Remote low-delay control method for unmanned delivery vehicle in complex road environment

Abstract

This application relates to the field of unmanned vehicle technology, specifically to a remote low-latency control method for unmanned delivery vehicles in complex road environments. The method includes: calculating the physical time required for the vehicle to come to a complete stop from its current speed using the vehicle's longitudinal speed and maximum braking deceleration at the current moment, thus obtaining a safe braking duration; predicting the network quality of the unmanned delivery vehicle at each predetermined driving path point within the time period corresponding to the future safe braking duration, using this prediction to calculate the target coding redundancy ratio of the acquired video stream, and dynamically adjusting the generation frequency of FEC packets; quantifying the total effective data volume that the communication link can carry, comparing it with the total data volume required for the vehicle's video receiver player to maintain current playback, assessing the bandwidth supply deficit, and adjusting the video coding parameters of the video transmitter. This application eliminates the feedback lag problem in the video stream control of unmanned delivery vehicles.

Description

Technical Field

[0001] This application relates to the field of unmanned vehicle technology, specifically to a remote low-latency control method for unmanned delivery vehicles in complex road environments. Background Technology

[0002] In the deployment scenarios of Level 4 autonomous driving, unmanned delivery vehicles need to operate in complex urban mixed traffic flow. Due to the long-tail problem of onboard perception, remote takeover becomes a necessary fallback mechanism to ensure the safe operation of vehicles. The remote takeover system relies on real-time communication technology to transmit video streams from the onboard cameras back to the remote cockpit, enabling the human driver to monitor and control the vehicle in real time.

[0003] However, in actual operation, the high-speed mobility of vehicles causes drastic spatiotemporal fluctuations in network channel quality. When a vehicle rapidly enters an area with signal obstruction, such as under an overpass, at the entrance of a tunnel, or during base station cell handover, the available network bandwidth drops instantly. Existing video flow control solutions mainly adjust the transmission bitrate based on the packet loss rate or delay gradient fed back from the receiver. However, existing technologies only reduce the bitrate after packet loss or delay has already occurred, which can easily miss the optimal time for risk avoidance and result in feedback lag. Furthermore, the optimization goals of existing technologies are usually to avoid network congestion or achieve optimal image quality, without considering the physical motion of the vehicle. When a vehicle is moving at a certain speed, its physical braking distance and the time required are fixed objective physical quantities. If the video transmission delay exceeds this physical braking time, the image seen by the remote driver will be severely lagging behind the real world, causing the driver's operating commands to be ineffective within a safe distance in the event of a sudden danger. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a remote low-latency control method for unmanned delivery vehicles in complex road environments, thereby resolving the existing issues.

[0005] The remote low-latency control method for unmanned delivery vehicles in complex road environments proposed in this application adopts the following technical solution:

[0006] One embodiment of this application provides a remote low-latency control method for unmanned delivery vehicles in complex road environments, the method comprising the following steps:

[0007] By calculating the longitudinal speed and maximum braking deceleration of the unmanned delivery vehicle at the current moment, the physical time required for the vehicle to come to a complete stop from the current speed is obtained, thus obtaining the safe braking duration;

[0008] Within the time period corresponding to the future safe braking duration, the network quality of the unmanned delivery vehicle at each predetermined driving path point is predicted by using the historical network quality of each predetermined driving path point and the network quality at the current moment. This prediction is then used to calculate the target coding redundancy ratio of the collected video stream and dynamically adjust the generation frequency of FEC packets.

[0009] During the time period, the total amount of effective data that the communication link can carry is quantified by using the predicted network quality of the vehicle at each predetermined driving path point and the characteristic of the confidence level of the predicted value decaying over time. This is then compared with the total amount of data required for the video receiver player of the vehicle to maintain the current playback, to assess the bandwidth supply deficit, and to adjust the video encoding parameters of the video transmitter.

[0010] In one embodiment, the process of obtaining the safe braking duration is as follows:

[0011] Calculate the ratio of the longitudinal vehicle speed to the maximum braking deceleration, amplify the ratio using a preset safety redundancy coefficient, and determine the sum of the amplified result and the inherent response delay constant of the vehicle control.

[0012] The safe braking duration is obtained by applying boundary constraints to the sum.

[0013] In one embodiment, the network quality evaluation metrics include network bandwidth and packet loss rate.

