Time-constrained scheduling computation in time-sensitive networks

Abstract

This disclosure relates to a method performed by an entity of a Time-Sensitive Network (TSN) for scheduling transmission times of intermediate bridges between terminal stations communicating with each other via the TSN, at least one of the intermediate bridges being a radio frequency (RF) bridge involving a radio frequency cellular network (5GS). The TSN entity (CNC) obtains a capability prediction (Qp) of the Quality of Service (QoS) for at least the radio frequency bridge from the 5GS, and schedules the transmission times based on the capability prediction (Qp) and further based on a prediction range (Hp) corresponding to the effective duration of the QoS capability prediction (Qp).

Description

Technical Field

[0001] This disclosure relates to a method and system for integrating wireless bridges in a Time-Sensitive Network (TSN) that takes into account variations in the quality of service of the radio link (e.g., due to terminal mobility). Priority is claimed to European Patent Application No. EP 24305257.8, filed on February 14, 2024, the contents of which are incorporated herein by reference. Background Technology

[0002] In a TSN network, to achieve guaranteed end-to-end latency for packet transmission between a speaker and a listener, it is necessary to rely on a predetermined data path or route and the scheduling of operations of network elements along that path. A TSN includes, for example, […]. Figure 1 Several components are shown, for example: ● Terminal stations, such as the terminals used by speakers and listeners. ● Top System for Centralized User Configuration (CUC): Terminal stations send their flow requests to the CUC system, and the CUC system sends the user configuration to the Centralized Network Configuration Node (CNC). ● The CNC node receives TSN flow requirements and TSN topology from the CUC system. It obtains egress and ingress port identifiers, service categories, and QoS metrics as minimum and maximum latency for each port pair from different TSN bridges. The CNC node calculates the scheduling (transmission time) for intermediate bridges between TSN terminals, as well as the time-aware shaping (TAS) control parameters and routing decisions (bridge selection) for different bridges. These calculations are performed to meet the data flow requirements of the TSN flows.

[0003] ● The entity called a "bridge" is a packet switching device of the TSN. Packets arriving at one port (ingress) of the bridge can be routed to another port (egress) with a bounded delay. The routing delay is advertised to the CNC node. Packets are buffered at an egress port until the port is opened after the time-gated window table configured by the CNC node.

[0004] A TSN bridge relays packets from one port to another based on parameters specific to each port pair, as indicated by the internal link. These parameters are typically: - The relevant latency factor (which can typically depend on the payload and can be considered as the inverse of the link's instantaneous throughput), and - Independent delay (e.g., typically a fixed delay for each communication).

[0005] In a centralized implementation of TSN, these parameters are provided to a CNC node, which collects information on the network link topology and capacity, as well as data flow requirements. After defining the data path used to transmit packets from one end station to another via bridges in the TSN, the CNC node calculates a gate-opening schedule that allows packets to flow from one ES (End Station) to another while guaranteeing their transmission delays. To do this, the CNC node calculates the cumulative delay from egress port to egress port of two consecutive bridges along the data path, where the egress port is defined at the output port of the bridge in the direction of packet flow along the path (while the ingress port is the bridge input port).

[0006] Therefore, the typical configuration cycle between the CNC node and the bridge in the TSN network can be described as follows, while now referring to... Figure 2 .

[0007] 1) In the first step, the CNC node collects QoS capabilities from the different TSN bridges that make up the TSN, typically independent latency and associated latency.

[0008] 2) The CNC node calculates the schedule by considering data flow characteristics and the QoS capabilities of the TSN network bridging, thus avoiding packet collisions in the network.

[0009] 3) CNC nodes typically provide each TSN bridge with the computed schedules involved in the process in the form of a time-gated window table.

[0010] From this point on, data streams can be delivered with time guarantees.

[0011] Executing the configuration loop can take a considerable amount of time, typically from a few seconds to several hours, depending on the topology complexity and the computing power of the CUC system and CNC nodes.

[0012] In existing technologies, this configuration cycle is executed when data flows with added, removed, or modified QoS requirements are encountered, and when the TSN topology is modified—that is, when a TSN bridge is added or removed. These events typically correspond to modifications to production lines within a factory, which are relatively rare occurrences.

[0013] In the existing technology, it is assumed that the QoS capabilities provided to the CNC node by the TSN bridge are always guaranteed.

[0014] Therefore, in the prior art, it is not a problem to define the operating window that is valid for a time period of the schedule (time gating configuration) calculated by the CNC, such as configuring the loop time.

[0015] However, when considering the integration of wireless network devices into a TSN (where typically at least one TSN bridge participates as a wireless network device, such as a cellular radio network like a 3GPP 5G system), the assumption that the QoS capabilities provided by the TSN bridge are always guaranteed does not apply well to wireless network devices where radio conditions may evolve spatially and temporally, leading to changes in throughput and transmission latency.

[0016] For another example, as a mobile component in a factory, the radio conditions can evolve as the robot moves around the factory. Existing time-gating windows can be determined solely based on worst-case scenarios in both space and time regarding the QoS capabilities (e.g., transmission latency) advertised by network devices (i.e., claimed performance related to the cell edge location), such as... Figure 3 As shown.

[0017] This leads to underutilization of wireless network resources.

[0018] In practice, when the current device lacks worst-case QoS capabilities, considering accurate local QoS capabilities would be beneficial for both the current device and other devices. For example, fewer resources need to be reserved for the current device, which can be better shared across all devices.

[0019] However, this is not possible with existing operating windows that do not assume that the QoS capabilities of network devices may change.

