Time-sensitive network gating scheduling system based on automatic driving

Abstract

The invention discloses a time-sensitive network gating scheduling system based on automatic driving. The system comprises a multi-source sensing module, a sensing collaborative fusion module and a hybrid gating scheduling module. The multi-source sensing module integrates a millimeter wave radar, a laser radar, a camera and an inertial navigation unit to obtain real-time environment data, and the real-time environment data is transmitted through a time-sensitive network of an IEEE 802.1 Qbv standard to reduce delay and jitter. The perception collaborative fusion module preprocesses and fuses multi-source data, and constructs an accurate three-dimensional environment model by using Kalman filtering, point cloud processing and a deep learning algorithm. The hybrid gating scheduling module integrates an IEEE 802.1 Qbv deterministic gating mechanism and an earliest deadline (EDF) real-time scheduling mechanism, periodic data are managed through a gating control list, and the EDF dynamically schedules event-driven data to ensure that key tasks are transmitted preferentially.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving, and more specifically, to a time-sensitive network gating scheduling system based on autonomous driving. Background Art

[0002] In recent years, with the rapid development of autonomous driving technology, the data traffic and complexity of in-vehicle networks have been continuously increasing, posing higher requirements for the real-time performance, latency, and bandwidth utilization of data transmission. Existing research has proposed some solutions for the scheduling of TT (Time-Triggered) flows in time-sensitive networks (TSN), but there are still certain limitations.

[0003] Existing public literature 1 (Research on Time-Sensitive Network Gating Scheduling Mechanism [D], 2022) proposed a scheduling based on the response time of TT flows. This solution optimizes the transmission order of TT flows through the ITIO algorithm and the slot sorting scheduling algorithm to minimize the sum of the response times of all TT flows in the network. The specific steps are as follows: First, design the ITIO algorithm to allocate the earliest available slot to each flow in a given order, and postpone the start time of the flow in case of conflicts; Second, propose the slot sorting algorithm, and according to the slot length and path hop count of the flow, give priority to scheduling short-slot flows to ensure that the flow is transmitted as early as possible in the link and reaches the destination node. However, this solution is only applicable to scenarios with known flow parameters and fixed network conditions, and there is a phenomenon of data transmission delay.

[0004] Existing public literature 2 (Research on Online Scheduling Algorithm for Real-Time Flows in Time-Sensitive Networks, 2021) proposed an online scheduling algorithm for TT flows. This algorithm dynamically determines the scheduling priority by setting the active interval and virtual deadline for TT flows on each switch: Only when the flow is in the active interval, it participates in the competition according to the virtual deadline, otherwise it is scheduled with the lowest priority. The algorithm determines the active interval length by iteratively calculating the worst-case response time (WCRT), and proves that if the sum of the active intervals of each switch ≤ the flow deadline, the scheduling can be guaranteed to succeed. However, as the number of flows increases, the success rate of scheduling decreases, and the utilization rate of network bandwidth is low.

[0005] Existing solutions are difficult to simultaneously meet the real-time requirements of periodic data and time-driven data when dealing with dynamic network environments, and there are problems such as transmission delay, large jitter, and insufficient bandwidth utilization efficiency. Therefore, there is an urgent need to develop a new scheduling system that can effectively reduce data transmission delay and jitter, ensure the real-time performance of data, and at the same time improve the utilization efficiency of network bandwidth to meet the higher requirements of autonomous driving technology for in-vehicle networks. Summary of the Invention

[0006] To overcome the above-mentioned defects of the prior art, the present invention provides a time-sensitive network gating scheduling system based on autonomous driving, which classifies and dynamically manages the scheduling of periodic data and event-driven data by integrating the deterministic gating scheduling mechanism of the IEEE 802.1Qbv standard and the earliest deadline first (EDF) real-time scheduling mechanism, so as to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A time-sensitive network gating scheduling system based on autonomous driving, including a multi-source perception module, a perception collaborative fusion module, and a hybrid gating scheduling module; the multi-source perception module obtains real-time environmental perception data from the vehicle-mounted sensor network by integrating millimeter-wave radar, lidar, camera, and inertial navigation unit; the perception collaborative fusion module integrates the real-time sensor data of the multi-source perception module; the hybrid gating scheduling module integrates the deterministic gating scheduling mechanism of the IEEE 802.1Qbv standard and the earliest deadline first real-time scheduling mechanism, and the specific steps are as follows:

[0009] Step S1, the gating scheduling analyzes and classifies the transmitted data frames to distinguish periodic data frames and event-driven data frames;

[0010] Step S2, for periodic data frames, the gating scheduling constructs a gating control list based on the IEEE 802.1Qbv standard;

[0011] Step S3, during the operation of the gating scheduling, when the sensor generates a burst of non-periodic data frames with deadlines, the data is identified as an event-driven data frame;

[0012] Step S4, when the event-driven data frame enters the scheduling queue, the gating scheduling immediately starts the earliest deadline first scheduling algorithm, dynamically sorts according to the deadline information carried by each data frame, and calculates and determines the real-time transmission priority of the data frame.

[0013] Step S5, when the network link is idle and the gating is open, the gating scheduling sends periodic data frames according to the control list rules. If the deadline of the event-driven data frame is lower than the urgency threshold within 10 milliseconds, the earliest deadline scheduling mechanism will give priority to sending the event-driven data frame.

[0014] As a further solution of the present invention, the multi-source perception module obtains real-time environmental perception data from the vehicle-mounted sensor network by integrating millimeter-wave radar, lidar, camera, and inertial navigation unit, including the following specific content: the multi-source perception module obtains real-time environmental perception data from the vehicle-mounted sensor network, and the sources of the environmental perception data include millimeter-wave radar, lidar, high-resolution camera, and inertial navigation unit.

[0015] To ensure the efficient and real-time transmission of various sensor data, the multi-source perception module uses a time-sensitive network based on the IEEE 802.1Qbv standard for data transmission. The IEEE 802.1Qbv standard realizes the precise division of transmission time slots through a precisely defined gating mechanism to ensure that the data frames of each sensor can be efficiently transmitted within a pre-planned time window, reducing latency and jitter during data transmission.

[0016] As a further solution of the present invention, the perception collaborative fusion module integrates the real-time sensor data of the multi-source perception module, including the following specific contents: The perception collaborative fusion module preprocesses and fuses the real-time sensing data obtained by the multi-source perception module. The implementation process of the preprocessing is as follows:

[0017] Step Y1, the perception collaborative fusion module synchronizes all sensor devices in the network based on the IEEE 802.1ASrev time synchronization protocol through a global clock to eliminate the time deviation caused by the independent operation of the sensors, so that the acquisition timestamps of various data are consistent.

[0018] Step Y2, for millimeter-wave radar data, a Kalman filter is used for real-time noise suppression and clutter filtering; for lidar point cloud data, voxelization and statistical filtering methods are used to eliminate point cloud noise and anomalies; for camera image data, Gaussian filtering, histogram equalization, and edge enhancement image processing algorithms are used to improve the clarity of the image, and the spatial consistency alignment of the image data with other sensor data is achieved through internal and external parameter calibration methods; for inertial navigation unit data, zero-bias correction, temperature compensation, and inertial navigation solution algorithms are implemented to eliminate sensor errors and cumulative drift, and attitude and motion trajectory information is output in real time.

[0019] Step Y3, the perception collaborative fusion module performs spatio-temporal difference and data association on the preprocessing results of various sensors in the same spatio-temporal framework to achieve the consistency and alignment of various data in both spatio-temporal dimensions.