[0014] In one embodiment, predicting the network quality of the unmanned delivery vehicle at each predetermined driving path point includes:

[0015] The system queries the historical network quality of the unmanned delivery vehicle at each predetermined driving route point. If the query is successful, the system obtains the historical average bandwidth and historical average packet loss rate for the corresponding location. If the query is unsuccessful, the system presets the bandwidth and packet loss rate for the corresponding location based on the measured bandwidth at the current moment.

[0016] The bandwidth obtained by the unmanned delivery vehicle at each predetermined driving path point is weighted and summed with the measured bandwidth at the current moment to obtain the bandwidth prediction result of the unmanned delivery vehicle at each predetermined driving path point. Correspondingly, the packet loss rate prediction result is obtained.

[0017] In one embodiment, the calculation of the target coding redundancy ratio of the acquired video stream includes:

[0018] Extract the maximum value of the predicted packet loss rate of the unmanned delivery vehicle among all predetermined driving path points, calculate the product of the maximum value and the preset redundancy multiple coefficient, and apply boundary constraints to the product to obtain the target coding redundancy ratio.

[0019] In one embodiment, the total effective data volume that the quantized communication link can carry includes:

[0020] The time period is divided into time units, and the corresponding time of the unmanned delivery vehicle at each predetermined driving path point is obtained and recorded as the path time. The bandwidth prediction value and packet loss rate prediction value corresponding to the path time that is closest to the center time of each time unit are used as the unit prediction bandwidth and unit prediction packet loss rate of each time unit.

[0021] Based on the predicted bandwidth and predicted packet loss rate of each time unit, the effective transmission volume of each time unit is calculated, and after summing them, the total effective data that the communication link can carry is obtained.

[0022] In one embodiment, calculating the effective transmission amount of each time unit includes:

[0023] A nonlinear positive mapping is performed on the predicted packet loss rate of each time unit to obtain the dynamic loss resistance cost coefficient. The product of the dynamic loss resistance cost coefficient and the predicted packet loss rate of the corresponding time unit is calculated and denoted as the first product. The difference between the natural number 1 and the first product is determined.

[0024] The product of the difference and the predicted bandwidth of the corresponding time unit is calculated and denoted as the second product. The second product is subject to a non-negative constraint. Based on the result of the non-negative constraint, the corresponding length of the time unit, and the preset time confidence attenuation weight, the effective transmission amount of each time unit is determined.

[0025] In one embodiment, the bandwidth supply deficit is the difference between the total amount of data and the total amount of effective data.

[0026] In one embodiment, adjusting the video encoding parameters of the video transmitter includes:

[0027] If the bandwidth supply deficit is greater than 0, the video transmitter degradation strategy is immediately triggered, and the target encoding bitrate of the transmitter's video encoder is reduced. If the bandwidth supply deficit is less than or equal to 0, the transmitter maintains the current video encoding parameters.

[0028] In one embodiment, the buffer backlog duration is calculated based on the amount of data to be decoded in the buffer of the vehicle's video receiver player, and compared with the safe braking duration to obtain the transmission delay deficit. Based on the magnitude of the transmission delay deficit, the playback rate of the vehicle's video receiver is dynamically adjusted to control the playback strategy of the player.

[0029] This application has at least the following beneficial effects:

[0030] This application calculates the safe braking time of the unmanned delivery vehicle and uses it as the time reference for video stream control. This ensures that the video transmission delay does not exceed the vehicle's current physical braking time, guaranteeing that the timeliness of the footage seen by the remote driver always meets the physical condition of seeing danger and being able to brake in time. This eliminates the risk of collisions caused by excessive delay and ensures the physical safety of takeover operations. Secondly, by predicting the network quality of the unmanned delivery vehicle at each predetermined driving path point, the target coding redundancy ratio of the acquired video stream is calculated. The generation frequency of FEC packets is dynamically adjusted, further regulating the playback rate of the vehicle's video receiver and controlling the player's playback strategy. This proactive mechanism can anticipate future capacity shortages and actively empty the buffer before the vehicle enters a weak network area. This proactive mechanism avoids video stuttering and data backlog caused by passive responses when bandwidth suddenly drops, unlike the reactive mechanism of existing technologies that only adjust the playback speed based on the current buffer level. It can intervene in advance when the buffer has not yet overflowed but the network is about to deteriorate, thus avoiding the safety hazards caused by feedback lag. This application solves the safety problem of takeover failure due to network fluctuations in remote driving by introducing vehicle dynamics constraints, significantly improving the safety level of remote takeover of unmanned delivery vehicles. Attached Figure Description