[0020] Therefore, considering mobile devices in TSN presents several challenges: - Wireless channels fluctuate due to environmental mobility or mobile terminals. Transmission latency in 5G systems varies over time. The scheduling calculations on the CNC node side take some time. - The TSN configuration closure (acquire delay / calculate schedule / transmit schedule) also has a minimum delay that is not necessarily compatible with variations in wireless transmission delay. Summary of the Invention

[0021] This disclosure aims to improve this situation.

[0022] It proposes a method executed by an entity of a time-sensitive network for scheduling the transmission time of intermediate bridges between terminal stations communicating with each other via a time-sensitive network, at least one of which is a radio frequency bridge involving a radio frequency cellular network.

[0023] More specifically, the aforementioned entity of the time-sensitive network obtains a capability prediction Qp of at least the Quality of Service (QoS) of the wireless bridge from the radio frequency cellular network, and schedules transmission time based on the capability prediction Qp and further based on a prediction range Hp corresponding to the effective duration of the QoS capability prediction Qp.

[0024] "Scheduled transmission time" refers to the calculation of the scheduling of these transmission times up to the time it takes to send a scheduling command to the relevant bridge.

[0025] In examples of implementation methods, the aforementioned entity can typically be a centralized network configuration node of a time-sensitive network.

[0026] The aforementioned wireless network bridge is typically a 5G cellular network.

[0027] Then, the method enables the ability to predict Qp for at least a given range of Hp from the 5G network, allowing entities in the time-sensitive network to determine efficient bridge scheduling for the next data transmission session, for example, by computing the data flow scheduling of the bridges.

[0028] In one implementation, the aforementioned entity sends the following to the radio frequency cellular network: - A first request to obtain at least the predicted QoS capability Qp, and - A second request for configuring the wireless bridge according to the scheduled transmission time. Furthermore, within an operating cycle that ends with a period of effective data transmission between terminal stations (e.g., corresponding to the aforementioned data transmission session), the transmission of a request for QoS capability prediction Qp precedes the transmission of a request for configuring the wireless bridge, the period ending before the duration of the prediction range Hp given for the operating cycle.

[0029] In the first embodiment, the entity is: - Obtain both the QoS capability prediction Qp and the prediction range Hp from the radio frequency cellular network. During the prediction range Hp, the radio frequency cellular network guarantees the QoS capability prediction Qp, and The transmission time is scheduled based on the predicted Qp and its range Hp provided by the radio frequency cellular network.

[0030] In the second embodiment, the entity is: - Apply a prediction range Hp to the radio frequency cellular network corresponding to the effective duration for which the QoS capability prediction Qp of the wireless bridge must be guaranteed by the radio frequency cellular network. - Receive from the radio frequency cellular network a QoS capability prediction Qp corresponding to the applied prediction range Hp, and The transmission time is scheduled based on the applied prediction range Hp and the received QoS capability prediction Qp.

[0031] In the third embodiment, which is between the first and second embodiments, the entity is: - Apply the minimum value of the prediction range Hp to the radio frequency cellular network, corresponding to the effective duration for which the QoS capability prediction Qp of the wireless bridge is guaranteed by the radio frequency cellular network. - Receive from the radio frequency cellular network a negotiated value of a predicted range Hp corresponding to the effective duration for which the predicted QoS capability Qp of the wireless bridge can be guaranteed by the radio frequency cellular network, the negotiated value being equal to or higher than the minimum value, and The transmission time is scheduled based on the negotiated value of the predicted range Hp and the predicted Qp based on the guaranteed QoS capability.

[0032] In one type of implementation of a time-sensitive network where entities operate according to consecutive operation cycles as described above, within the same operation cycle, the entity can: - During the data acquisition period (as shown in the attached figure) Figure 5 and Figure 6 The middle is marked as "T" D ")period: Send a request to the entity of the radio frequency cellular network to obtain data on the QoS capability prediction Qp of at least the wireless bridge, and Obtain a prediction range Hp corresponding to the validity duration of the obtained QoS capability prediction Qp (depending on the choice of the second or first implementation, Hp is determined by the entity or obtained directly from the wireless network). - And, during the scheduling period (in Figure 5 and Figure 6 The mark "T" in S During this period, the scheduling of the transmission time can be calculated based on the QoS capability prediction Qp and its prediction range Hp. In one operation cycle, the scheduling period (T) S Following that is: - Configure time period (T) C This corresponds to the time taken for the Time-Sensitive Networking (TSN) entity (CNC) to send configuration data to the intermediate bridge to prepare for effective data transmission between terminal stations, and for the bridge to receive the sent configuration data and thus be configured to follow the scheduled transmission time. - The transmission period (T) of effective data transmission between terminal stations based on the scheduled transmission time. O Transmission period (T)O The process ends before the validity of the QoS capability prediction Qp expires.

[0033] In this implementation, entities in a time-sensitive network can typically be scheduled during the time period (T). S During this period, a time-gated window table is calculated, which defines the time when data packets buffered at the intermediate bridge will be transmitted by the intermediate bridge.

[0034] Furthermore, in this embodiment, the transmission period (T) of the current operation cycle O This can typically depend on the prediction range Hp given for the current operating cycle, and each operating cycle therefore has a total duration that depends on the prediction range Hp given for each operating cycle.

[0035] In addition, at least one of the following: - The data acquisition period (T) D ), corresponding to the duration of the data collection phase, and - The configuration period (T) C This corresponds to the duration for which an entity in a time-sensitive network sends a schedule for the data stream to each intermediate bridge. It can be defined as an untunable timing parameter based on the worst-case estimate of the response time.

[0036] Then, the entities of the time-sensitive network can satisfy the condition that the data acquisition period (T) is considered within a given operation cycle. D The configuration period (T) C ) and the transmission period (T) O Meanwhile, the remaining scheduling period (T) within the given operation cycle S The constraints of the transmission time during the scheduling period.