[0020] After completing the preprocessing and spatio-temporal alignment of the data from the millimeter-wave radar, lidar camera, and inertial navigation unit, the perception collaborative fusion module further fuses the data according to the characteristics of different sensor data and the real-time perception requirements of the intelligent driving system. Specifically, it includes: taking the distance and speed measurements provided by the millimeter-wave radar and lidar as inputs, using the unscented Kalman filter to fuse and generate a more accurate and stable target motion state estimation, reducing the random noise and errors in the measurement process of a single sensor; for the camera images and lidar point cloud data, applying joint calibration and feature matching techniques, through the spatial transformation and field-of-view calibration of the point cloud and image, realizing the fusion of three-dimensional spatial data and two-dimensional visual data, and constructing a three-dimensional environmental model with rich spatial information and visual semantic features; combining the convolutional neural network YOLOv5 and the point cloud network PointNet++, extracting the category, position, size, and orientation information of the targets in the image and the three-dimensional geometric information of the point cloud data respectively, and then using a multi-layer perceptron for deep fusion processing of multi-feature vectors to improve the accuracy and robustness of environmental perception. The motion information such as the vehicle attitude, speed, and angular velocity measured by the inertial navigation unit is used as dynamic constraint conditions to assist in real-time correction of the data errors caused by vehicle operation during the fusion process.

[0021] As a further solution of the present invention, the hybrid gating scheduling module integrates the deterministic gating scheduling mechanism of the IEEE 802.1Qbv standard and the earliest deadline real-time scheduling mechanism, including the following specific contents: The core basis of the hybrid gating scheduling module is the time slot division gating mechanism defined by the IEEE 802.1Qbv standard. The time slot division gating mechanism controls the data sending queue in terms of time on the ports of the Ethernet switch and the end system, and schedules the transmission of network data frames in a deterministic manner. Each sending queue in the network is equipped with a corresponding gate, which has two states: open and closed. A data frame is only allowed to be sent when the gate of the corresponding queue is in the open state, otherwise it will be temporarily stored in the sending queue waiting for the next gate opening moment. The specific moments of gate opening and closing are determined by the gate control list and executed periodically. The gate control list consists of a series of time slots, and each time slot details the state and duration of the gate of each queue. However, the network communication in autonomous vehicles not only has periodic data, which includes millimeter-wave radar and lidar data streams, but also has a large amount of event-driven data, which includes emergency obstacle avoidance instructions and suddenly appearing obstacle alarms. The event-driven data has the characteristics of suddenness and strict real-time requirements. Therefore, the hybrid gating scheduling module introduces a real-time online scheduling mechanism based on the earliest deadline (EDF) to more flexibly meet the dynamic real-time transmission requirements of event-driven data in autonomous vehicles. The real-time online scheduling mechanism based on the earliest deadline dynamically adjusts the priority according to the deadline of each data frame to be sent, so as to ensure the timeliness of real-time data transmission. The specific steps are as follows:

[0022] Step Z1, a data frame is generated from a sensor or a task source and enters the data sending queue. At this time, each data frame carries a clear deadline.

[0023] Step Z2, the earliest deadline scheduler monitors and records the deadline of each data frame in the queue in real time and maintains the deadline information table of the data frame.

[0024] Step Z3, the earliest deadline algorithm continuously sorts the data frames in the queue in real time, arranging them in the order from the nearest to the farthest deadline, that is, the data frame with the nearest deadline is at the front of the queue and has the highest priority; the data frame with a farther deadline is at the back of the queue and has a lower priority.

[0025] Step Z4, when a new data frame arrives at the queue, the earliest deadline algorithm immediately compares the deadline of the new data frame with the deadlines of the data frames already in the queue; if the deadline of the newly arrived data frame is more urgent than that of the highest-priority data frame in the current queue, the new data frame will obtain a higher transmission priority and be immediately placed at the front of the queue, while updating the sorting order of the entire queue.

[0026] Step Z5, when the network link or transmission channel is idle, the earliest deadline scheduler immediately selects the data frame at the forefront of the current sorting in the queue for transmission.

[0027] Step Z6, when a certain data frame is successfully transmitted, the earliest deadline scheduler removes it from the queue, updates the queue status, and then executes Step Z3 again to dynamically sort the remaining data frames in the queue according to the deadlines.

[0028] By combining the deterministic gating scheduling of the IEEE 802.1Qbv standard and the earliest deadline first (EDF) real-time scheduling mechanism, the utilization efficiency of network resources is improved, and the risks of data transmission jitter and delay caused by improper network resource allocation are reduced.

