network TSN flow exceeds the window / expiration period or the window utilization rate is abnormally high and conflicts occur frequently, it is determined to be a policy mismatch or insufficient resources, and a reordering needs to be triggered; if the critical business stably hits the window and the congestion is controllable, it is determined to be a fine-tuning optimization, and the weight / quota / integrated rate increment and hysteresis threshold adjustment strategy are output to complete the scheduling of dynamic data packets.

9. A dynamic packet scheduling system for executing the dynamic packet scheduling method according to any one of claims 1-8, characterized in that, include: The real-time traffic awareness module is used to identify and count the traffic flow of data packets from the 5G network side and the time-sensitive network (TSN) side, and form a traffic status snapshot. The wireless state awareness module is used to collect dynamic information of wireless links on the 5G network side, transform the dynamic information into wireless risk factors that directly affect scheduling, bind the wireless risk factors with traffic state snapshots to form the wireless part of the joint input frame, and perform time alignment processing. The system status monitoring module is used to continuously monitor the resource and congestion status of DS-TT itself; The TSN time constraint module is used to obtain the periodicity and window boundary of Time-Sensitive Network (TSN) services, and output conflict warning information by calculating the remaining deadline and constraint level. The scheduling controller (SC) is used to integrate traffic state snapshots, radio risk factors, and time constraints of time-sensitive networks (TSN) to generate a complete policy of weight, quota, dequeue order, in-window transmission plan, and conflict avoidance. The strategy adjustment module is used to perform parameter adaptation and threshold hysteresis control on the complete strategy; The queue scheduling module is used to optimize and protect critical time-sensitive network (TSN) services within a window based on the threshold hysteresis control results and the currently effective executable policies. The data forwarding module is used to perform data forwarding and gated transmission based on the optimized protection processing results; The scheduling execution and feedback module is used to statistically analyze the data forwarding and gating transmission results, generate the feedback report required for closed-loop scheduling, and perform parameter adaptation or trigger re-decision to complete the scheduling of dynamic data packets.

Images