[0037] In this embodiment, the operation cycle can therefore have a total duration as a fixed and non-tunable time parameter.

[0038] Furthermore, entities in a time-sensitive network can use optimization algorithms (such as so-called genetic algorithms) to compute the scheduling of transmission times in order to satisfy the scheduled time period (T). S Given the remaining time constraints within the scheduled period, the optimization algorithm stops at the end of the scheduled period to provide the latest optimal solution obtained as a result for scheduling.

[0039] Alternatively, computing resources can be dynamically adapted to meet computing time constraints (e.g., typically by using a variable number of computing processors).

[0040] In this implementation using the optimization algorithm, in the current operating cycle, the optimization algorithm starts scheduling computation from the previous scheduling computation obtained for the previous operating cycle (instead of starting from scratch, thus making the algorithm converge faster).

[0041] In this implementation, the entities in the time-sensitive network can at least be in the effective data transmission period (T) of the previous operating cycle preceding the current operating cycle. O Before the end, send a request for a QoS capability prediction Qp for the current operation period.

[0042] This implementation is referred to below as the "pipeline mode" and Figure 10 The example is presented.

[0043] In one implementation, the entity in the time-sensitive network can be active before the end of a previous operating cycle, and during the data acquisition period (T) of the previous operating cycle. D After that, a request for a QoS capability prediction Qp for the current operation period is sent.

[0044] This implementation method is referred to below as "parallel mode" and Figure 11 The example is presented.

[0045] This specification also relates to a computer program that includes instructions that, when executed by a processing unit, cause the above-described method to be implemented.

[0046] This specification also relates to the aforementioned entities of time-sensitive networks, which include processing units for performing the methods described above.

[0047] This entity can be configured to be incorporated into a centralized network configuration node of a time-sensitive network. Attached Figure Description

[0048] Other features, details, and advantages will be shown in the following detailed description and accompanying drawings, wherein: - Figure 1 The general architecture of Time-Sensitive Networks (TSN) is illustrated schematically. - Figure 2 This illustrates the messages typically exchanged between CNC nodes and 5G system components, particularly those used to expose QoS capabilities to the TSN-CNC. - Figure 3 The diagram illustrates a typical "worst-case" delay statement based on existing technical methods. - Figure 4 A method according to this specification with variable delay declarations is shown. - Figure 5 The timeline of the sequential steps to be considered is shown. - Figure 6 It shows Figure 5 A timeline, but without gaps. - Figure 7 The diagram illustrates messages exchanged between a CNC node and a 5G system element according to a first embodiment, wherein the 5G system element provides a prediction range Hp, the prediction range Hp being determined by a duration T. P and its optional start time t p composition, - Figure 8 The diagram illustrates messages exchanged between a CNC node and a 5G system element according to a second embodiment, wherein the predicted range duration Hp is provided to the 5G system element, which takes it into account to backhaul a prediction of QoS capabilities that should remain effective during the provided predicted range. - Figure 9 The diagram illustrates messages exchanged between a CNC node and 5G system components according to an embodiment, wherein, for example, a prediction range Hp is provided to the CNC node by the 5G system components, defining the remaining time for the CNC node to calculate the schedule, and thereby applying the time T spent on that calculation. s , - Figure 10 This illustrates a "pipeline" operation mode for processing several consecutive operation windows. - Figure 11 This illustrates a "parallel" operation mode for processing several consecutive operation windows. - Figure 12 Examples of operation modes used for consecutive windows in the first or second embodiment according to the pipeline mode are shown. - Figure 13 This shows the effect in pipeline mode. Figure 9 Examples of operation modes used for continuous windows in the implementation method, - Figure 14 A hybrid implementation between the first and second embodiments is shown, which establishes a trade-off for the prediction range Hp to be applied to scheduling through negotiation between the CNC and 5GS.

[0049] - Figure 15 The schematic illustration shows an entity CNC of a Time-Sensitive Network (TSN) implementing the above-described method in a system including a 5GS cellular network, according to an example of an embodiment. Detailed Implementation

[0050] In the following text, the term "CNC" or "CNC node" refers to the entity responsible for scheduling data flow on bridges in a Time-Sensitive Network (TSN). It can be, for example (but not limited to), a centralized network configuration node of the TSN.

[0051] One type of bridge in TSN involves a wireless bridge that operates on a radio frequency cellular network (such as a 5G network). The term "5GS" refers to this radio frequency cellular network and typically includes some radio components and core network components. 5GS includes an entity responsible for exchanging messages with the CNC node to perform calculations for data flow scheduling of the bridge, also known as "bridge scheduling." In the following text, the term "5G" or "5G network" refers to such a radio frequency cellular network (with 5G type or any other network generation).

[0052] In the method described below, with Figure 3 Compared to the existing technology, this method assesses the current accurate Quality of Service (QoS) conditions of the 5G network for the current operating window. Therefore, a time parameter (which is then assumed to be valid for a given duration) can be introduced into the operating window determination for bridge scheduling, such as... Figure 4 As shown.

[0053] In practice, since the Quality of Service (QoS) in a radio frequency cellular network can vary depending on the conditions of the wireless network (including the location of the terminal), QoS can be predicted for a limited effective time period. This prediction is referred to below as "QoS Capability Prediction" and labeled "Qp", and its effective time period is referred to as "Prediction Range" and labeled "Hp".