[0029] Technical effects and advantages of a time-sensitive network gating scheduling system based on autonomous driving according to the present invention: The present invention integrates the deterministic gating scheduling mechanism of the IEEE 802.1Qbv standard and the earliest deadline first (EDF) real-time scheduling mechanism. In the communication network of autonomous driving vehicles, it realizes the classified scheduling and dynamic management of periodic data and event-driven data, and solves the problem in the prior art that it is difficult to simultaneously consider the real-time performance, low latency, low jitter, and high bandwidth utilization of data transmission. The present invention precisely divides the time slots of periodic data frames through a gating control list to ensure their transmission within a strictly controlled time window, meeting their requirements for deterministic and low-jitter transmission. At the same time, the system introduces an online scheduling strategy based on EDF to dynamically respond to bursty event-driven data frames with high real-time requirements. By calculating the deadlines of data frames in real time and adjusting the priorities, the priority scheduling of critical task data is realized, effectively reducing the risk of data failure due to delay. Brief Description of the Drawings

[0030] Figure 1 It is a schematic structural diagram of a time-sensitive network gating scheduling system based on autonomous driving according to one embodiment of the present invention.

[0031] Figure 2 It is a flowchart of the ITIO algorithm in a scheduling based on the response time of TT flows in the prior art.

[0032] Figure 3 It is a flowchart of an online scheduling algorithm for TT flows in the prior art.

[0033] Figure 4 Schematic diagram of the response times of three scheduling strategies, one of the embodiments of the present invention, in 10 burst emergency events.

[0034] Figure 5 Schematic diagram of the end-to-end delay comparison of the data transmission in the hybrid gating scheduling module, one of the embodiments of the present invention.

[0035] In the figure: Qbv Only represents the IEEE 802.1Qbv deterministic scheduling; EDF Only represents the earliest deadline first scheduling; Hybrid Qbv+EDF represents the combination of the IEEE 802.1Qbv standard deterministic gating scheduling and the earliest deadline real-time scheduling mechanisms of the present invention. Detailed implementation manners

[0036] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0037] Embodiment 1

[0038] Referring to Figure 1 the schematic diagram shown, an embodiment of the present invention provides a time-sensitive network gating scheduling system based on autonomous driving, which includes a multi-source perception module, a perception collaborative fusion module, and a hybrid gating scheduling module; the multi-source perception module obtains real-time environmental perception data from the vehicle-mounted sensor network by integrating millimeter-wave radar, lidar, camera, and inertial navigation unit; the perception collaborative fusion module integrates the real-time sensor data of the multi-source perception module; the hybrid gating scheduling module fuses the IEEE 802.1Qbv standard deterministic gating scheduling mechanism and the earliest deadline real-time scheduling mechanism.

[0039] In this embodiment, referring to Figure 2 , a scheduling based on the response time of TT flows disclosed in the prior art is provided. This solution optimizes the transmission order of TT flows through the ITIO algorithm and the time slot sorting scheduling algorithm to minimize the sum of the response times of all TT flows in the network. The specific steps are as follows: First, design the ITIO algorithm to allocate the earliest available time slot to each flow in the given order, and postpone the start time of the flow when there is a conflict; secondly, propose the time slot sorting algorithm, and preferentially schedule the short time slot flows according to the time slot length and path hop count of the flow to ensure that the flow is transmitted as early as possible in the link and reaches the destination node. However, this solution is only applicable to scenarios with known flow parameters and fixed network conditions, and there is a phenomenon of data transmission delay. Referring toFigure 3 , which is a TT flow online scheduling algorithm disclosed in the prior art. This algorithm dynamically determines the scheduling priority by setting the active interval and virtual deadline for TT flows on each switch: only when the flow is in the active interval, it participates in the competition according to the virtual deadline, otherwise it is scheduled with the lowest priority. The algorithm determines the length of the active interval by iteratively calculating the worst-case response time (WCRT), and proves that if the sum of the active intervals of each switch ≤ the flow deadline, the scheduling can be guaranteed to succeed. However, as the number of flows increases, the ratio of successful scheduling of this algorithm decreases, and the utilization rate of network bandwidth is low.