[0031] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 A flowchart illustrating the steps of a remote low-latency control method for unmanned delivery vehicles in complex road environments provided in this application. Detailed Implementation

[0033] To further illustrate the technical means and effects adopted by this application to achieve the intended inventive objective, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a remote low-latency control method for unmanned delivery vehicles in complex road environments proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0035] The following description, in conjunction with the accompanying drawings, details a specific scheme for a remote, low-latency control method for unmanned delivery vehicles in complex road environments provided in this application.

[0036] This application provides a method for remote low-latency control of unmanned delivery vehicles in complex road environments, specifically, the following method is provided. Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps:

[0037] Step S001: Synchronous acquisition and alignment of multi-source heterogeneous data.

[0038] In the remote control system of unmanned delivery vehicles, vehicle motion data, path planning data, and video transmission status data are typically generated by different hardware modules or software processes, each with its own independent timestamp. To eliminate data timing discrepancies caused by processing delays, a data alignment operation based on a unified clock is first performed.

[0039] Specifically, a system clock of a master control unit is set as a unified time base. At the trigger time of each control cycle... In this embodiment, the process is triggered every 50 milliseconds, and the following three types of data are collected in parallel and aligned:

[0040] Vehicle dynamics data: This data is obtained by reading the vehicle's real-time motion status via the vehicle's internal CAN (Controller Area Network) bus interface. This data includes the original timestamp of its generation. longitudinal speed Calculate the time deviation. Combined with the acceleration information from the previous moment, the velocity value is calibrated to [the specified value] using a linear extrapolation algorithm. At any given moment, obtain the vehicle's current speed. Simultaneously, the vehicle's maximum braking deceleration is read from the preset vehicle chassis configuration file. (unit: This parameter characterizes the ultimate capability of a vehicle's mechanical braking system.

[0041] Navigation path data: Obtain the sequence of predetermined driving path points from the autonomous driving planning module of the unmanned delivery vehicle. The sequence consists of several pairs of bytes. Composition, in which, This indicates the geographical coordinates (latitude and longitude) that the vehicle is expected to travel through. Relative to The predicted arrival time offset (in seconds). Conduct inspections and remove [illegible characters] Expired path points ensure that the path sequence only contains future valid trajectories.

[0042] Transmission status data: Obtain the status information of the receiving player through the callback interface of the video transmission SDK. Read the amount of data to be decoded from the buffer. (Unit: bits), this value reflects the current time. Backlogged data that has arrived at the receiving end but has not yet been played. Simultaneously, obtain the real-time bandwidth estimate of the communication link at the current moment. (Unit: bps), serving as the baseline value for subsequent network predictions.

[0043] Step S002: Calculate the physical time required for the vehicle to come to a complete stop from its current speed using the longitudinal speed and maximum braking deceleration of the unmanned delivery vehicle at the current moment, and obtain the safe braking duration.

[0044] To quantify the physical process of vehicle braking as a time constraint for network transmission, the physical time required for the vehicle to come to a complete stop from its current speed is calculated and defined as the safe braking duration. This duration constitutes the upper limit of the time integral for subsequent network capacity assessment.

[0045] To improve the system's engineering robustness and avoid hazards caused by excessively short calculated braking times due to changes in road surface friction coefficients or sensor noise, this embodiment introduces a safety redundancy coefficient and boundary clamping logic. Safe braking duration. The calculation formula is:

[0046] In the formula, This indicates that under ideal friction conditions, the vehicle's speed from its current speed... The purely physical braking time required to decelerate to 0. The larger the value, the larger this item is, indicating that the upper limit of allowable transmission delay can be appropriately relaxed when traveling at high speeds; The larger the value, the smaller this value, indicating a car with better braking performance and a shorter reaction time for the network. This represents the inherent response delay constant of vehicle control, which is set to 0.2 seconds in this embodiment. This item covers the mechanical delay from when the network issues a command to when the vehicle actuator (such as the brake pump) begins to act, as well as the minimum physical delay of the network round-trip transmission. The value represents the safety redundancy coefficient, which is set to a range of 1.1 to 1.3. In this embodiment, the value is 1.2. Considering that rain and snow will reduce the road surface friction coefficient and the actual braking distance may become longer, the safety redundancy coefficient is introduced to amplify the calculation results in order to reserve a more sufficient safety margin. This represents the boundary constraint function, used to convert the calculation results... Limited to the range Inside. This represents the minimum look-ahead time, which is set to 1.0 second in this embodiment. Even if the vehicle is stationary or moving at extremely low speeds, a minimum look-ahead prediction window needs to be maintained to prevent calculation divergence due to an excessively small window. This indicates the maximum effective prediction time, which is set to 5.0 seconds in this embodiment. Since the long-term prediction accuracy of network state is low, an excessively long time window will introduce huge cumulative errors. Therefore, this upper limit is set to ignore network states that exceed this range.

[0047] Step S003: Within the time period corresponding to the future safe braking duration, predict the network quality of the unmanned delivery vehicle at each predetermined driving path point by using the historical network quality of each predetermined driving path point of the unmanned delivery vehicle and the network quality at the current moment.

[0048] At the present moment Subsequent safe braking duration Within the time window, it is necessary to predict the future. The network quality along the vehicle's route during this period. Combining historical database data with current measured values, a path-related network state sequence is generated. .

[0049] A cold start degradation strategy is configured for situations with no historical data to ensure operation on any road segment. The specific execution logic is as follows:

[0050] Database query: Traversing the future Each waypoint the vehicle passed through during this period using geographical coordinates Use the index to query the pre-built historical network quality database. If the query finds a match, retrieve the historical average bandwidth for that location. and historical average packet loss rate .

[0051] If the query fails to find a match, it falls under the cold start scenario, where the vehicle is traveling on a new road segment not yet included in the database. This triggers a degradation strategy: the bandwidth at that location is set to the currently measured bandwidth. The conservative estimate is used to set the bandwidth at this location point in this embodiment. And set the packet loss rate to a preset safe default value, specifically as follows: .

[0052] Furthermore, considering that historical data may be outdated, the current measured bandwidth is used. For the query results or those generated by downgrading After weighted correction, the predicted bandwidth is obtained. The correction logic uses time-decay weighting, and the specific expression is as follows:

[0053] ;in, The trust decay factor is set to 0.8 in this embodiment. The time normalization constant is set to 1 second in this embodiment to eliminate... Dimensions. When When it approaches 0 (i.e., at the current moment), When the value is close to 1, the predicted value is mainly determined by the measured value; as... Increase (i.e., future time). As the value decreases, the predicted value gradually returns to the historical statistical mean. This ensures that the acquisition of the predicted bandwidth conforms to the engineering principles of recent data measurement and long-term statistical data.

[0054] Similarly, using the current measured packet loss rate For the query results or those generated by downgrading After weighted correction, the predicted packet loss rate is obtained. The specific expression is:

[0055]

[0056] Finally, a path-related network state sequence is generated. The corrected forecast data is organized in chronological order, and the output sequence is generated. Until the covered time range exceeds Up to this point. This sequence precisely describes the expected communication link capacity at every moment during the braking process the vehicle is about to undergo.

[0057] Step S004: Calculate the target coding redundancy ratio of the acquired video stream and dynamically adjust the generation frequency of FEC packets.

[0058] Based on path-associated network state sequences Adjusting the loss-resistance strategy at the video transmitter ensures the integrity of the video stream during future transmission, preventing effective capacity reduction due to packet loss. Transmitter control is designed to defend against impending packet loss. If the sequence... The system indicates that the vehicle is about to enter a high packet loss area. Therefore, it is necessary to increase the redundancy of the video stream before the vehicle arrives. Specifically:

[0059] First, traverse the sequence. The predicted packet loss rate for all time points is used to extract the maximum value, which is then denoted as the maximum predicted packet loss rate. This value represents the duration of safe braking. The worst possible operating conditions for the communication link.