[0054] Several parameters in a 5G network influence the fact that scheduling, utilizing bridge QoS capability calculations based on QoS capability predictions from 5GS, will remain effective during the operational window for TSN terminal stations that may be embedded in mobile 5G terminals (and thus ensure overall guaranteed control of end-to-end latency). Depending on the scenario, some of these parameters are non-tunable (and can be predetermined, measurable, fixed, or variable but calculated only once), while others can be tuned. These parameters at least characterize: The time required to collect QoS capability predictions and their prediction range from 5G components. The range may not be provided when seeking QoS capability predictions, and it can be collected later, for example, for future calculations. In this case, available QoS capability predictions (obtained before the current cycle) can be used. Finally, QoS capabilities can be collected at a lower rate than scheduled computations.

[0055] Obtain the relevant TSN bridge QoS capabilities and calculate the time required for TSN scheduling at the CNC node.

[0056] Configure the TSN bridge timing accordingly.

[0057] The time and duration of the operation window configured for applying the TSN bridge (preferably based on the predicted range of the collected QoS capabilities).

[0058] One implementation method is thus proposed for: The values ​​of the aforementioned untunable parameters are collected from the different network elements (including 5GS) in the TSN. Calculate the tunable parameters used to ensure the quality of service in the operation window. During the operation window, the corresponding configuration is applied to TSN data transmission.

[0059] Therefore, the duration of the tuning sequence steps is proposed to ensure good operation of TSN in 5G involving terminal mobility.

[0060] exist Figure 7 In the first embodiment shown, the wireless network provides feedback on the QoS capability prediction and prediction range for each device (such as the QoS prediction predicted by 5GS, and the time during which the QoS prediction is guaranteed), and the duration of the operating window is adapted to ensure that the calculated schedule will operate within a time window that is compatible with the lowest prediction range found in 5G.

[0061] exist Figure 8 In the second embodiment shown, the 5GS can predict QoS capabilities for any given prediction range. The CNC needs the prediction range Hp for operation during the target operation window, and the prediction range Hp is notified to the 5GS for use in calculating its QoS capability prediction.

[0062] Within this range (Hp), the CNC must calculate bridge scheduling and send configuration commands to the bridges to ensure data packet transmission according to the schedule until the range duration ends. Therefore, predicting the end of the range duration can define the next time to collect QoS capability predictions and schedule the next future bridge scheduling.

[0063] In the second implementation, typically, if the applied range Hp is too long, the QoS capability prediction may yield poor QoS because it will be the only QoS that can be guaranteed during the long range, and the prediction will be suboptimal. On the other hand, a short range Hp will be able to guarantee good QoS, but may not allow the CNC enough time to effectively compute bridge scheduling.

[0064] In a hybrid implementation between the first and second embodiments (e.g.) Figure 14As shown, the CNC applies a minimum value, Hpmin, of the predicted range Hp to the 5GS to allow sufficient time for bridge scheduling, etc. The 5GS can then base its QoS capability predictions on this minimum range. The 5GS can also estimate QoS capabilities for a longer range Hp, for example, when the radio network situation is sufficiently stable due to slowly moving terminals, or when it has a small amount of data flow to process, resulting in limited radio resource usage. The 5GS then transmits this longer predicted range Hp along with the associated QoS capability predictions to the CNC, which allows the CNC to have more than the minimum range Hpmin for scheduling.

[0065] Finally, due to the "negotiation" between the CNC and 5GS, this hybrid implementation establishes a trade-off between the first and second implementations for the range of predictions to be applied to scheduling.

[0066] According to this specification and any of these embodiments, the CNC node adjusts the scheduled computation time, for example, by defining a maximum computation time. The maximum computation time is calculated by taking into account the duration of the target operating window and, of course, the QoS capability prediction range of the 5G network.

[0067] A more detailed implementation method is given below. To set up the operation window, the following steps are typically required: Step D: Obtain information related to the QoS capability prediction Qp of at least one data link (the radio link between the terminal and its serving base station), which can be calculated in the 5GS at the terminal, in the eNodeB or gNB, or in any other 5G core network element (collectively referred to here as "5GS"): ○ T D and t D These are the duration and start time of the information data collection phase, respectively. Figure 5 As shown, ○ T P and t P These are the duration and start time of the range window for information related to the QoS capability prediction Qp of at least one data link, respectively. In practice, the value T... P and t P For example, it relates to the radio link conditions of the data link, which may vary due to terminal mobility or changes in cell load.

[0068] Step S: At the CNC node, calculate the scheduling based on the obtained QoS capabilities and predict Qp. ○ T S and t S These are the duration and start time of the computation phase of the scheduling process in the CNC node, respectively.

[0069] Step C: The CNC node configures the TSN element based on the calculated schedule. ○ T C and t C These refer to the duration and start time of the TSN device configuration phase, respectively.

[0070] Step O: Send data packets via TSN elements configured according to the calculated schedule: ○ T O and t O These are the duration and start time of the operation window, during which the TSN device uses the calculation and configuration of steps S and C for scheduling.

[0071] Therefore, in order to achieve the correct ordering of operations and ensure the availability of information for computation, the following properties must be satisfied: Figure 5 As shown: Condition t D +T D ≤t S Ensure that information related to the QoS capability prediction (Qp) for at least one data link has been received before scheduling calculations begin. If no information is received for some affected data links, the last received information for those data links is used instead.

[0072] Condition t s +T S ≤t c Ensure that scheduling calculations have finished before the TSN network configuration is executed.

[0073] Condition t c +T c ≤t O Ensure the operation window starts after the TSN element configuration is complete.

[0074] Condition t P ≤t O Ensure that the scope window for calculating scheduling information begins before the operation window begins.

[0075] Condition t O +T O ≤t P +T P Ensure that the operation window closes before the range window used for calculating scheduling information closes.

[0076] Therefore, the parameters listed above appear to be interdependent.