[0040] Furthermore, the multi-source perception module obtains real-time environmental perception data from the vehicle-mounted sensor network by integrating millimeter-wave radar, lidar, camera, and inertial navigation unit, including: the multi-source perception module obtains real-time environmental perception data from the vehicle-mounted sensor network, and the sources of the environmental perception data include millimeter-wave radar, lidar, high-resolution camera, and inertial navigation unit. The millimeter-wave radar uses its all-weather detection ability and high-dynamic ranging characteristics to continuously detect and measure key information such as the distance, speed, and direction of target objects under different weather and lighting conditions, thus providing accurate target tracking and obstacle avoidance basis for autonomous driving. The lidar generates high-precision three-dimensional spatial point cloud data by emitting pulsed laser beams and receiving the signals reflected from surrounding environmental objects, and can express the spatial structure and object contour of the vehicle's surrounding environment in detail and clearly. The high-resolution camera is used to capture visual information such as road conditions, traffic signs, lane lines, and pedestrians, and relies on deep learning algorithms and visual target detection technologies of convolutional neural networks to identify road objects, traffic signs, and signal lights in real time. The inertial navigation unit, as a compensation and auxiliary sensing device, continuously measures the acceleration, angular velocity, and direction angle changes of the vehicle, and provides real-time information on the vehicle's motion posture and position changes through inertial navigation algorithms.

[0041] To ensure the efficient and real-time transmission of various sensor data, the multi-source perception module uses a time-sensitive network based on the IEEE 802.1Qbv standard for data transmission. The IEEE 802.1Qbv standard realizes the precise division of transmission time slots through a precisely defined gating mechanism to ensure that the data frames of each sensor can be efficiently transmitted within a pre-planned time window, reducing the delay and jitter during data transmission.

[0042] Furthermore, the perception collaborative fusion module integrates the real-time sensor data of the multi-source perception module, including: the perception collaborative fusion module preprocesses and fuses the real-time sensing data obtained by the multi-source perception module. The implementation process of the preprocessing is as follows:

[0043] Step Y1: The perception collaborative fusion module synchronizes all sensor devices in the global clock synchronization network based on the IEEE 802.1ASrev time synchronization protocol, eliminates the time deviation caused by the independent operation of sensors, and makes the acquisition timestamps of various types of data consistent.

[0044] Step Y2: For millimeter-wave radar data, a Kalman filter is used for real-time noise suppression and clutter filtering; for lidar point cloud data, voxelization and statistical filtering methods are used to eliminate point cloud noise and anomalies; for camera image data, Gaussian filtering, histogram equalization, and edge enhancement image processing algorithms are used to improve the clarity of the image, and the spatial consistency alignment between the image data and other sensor data is achieved through internal and external parameter calibration methods; for inertial navigation unit data, zero-bias correction, temperature compensation, and inertial navigation solution algorithms are implemented to eliminate sensor errors and cumulative drifts, and attitude and motion trajectory information is output in real time.

[0045] Step Y3: The perception collaborative fusion module performs spatio-temporal difference and data association on the preprocessing results of various sensors in the same spatio-temporal framework, and realizes the consistency and alignment of various types of data in both spatio-temporal dimensions.

[0046] After the perception collaborative fusion module completes the preprocessing and spatio-temporal alignment of millimeter-wave radar, lidar, camera, and inertial navigation unit data, it further fuses according to the characteristics of different sensor data and the real-time perception requirements of the intelligent driving system. Specifically, it includes: taking the distance and speed measurements provided by the millimeter-wave radar and lidar as inputs, using the unscented Kalman filter to fuse and generate a more accurate and stable target motion state estimate, reducing the random noise and errors in the single-sensor measurement process; for camera images and lidar point cloud data, applying joint calibration and feature matching technologies, through spatial transformation and field-of-view calibration of the point cloud and image, to achieve the fusion of three-dimensional space data and two-dimensional visual data, and construct a three-dimensional environment model with rich spatial information and visual semantic features; combining the convolutional neural network YOLOv5 and the point cloud network PointNet++, extracting the category, position, size, and orientation information of the target in the image and the three-dimensional geometric information of the point cloud data respectively, and then using a multi-layer perceptron for deep fusion processing of multi-feature vectors to improve the accuracy and robustness of environmental perception. The motion information such as vehicle attitude, speed, and angular velocity measured by the inertial navigation unit is used as dynamic constraint conditions to assist in real-time correction of data errors caused by vehicle operation during the fusion process.