[0077] The condition check can be performed by one of the network elements involved (e.g., a dedicated function in the CNC or a dedicated node in the TSN) after collecting the necessary information, and provides feedback on whether the configuration ensures the effectiveness of the scheduling calculated from QoS capabilities during the operating window.

[0078] In the method detailed below, the calculation of tunable parameters is performed based on non-tunable parameters, ensuring the effectiveness of scheduling calculated from QoS capability predictions during the operating window. According to the implementation, this parameter determination (DoP) can be performed by different network elements at different times.

[0079] Some conditions are ensured through the definition of a protocol, which involves: The schedule is calculated only after a message is received from the element performing the DoP (this ensures that t is always satisfied). D +T D ≤t S ) The configuration only begins with the CNC after the CNC has completed the scheduling calculations (this ensures that t is always satisfied). s +T S ≤t c ) The operation window only opens after the configuration is complete, via CNC commands (this ensures that t is always met). c +T c ≤t O ) In addition, the intervals between operations should generally be included in T. D T S T C The waiting time in the value. Therefore, it is assumed below that there is no gap between steps, which involves the t. D +T D =t S t S +T S =t c t C +T C =t O .

[0080] Finally, the system consists of t D T D T S T C T O and T P and tP Parameterization is required, and the following two constraints must be met: Condition t P ≤t O Ensure that the range of QoS capability predictions used for computational scheduling begins before the operation window begins.

[0081] Condition t O +T O ≤t P +T P Ensure that the operation window ends before the range of QoS capability prediction information used for calculating scheduling ends.

[0082] Figure 6 The diagram shows the timeline of these sequential steps without time gaps between them.

[0083] Of all the wireless devices that can intervene in the next transmission session, the 5GS (the core network entity, or eNodeB, or base station, etc.) is responsible for collecting QoS predictions Qp (and their possible range Hp).

[0084] Therefore, as a note, when collecting several T for DoP on several data links... P and t P When values ​​are used, combinations of these values ​​can be used to determine a single representative T. P and t P For example, when considering the worst-case scenario, max({t P}) used as t P The representative value, and min({t P +T P})-max({t P}) can be used as T P The representative value.

[0085] Regarding the obtained parameters, T D This refers to the duration of QoS capability prediction collection. During this phase, network elements calculating DoP typically send messages to one or more 5G network elements and receive QoS capability predictions in response. D These are typically non-tunable parameters, and can be estimated by considering the worst-case response time of different bridges and the number of bridges in the network.

[0086] T C This refers to the duration of the configuration phase. During this phase, the CNC node sends a schedule for the data flow to each bridge in the network. C These are typically non-tunable parameters, and can be estimated by considering the worst-case response time of different bridges and the number of bridges in the network.

[0087] T S This is the duration used to calculate the scheduling. In one implementation, this duration is fixed and determined, for example, by pre-measuring the maximum computation time of the CNC node. In subsequent alternative implementations, the parameter T... S These can be tunable parameters that can be calculated as described below.

[0088] T P This is a parameter representing the QoS capability prediction duration and prediction range (Hp). Depending on the implementation, T P These parameters can be tunable or non-tunable.

[0089] In the first embodiment discussed above, the value T C T D T S It is non-tunable and pre-obtained. Parameter t O The duration T to be determined O An instance of a repeating set of operation windows. Parameter t D By t D =t O -(T C +T D +T S ) Determine. For simplicity, assume t P =t D (Assuming QoS request prediction starts from the request time). Prediction range T P Collected from 5GS by the CNC node, therefore, in this first embodiment, it is applied to the CNC node.

[0090] Duration T O By T O =T P -(T C +T D +T S )Calculate. Based on the obtained T P Calculate T O .

[0091] exist Figure 7 An example of the first implementation is presented, where Hp is the prediction range for QoS capability prediction. Typically, Hp is a tuple ([t P ], T P ), where t P It can be optional. When t is not provided. P This means that the prediction starts from the current time (e.g., the time the collection request message was received). Otherwise (t P and t D (Not equal), except for the predicted range duration TP In addition, the 5GS also provides its start time t p .

[0092] The operation window can also be used to adapt the configuration cycle duration to the prediction range provided by 5GS. For example, it can be used to determine the next time to start a new collection request by considering the next operation window that begins at the end of the previous operation window. t O (i+1)=t O (i)+T O (i) Therefore, if, for example, a mobile device moves rapidly within a given part of a factory, radio conditions can evolve rapidly. Consequently, the duration of accurate QoS capability predictions can be shorter, and configuration cycles can be adapted accordingly to ensure optimized scheduling at all times. When radio conditions are more stable, the range of QoS capability predictions can be longer. Thus, configuration cycles can evolve accordingly, thereby saving signaling and computational resources.

[0093] A possible constraint is that the operation window is determined after the reception of the collected response. Therefore, it can occur either after or before the scheduling calculation.

[0094] In the second embodiment, the value T D T S T C It is non-tunable and pre-obtained. Parameter t O It is a fixed (pre-determined) duration T O An instance of a repeating set of operation windows. Parameter t D By t D =t O -(T C +T D +T S ) Determine. For simplicity, assume t P =t D To minimize the prediction range T P Calculated as T P =T O -(T C +T D +T S ). Parameter T P Provided to 5GS, and taken into consideration by 5GS for backhaul, it can guarantee effective QoS capability prediction during the range applied to 5GS.

[0095] exist Figure 8 An example of the second implementation is described, where Hp is the target prediction range for QoS capability prediction. Typically, Hp is a tuple ([t... P ], T P ), tP It is optional. When t is not provided... P This means that the prediction should be made from the current time (e.g., the time when the collection request message is received) (t P and t D (Equal) begins. If provided, then t P It can be included in t D (Predicting QoS requests starts from the request time) and t O (The predicted QoS request starts from the beginning of the target operation window) between.