[0047] Furthermore, the hybrid gating scheduling module integrates the deterministic gating scheduling mechanism of the IEEE 802.1Qbv standard and the earliest deadline real-time scheduling mechanism, including: The core basis of the hybrid gating scheduling module is the time slot division gating mechanism defined by the IEEE 802.1Qbv standard. The time slot division gating mechanism controls the data sending queue in terms of time on the ports of the Ethernet switch and the end system, and schedules the transmission of network data frames in a deterministic manner. Each sending queue in the network is equipped with a corresponding gate, which has two states: open and closed. A data frame is only allowed to be sent when the gate of the corresponding queue is in the open state; otherwise, it will be temporarily stored in the sending queue and wait for the next gate opening moment. The specific moments of gate opening and closing are determined by the gate control list and executed periodically. The gate control list consists of a series of time slots, and the state and duration of each queue gate are specified in detail in each time slot.

[0048] However, in the network communication of autonomous vehicles, there is not only periodic data, which includes millimeter-wave radar and lidar data streams, but also a large amount of event-driven data, which includes emergency obstacle avoidance instructions and sudden obstacle alarms. The event-driven data has the characteristics of suddenness and strict real-time requirements. Therefore, the hybrid gating scheduling module introduces a real-time online scheduling mechanism based on the earliest deadline first (EDF) to more flexibly meet the dynamic real-time transmission requirements of event-driven data in autonomous vehicles. The real-time online scheduling mechanism based on the earliest deadline adjusts the priority dynamically according to the deadline of each data frame to be sent, so as to ensure the timeliness of real-time data transmission. The specific steps are as follows:

[0049] Step Z1, a data frame is generated from a sensor or a task source and enters the data sending queue. At this time, each data frame carries a clear deadline.

[0050] Step Z2, the earliest deadline scheduler monitors and records the deadline of each data frame in the queue in real time and maintains the deadline information table of the data frames.

[0051] Step Z3, the earliest deadline algorithm continuously sorts the data frames in the queue in real time, arranging them in the order from the nearest to the farthest deadline, that is, the data frame with the nearest deadline is at the front of the queue and has the highest priority; the data frame with a farther deadline is at the back of the queue and has a lower priority.

[0052] Step Z4, when a new data frame arrives at the queue, the earliest deadline algorithm immediately compares the deadline of the new data frame with the deadlines of the existing data frames in the queue; if the deadline of the newly arrived data frame is more urgent than the data frame with the highest priority in the current queue, the new data frame will be assigned a higher transmission priority and immediately placed at the front of the queue, while updating the sorting order of the entire queue.

[0053] Step Z5, when the network link or transmission channel is idle, the earliest deadline scheduler immediately selects the data frame at the front of the current sorting in the queue for transmission.

[0054] Step Z6, when a data frame is successfully transmitted, the earliest deadline scheduler removes it from the queue, updates the queue status, and then executes Step Z3 again to dynamically re-sort the remaining data frames in the queue according to their deadlines.

[0055] By combining the deterministic gating scheduling of the IEEE 802.1Qbv standard and the earliest deadline first (EDF) real-time scheduling mechanism, the utilization efficiency of network resources is improved, and the risks of data transmission jitter and delay caused by improper network resource allocation are reduced. The specific steps are as follows:

[0056] Step S1, the gating scheduling analyzes and classifies all the data frames to be transmitted in the autonomous driving system, distinguishing periodic data frames and event-driven data frames.

[0057] Step S2, for periodic data frames, the gating scheduling constructs a gating control list based on the IEEE 802.1Qbv standard. In the gating control list, the opening and closing times of each queue gate are clearly specified for each time slot, ensuring that the periodic data frames are transmitted at the scheduled time points according to the predetermined transmission rhythm.

[0058] Step S3, during the operation of the gating scheduling, when the sensor generates sudden, non-periodic data frames with strict deadlines, such data are identified as event-driven data frames. At this time, the time-driven data frames are immediately incorporated into an independent real-time scheduling queue for management.

[0059] Step S4, when the event-driven data frames enter the scheduling queue, the gating scheduling immediately starts the earliest deadline first scheduling algorithm, dynamically sorts them according to the deadline information carried by each data frame, and calculates and determines the real-time transmission priority of the data frames.

[0060] Step S5, when the network link is idle and in the gated open state defined by the gated control list, the gated scheduling sends periodic data frames according to the deterministic scheduling rules of the gated control list to maintain the timeliness and stability of periodic data transmission; if the deadline of the event-driven data frame is less than the preset urgent threshold within a 10-millisecond time window, the earliest deadline real-time scheduling mechanism automatically intervenes and preferentially sends the event-driven data frame with a nearby deadline.

[0061] In this embodiment, referring to Figure 4 , the average response time of the IEEE 802.1Qbv static gated scheduling is close to or exceeds 10 ms, reaching up to 10.7 ms at most, and fails to complete the response within the common 10 milliseconds in real-time systems multiple times. Due to its fixed scheduling structure, it cannot handle sudden events immediately, so there is a risk of delay in emergency data transmission. The average response time of the earliest deadline first scheduling is better, maintaining in the range of 4.0 - 4.5 ms, which can significantly shorten the data processing time of sudden events. However, due to the lack of a gated mechanism to guarantee the timing resources of the data channel, there are still certain fluctuations. The response time of the combination of the IEEE 802.1Qbv standard deterministic gated scheduling and the earliest deadline real-time scheduling mechanisms is the best, stabilizing between 2.5 - 2.8 ms, far better than the 10-ms real-time threshold. This shows that the hybrid scheduling of the present invention not only inherits the response ability of the EDF, but also reserves a dynamically allocable high-priority transmission channel with the help of the Qbv mechanism, enabling the system to schedule event data faster, and significantly enhancing the stability and reliability.

[0062] In this embodiment, referring to Figure 5 , the overall delay of the IEEE 802.1Qbv deterministic scheduling is relatively stable, about 3.0 - 3.3 ms, showing strong determinism and low jitter characteristics. However, due to its static allocation scheduling, it cannot dynamically adapt to bursty data streams, so there is a problem of slightly higher delay. The delay fluctuation of the earliest deadline first scheduling is obvious, ranging from 2.4 - 4.3 ms. Although it can preferentially process emergency data frames, due to the lack of structured scheduling for periodic data, the overall jitter is large, and network resources are prone to conflict or congestion. The delay of the combination of the IEEE 802.1Qbv standard deterministic gated scheduling and the earliest deadline real-time scheduling mechanisms is between 2.4 - 2.6 ms, showing extremely small fluctuations and low average delay, indicating that the hybrid scheduling strategy not only inherits the determinism of the Qbv, but also has the real-time response ability of the EDF, achieving better real-time performance and stability while meeting the scheduling requirements of multiple types of data frames.

[0063] The present invention integrates the deterministic gating scheduling mechanism of the IEEE 802.1Qbv standard and the earliest deadline first (EDF) real-time scheduling mechanism, and realizes the classified scheduling and dynamic management of periodic data and event-driven data in the communication network of autonomous vehicles, solving the problem in the prior art that it is difficult to simultaneously consider the real-time data transmission, low latency, low jitter and high bandwidth utilization. The present invention precisely divides the time slots of periodic data frames through a gating control list to ensure their transmission within a strictly controlled time window, meeting their requirements for deterministic and low-jitter transmission. At the same time, the system introduces an online scheduling strategy based on EDF to dynamically respond to bursty event-driven data frames with high real-time requirements. By calculating the deadline of the data frames in real time and adjusting the priorities, the priority scheduling of critical task data is realized, effectively reducing the risk of data failure due to delay.