[0096] In a possible third implementation, the value T C T D It is untunable and pre-obtained. To determine T... P and T S For simplicity, assume the parameter t P =t D (QoS request prediction starts from the request time). Prediction range T P Collected from 5GS by the CNC node. Then based on the obtained T P Calculate T S The CNC node must be at time T. S Internal computation scheduling.

[0097] exist Figure 9 An example of the third implementation is shown, where Hp is the prediction range for QoS capability prediction. Typically, Hp is a tuple ([t... P ], T P ), t P It is optional. When t is not provided... P This means that the prediction starts from the current time (e.g., the time when the collection request message is received).

[0098] In this third embodiment, typically, the next time to collect Qp (and possibly Hp) from 5GS can be imposed by a fixed duration of the operation window To, independent of the current prediction range Hp. Therefore, the CNC will satisfy constraints to compute the schedule within a limited remaining computation cycle Ts within an operation window.

[0099] In practice, time constraints arising from scheduling computation will be considered in this case. Scheduling computation itself has a time constraint T. S Furthermore, the optimization algorithm used for scheduling computation will provide results within a given time limit.

[0100] This can be achieved in the following ways: - By using optimization algorithms such as genetic algorithms, which can stop computation at any time. In this case, the latest optimal solution obtained is the result used for scheduling. - Or by dynamically adapting computing resources to meet computing time constraints.

[0101] Furthermore, optimization algorithms can begin their search from previous solutions instead of starting from scratch. Therefore, the algorithms converge faster.

[0102] In the worst-case scenario where scheduling calculations cannot be performed due to insufficient time left for the CNC, an infinite prediction range Hp, which provides suboptimal Qp, can be used in abnormal situations, as is done in the prior art.

[0103] Alternative algorithms can be learning algorithms that typically use artificial intelligence, in order to provide: - The first preliminary step involves the algorithm learning how to compute a typical scheduling configuration offline, for example, based on a given prediction Qp (or a prediction vector Qpi for each wireless device involved in a communication cycle) and the relevant range Hp. - and the second current step, performed online and especially under time constraints, wherein the algorithm determines the schedule based on previous observations of having the same or similar pairs (Qp, Hp) (or vectors (Qpi, Hp)).

[0104] To enable scheduling calculations and the use of their results, data flow characteristics and requirements, such as source, destination, maximum transmission delay, and priority, are provided to the CNC. The scheduling calculations will take into account the QoS capabilities indicated by the network devices and provide scheduling results that meet the data flow requirements.

[0105] Since the QoS capabilities indicated by 5GS may evolve at each operating window, the scheduling outcome may also evolve. This means that the end-to-end transmission time of the data stream can vary between different operating windows.

[0106] What may happen is that, given a QoS capability prediction, some data flows cannot be accepted within the operating window. If the scheduler can compute several schedules with different priorities, it moves the data flows between the different schedules according to the data flow priority, ensuring that the higher-priority schedule at least satisfies the data flow requirements. Alternatively, lower-priority data flows can serve as "best-effort" data flows for the duration of the operating window.

[0107] Typically, the third implementation is characterized by time-constrained scheduling computation. Therefore, the scheduling result will depend on QoS capability predictions, but also on the time constraints imposed on the algorithm. If the time constraint is large, the scheduling result is optimized, corresponding to the optimal solution the algorithm can find. When the time constraint becomes a true constraint, the result provided within the time constraint may be a less optimized solution, potentially making it impossible to accept all data streams. In this case, one solution is to dynamically adapt computing resources. Another possibility is to assign priorities to data streams as described above.

[0108] Of course, some of the above embodiments can be combined. In fact, the above embodiments are not exclusive. For example: - CNC nodes can have a target minimum operating window (in implementations 2 and 3). - The CNC node can provide minimum range predictions to 5GS (in implementation 2 and hybrid implementations). - 5GS can make QoS predictions for a predicted range Hp (typically higher than the minimum range threshold) based on cell load, device location, and mobility backhaul (in implementations 1, 2, and 3). - If necessary, the CNC node can determine the time constraints for scheduling calculations based on Hp (in implementation 3). - The operation window can be updated (in implementation 1).

[0109] Therefore, the typical exchange data message for collecting QoS capabilities is as follows: - From CNC node to 5GS: Collect_Req(([Hp]) - From 5GS to CNC node: Collect_Rep(Qp, Hp) Where Hp=([t P ],T P ) Instead of regular collection, "asynchronous" collection can be performed. In fact, in the previous implementation, it was assumed that the CNC sent a message to the 5GS each time it wanted to start a collection cycle, and the 5GS provided its response when it received the request.

[0110] Collection can also be asynchronous. In another implementation: the CNC node sends a first message indicating a request for QoS capability predictions and prediction ranges. The 5GS can then send its response without further requests from the CNC node. For example, when the 5GS estimates that changes in radio conditions are sufficient and it has better predictions to provide, the 5GS can send new values. In this case, the CNC node triggers the determination of a new operating window based on the new QoS capability predictions and prediction ranges received from the 5GS.

[0111] More generally, the above description presents a sequential sequence of steps for a single operation window. In practice, it is reasonable to assume that operation windows occur periodically and continuously. In this case, additional constraints can be satisfied to ensure that there is no overlap in the pipeline of sequential steps, as described below.

[0112] It is assumed that the next operation window will begin when the previous operation window ends; therefore: t i+1O ≤t i O +T i O From t i+1 P The next range window can start at t i D and t i+1 O It begins between.

[0113] To achieve continuous and non-overlapping efficient operating modes, the following settings can be determined: t i+1 O =t i O +T i O .