[0064] As described above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

[0065] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A time-sensitive network gating scheduling system based on autonomous driving, characterized in that: The system includes a multi-source perception module, a perception collaborative fusion module and a hybrid gating scheduling module; The multi-source perception module acquires real-time environmental perception data from the vehicle-mounted sensor network by integrating millimeter-wave radar, laser radar, camera and inertial navigation unit; The perception collaborative fusion module integrates real-time sensor data from multi-source perception modules; The hybrid gated scheduling module integrates the IEEE 802.1Qbv standard deterministic gated scheduling mechanism and the earliest deadline real-time scheduling mechanism. The specific steps are as follows: Step S1, the gated scheduling analyzes and classifies the transmitted data frames to distinguish between periodic data frames and event-driven data frames; Step S2, for periodic data frames, the gating scheduling constructs a gating control list based on the IEEE 802.1Qbv standard; Step S3, in the gated scheduling operation, when the sensor generates a burst of non-periodic data frames with a deadline, the data is identified as an event-driven data frame; Step S4, when the event-driven data frame enters the scheduling queue, the gated scheduling starts the earliest deadline priority scheduling algorithm in real time, dynamically sorts the data frame according to the deadline information carried by each data frame, and calculates and determines the real-time sending priority of the data frame. Step S5, when the network link is idle and the gating is open, the gating scheduling sends periodic data frames according to the control list rules. If the deadline of the event-driven data frame within 10 milliseconds is lower than the urgency threshold, the earliest deadline scheduling mechanism will give priority to sending the event-driven data frame.

2. A time-sensitive network gating scheduling system based on autonomous driving according to claim 1, characterized in that ,The perception collaborative fusion module aligns the sensor data timestamps based on the IEEE 802.1ASrev time synchronization protocol, and preprocesses the data through Kalman filter, Gaussian filter, histogram equalization and edge enhancement algorithm.

3. A time-sensitive network gating scheduling system based on autonomous driving according to claim 1, characterized in that The millimeter-wave radar of the multi-source perception module is used to detect the distance, speed and direction of the target object, the lidar generates high-precision three-dimensional point cloud data, the camera captures visual information, and the inertial navigation unit measures the vehicle acceleration, angular velocity and direction angle changes.

4. The time-sensitive network gating scheduling system based on autonomous driving according to claim 1 is characterized in that: The perception collaborative fusion module fuses millimeter-wave radar and lidar data through unscented Kalman filtering, and uses joint calibration and feature matching technology to fuse camera images and lidar point cloud data to build a three-dimensional environment model.

5. The time-sensitive network gating scheduling system based on autonomous driving according to claim 1 is characterized in that: The hybrid gating scheduling module divides data into periodic data frames and event-driven data frames. The periodic data frames are divided into time slots through a gating control list, and the event-driven data frames dynamically adjust the priority through an earliest deadline algorithm.

6. The time-sensitive network gating scheduling system based on autonomous driving according to claim 1 is characterized in that: The steps of the earliest deadline algorithm include: maintaining a data frame deadline information table; sorting queues from near to far according to deadlines; re-sorting when new data frames arrive; and sending queue front data frames when the link is idle.

7. The time-sensitive network gating scheduling system based on autonomous driving according to claim 1 is characterized in that: The gating control list includes a series of time slots, each of which specifies the opening and closing times of the gating of each queue, and periodic data frames are transmitted at a predetermined rhythm; event-driven data frames enter an independent real-time scheduling queue, and the gating scheduling starts the earliest deadline algorithm in real time, dynamically sorts and adjusts the priority according to the deadline.

8. The time-sensitive network gating scheduling system based on autonomous driving according to claim 1 is characterized in that: The periodic data includes millimeter wave radar and lidar data streams, and the event-driven data includes emergency obstacle avoidance instructions and sudden obstacle alerts.

9. The time-sensitive network gating scheduling system based on autonomous driving according to claim 1 is characterized in that: The perception collaborative fusion module uses the convolutional neural network YOLOv5 and the point cloud network PointNet++ to extract the features of the image and point cloud, and performs deep fusion processing through a multi-layer perceptron.

Images