[0114] For continuous operations, t(i+1) can be set before the current window ends. Setting it at the end can be seen as an optimization of the window duration. However, for example, if the CNC node receives a new (Qp, Hp) asynchronous event, it can terminate window i before window i ends.

[0115] According to the previous implementation, the constraints should remain in effect.

[0116] From t i+1 D The initial definition includes two operation modes: - If t i+1 D ≥t i O In the so-called "pipeline" operation mode, steps D, S, and C of the next window i+1 are completed during step O of the previous window i, such as... Figure 10 As shown, and - If t i+1 D ≤t i O And t i+1 D ≥t i D In the so-called "parallel" operation mode, one or more of the next window i+1 steps D, S, and C can be completed during one or more previous window i steps, such as... Figure 11 As shown. Here, the CNC node calculates the bridge schedule for the current operating cycle and initiates the collection of the predicted Qp (in the first implementation or hybrid implementation, and ultimately its corresponding range Hp) for the next operating cycle in parallel.

[0117] Depending on the operating mode, some constraints of the previous implementation methods need to be effective.

[0118] Typically, for the first and second embodiments, in Figure 12 The process involves continuous window operation. In continuous operation mode, the collection time is calculated as t. i+1 D =t i+1 O -(T C +T D +T S ), where T S =T i S =T i+1 S And t i+1 O =t i O +T i O .

[0119] Then, depending on when the next collection occurs, two scenarios can be considered: - When t i+1 D ≥t i O At that time, the assembly line operation mode, in which: T i O ≤T i P -(t i P -t i O ) (where T) i P T must be included i O ),as well as T i O ≥T D +T S +T C

[0120] (where T) i O Therefore, the following window steps D, S, and C must be included.

[0121] - When t i+1 D ≥t iD +T D And t i+1 D ≤t i O In parallel operation mode, where: ■ T i P It is known at the end of the collection request. ■ If t i+1 D =t i D +T D Then T i O ≤ T i P – (t i P - t i D ),and ■ If t i+1 D =t i D +T D Then T i O ≥T D +T S +T C -(T S +T C ), so T i O ≥T D And when t i+1 D ≥t i D At that time, t i+1 D ≤t i D +T D And t i+1 D ≤t i O ,in: ■ If t i+1 D =t i D Then T i O ≤T i P -(T D +T S +T C ) Now regarding the third implementation method, Figure 13 The example presents continuous window operations (implemented in this example according to the aforementioned pipeline pattern), where the scheduling time T... i S and T i+1 S Depends on T i P and T i+1 P .

[0122] In continuous operation mode, t i+1 O =t i O +T i O .

[0123] RT must be determined i O .

[0124] Depending on when the next collection occurs, two scenarios can be considered: - When t i+1 D ≥t i O At that time, the assembly line operation mode is as follows: T i O ≤T i P -(t i P -t i O ) (where T) i P T must be included i O ) T i O ≥T D +T i+1 S +T C Then T i+1 S ≤T i O -(T D +T C ) (where T) i O The next window steps D, S, and C must be included. T i P +(ti P -t i D )≥T i S +T i+1 S +2T D +2T C

[0125] T i S and RT i+1 S A balance must be struck between the current window step and the next window step. T i+1 S It should be set to the minimum value T Smin ,as well as - Parallel operation mode, where: If t i+1 D ≥t i D +T D And t i+1 D ≤t i O : ■ T i P It is known at the end of the collection request. ■ If t i+1 D =t i D +T D Then T i O ≤ T i P – (t i P - t i D ) ■ T i O +T C +T i S ≥T D +T i+1 S +T C ■ If T i S =T i+1 S =T S Then T iO ≥T D If t i+1 D ≥t i D , then t i+1 D ≤t i D +T D And t i+1 D ≤t i O : If t i+1 D =t i D Then T i O ≤T i P -(T D +T i S +T C And T i O ≤T i P -(T D +T i+1 S +T C ) Figure 15 The illustration schematically depicts a Time-Sensitive Network (TSN) where a transmitter Em sends data packets to a receiver Re via bridges BR1, BR2, ... of the TSN. One of these bridges is a wireless bridge, and it can be, for example, a radio frequency cellular network 5GS, where one or more wireless devices WD1, WD2, ... participate in data transmission between the transmitter and receiver. This is similar to what is used in cellular networks. Figure 15 As shown in the example, wireless devices are connected to one or more base stations (BS), which may be connected to the core network (CNW). The 5GS network (e.g., an entity of the core network) can collect QoS predictions (Qpi) of the wireless devices involved in data transmission between transmitters and receivers (and possibly their ranges (Hpi), according to the aforementioned first embodiment), and can determine, for example, the minimum prediction range (Hp) among the devices involved (e.g., based on the device with the shortest prediction range).

[0126] When the wireless network is one of the bridges, the entity of the network TSN (such as the centralized network configuration node CNC) can then use the predicted range Hp of the wireless network 5GS (e.g., provided by the core network CNW) to calculate the scheduling of the transmission time of the TSN bridge according to this specification.

[0127] Therefore, the physical CNC includes a processing unit, which includes at least: - A communication interface COM for receiving predicted Qp (and possibly its range Hp) from at least the 5GS entity and sending scheduling data to the TSN bridge. - A memory MEM that stores at least the instructions of a computer program, which, when read by the processor PROC, cause the implementation of the methods described above, and - This processor PROC is connected to memory MEM to execute computer program instructions to calculate bridge scheduling and corresponding transmission times based on the predicted Qp and its range Hp, and is connected to interface COM to send transmission times to the TSN bridge.

Claims

1. A method performed by an entity of a Time-Sensitive Network (TSN), the method being used to schedule the transmission time of intermediate bridges between terminal stations communicating with each other via the TSN, at least one of the intermediate bridges being a radio frequency bridge involving a radio frequency cellular network (5GS). in, The entity (CNC) of the Time-Sensitive Network (TSN) obtains a capability prediction Qp for the Quality of Service (QoS) of at least the wireless bridge from the radio frequency cellular network (5GS), and schedules the transmission time based on the capability prediction Qp and further based on a prediction range Hp corresponding to the effective duration of the QoS capability prediction Qp.

2. The method according to claim 1, wherein, The entity sends the following to the radio frequency cellular network (5GS): - A first request to obtain at least the predicted QoS capability Qp, and - A second request ("Configure (Schedule)") for configuring the wireless bridge according to the scheduled transmission time. Furthermore, during the effective data transmission period (T) between the terminal stations... O Within the operation period ending at (t), the transmission of requests for QoS capability prediction Qp precedes (t) D The transmission of requests for configuring the wireless bridge (t) C The time period (T) O The process ends before the duration (Tp) of the predicted range Hp given for the operation cycle.

3. The method according to any one of the preceding claims, wherein, The entity (CNC): - Obtain both the QoS capability prediction Qp and the prediction range Hp from the radio frequency cellular network (5GS), during the prediction range Hp, the radio frequency cellular network (5GS) guarantees the QoS capability prediction Qp, and The transmission time is scheduled based on the predicted Qp and its range Hp provided by the radio frequency cellular network (5GS).

4. The method according to any one of claims 1 and 2, wherein, The entity (CNC): - Apply a prediction range Hp to the radio frequency cellular network (5GS) corresponding to the effective duration for which the QoS capability prediction Qp of the wireless bridge needs to be guaranteed by the radio frequency cellular network (5GS). - Receive the QoS capability prediction Qp corresponding to the applied prediction range Hp from the radio frequency cellular network (5GS), and - The transmission time is scheduled based on the applied prediction range Hp and the received QoS capability prediction Qp.

5. The method according to any one of claims 1 and 2, wherein, The entity (CNC): - Apply the minimum value of the prediction range Hp to the radio frequency cellular network (5GS) corresponding to the effective duration for which the QoS capability prediction Qp of the wireless bridge is guaranteed by the radio frequency cellular network (5GS). - A negotiated value received from the radio frequency cellular network (5GS) corresponding to the effective duration for which the QoS capability prediction Qp of the wireless bridge can be guaranteed by the radio frequency cellular network (5GS), the negotiated value being equal to or higher than the minimum value, and - The transmission time is scheduled based on the negotiated value of the predicted range Hp and the predicted Qp of the guaranteed QoS capability.

6. The method according to any one of the preceding claims, wherein, The entity (CNC) of the Time-Sensitive Network (TSN) operates according to consecutive operating cycles, and within the same operating cycle: - During the data acquisition period (T) D )period: Send a request to the entity of the radio frequency cellular network (5GS) to obtain data on the QoS capability prediction Qp of at least the wireless bridge, and Obtain the prediction range Hp corresponding to the effective duration of the obtained QoS capability prediction Qp. - During the scheduling period (T) S During this period, the scheduling of the transmission time is calculated based on the QoS capability prediction Qp and its prediction range Hp. During an operating cycle, in the scheduling period (T) S Following that is: - Configure time period (T) C The configuration time period (T) C The entity (CNC) corresponding to the Time-Sensitive Network (TSN) sends configuration data to the intermediate bridge to prepare for effective data transmission between the terminal stations, and causes the bridge to receive the sent configuration data and thus be configured to follow the time spent on the scheduled transmission time. - The transmission period (T) of effective data transmission between the terminal stations according to the scheduled transmission time. O The transmission period (T) O The process ends before the validity of the QoS capability prediction Qp expires.

7. The method according to claim 6, wherein, The transmission period (T) of the current operating cycle O The duration of each operation cycle depends on the prediction range Hp given for the current operation cycle.

8. The method according to any one of claims 6 and 7, wherein, The entity (CNC) of the time-sensitive network satisfies constraints to ensure that, during the data acquisition period (T) considering a given operating cycle... D The configuration period (T) C ) and the transmission period (T) O Meanwhile, the remaining scheduling period (T) within the given operation cycle S The scheduling of the transmission time is calculated during the period.

9. The method according to claim 8, wherein, The Time-Sensitive Network (TSN) entity (CNC) uses an optimization algorithm to calculate the scheduling of the transmission time in order to satisfy the scheduling period (T). S Given the remaining time constraints within the scheduled period, the optimization algorithm stops at the end of the scheduled period to provide the latest optimal solution obtained as a result for scheduling.

10. The method according to claim 9, wherein, In the current operating cycle, the optimization algorithm begins scheduling computation from the previous scheduling computation obtained for the previous operating cycle.

11. The method according to any one of claims 6 to 10, wherein, The entity (CNC) of the Time-Sensitive Network (TSN) has at least the effective data transmission period (T) of the previous operating cycle preceding the current operating cycle. O Before the end, send the request for QoS capability prediction Qp for the current operation period.

12. The method according to any one of claims 6 to 11, wherein, The entity (CNC) of the Time-Sensitive Network (TSN) occurs before the end of a previous operating period preceding the current operating period and during the data acquisition period (T) of the previous operating period. D After that, the request for QoS capability prediction Qp for the current operating period is sent.

13. A computer program comprising instructions that, when executed by a processing unit, cause the implementation of the method according to any one of the preceding claims.

14. An entity for a Time-Sensitive Network (TSN), the entity comprising a processing unit for performing the method according to any one of claims 1 to 12.

15. The entity of claim 14, wherein the entity is configured to be incorporated into a centralized network configuration (CNC) node of the Time-Sensitive Network (TSN).

Images