[0060] Secondly, according to Calculate the target coding redundancy ratio (Forward error correction redundancy). The calculation formula is as follows, with boundary constraints:

[0061] ;in, This represents the redundancy factor, which is set to 1.5 in this embodiment. This means that in order to resist 1% packet loss, it is usually necessary to configure more than 1% of redundant data packets to ensure that the original data can be recovered with a high probability in the event of random packet loss. The larger the size, the stronger the damage resistance, but the more bandwidth it consumes. This represents the minimum basic redundancy, which is set to 5% in this embodiment. Even if the prediction network is very good, a small amount of redundancy is reserved to deal with sudden noise. This indicates the maximum redundancy limit, which is set to 50% in this embodiment to prevent redundant data from occupying all the bandwidth under extremely weak network conditions, thus preventing the transmission of effective video.

[0062] Calculate the target coding redundancy ratio The data is sent in real time to the vehicle-mounted video encoder (e.g., H.264 / H.265 hardware encoder). The encoder dynamically adjusts the generation frequency of FEC packets based on this parameter, realizing forward defense based on path prediction. It adjusts in advance according to the predicted value of the path ahead, ensuring that the video stream has sufficient damage resistance the moment the vehicle enters the weak network area.

[0063] Step S005: By using the predicted network quality of the vehicle at each predetermined driving path point and the characteristics of its confidence decay over time, the total amount of effective data that the communication link can carry is quantified, and compared with the total amount of data required by the vehicle's video receiver player to maintain the current playback, the bandwidth supply deficit is assessed, and the video encoding parameters of the video transmitter are adjusted.

[0064] In order to accurately measure the network bandwidth that changes nonlinearly over time, the continuous time window is first discretized and sliced.

[0065] With safe braking duration Assuming the total time span for the evaluation, a small time step is set. In this embodiment The step size should be less than the typical period of network state changes to ensure computational accuracy. The time period is divided into A series of evaluation time units ( ),in .

[0066] For each time unit Calculate its center time and in sequence Searching for and The closest moment is denoted as the path moment. The bandwidth prediction value and packet loss rate prediction value corresponding to the path moment are used as the unit prediction bandwidth for that time unit. And unit predicted packet loss rate .

[0067] Furthermore, calculate the actual number of effective bits that can be transmitted in each time unit after deducting various losses, i.e., the effective transmission amount of the unit. .

[0068] To address the issue of potentially negative results due to simple subtraction in existing technologies, and to tackle the engineering challenges of inaccurate long-term predictions, this embodiment employs a computational model that incorporates non-negativity constraints and confidence decay. (Effective transmission volume of the computational unit) The specific expression is:

[0069] In the formula, This represents a non-negativity constraint function. In extremely poor network environments, the retransmission overhead caused by packet loss may exceed the physical bandwidth. This function ensures that the calculated effective capacity is at least 0, avoiding the physical fallacy of negative capacity and guaranteeing the logical stability of the algorithm under extremely weak network conditions. The dynamic loss mitigation cost coefficient is obtained by performing a nonlinear positive mapping on the predicted packet loss rate of each time unit. In this embodiment... This represents the effort to resist The packet loss rate, the proportion of redundant data that needs to be added, A minimum constant greater than 0 is preset to avoid a denominator of 0 in this embodiment. . The higher, The larger this term is (showing non-linear growth), the more it reflects the erosive effect of packet loss on effective bandwidth. The time confidence decay weight is used because the error of network prediction accumulates over time, and the confidence of long-term predictions is lower than that of recent predictions. It is a time index The decreasing coefficient, in this embodiment This weighting is used to discount the long-term capacity, employing a conservative strategy to address future uncertainties and ensuring that the calculated total capacity represents a safe lower bound with high confidence. This is denoted as the first product. This is denoted as the second product.

[0070] All The cumulative transmission capacity is obtained by summing the effective transmission amounts of each time unit. The total effective data volume that a communication link can carry is specifically expressed as:

[0071]

[0072] The cumulative transmission capacity limit represents the maximum amount of data that the communication network can objectively transmit before the vehicle must stop due to a hazard, after deducting all losses and reserving a safety margin. Bits of data. This metric integrates the vehicle's physical motion constraints (represented by the upper limit of integration time) with network environment constraints (represented by bandwidth integration) across layers.

[0073] Meanwhile, to prevent the risk of future supply disruptions, a supply shortage assessment is introduced, and future supply shortages are calculated. Total amount of data required to maintain current playback within a given time period , ,Compare and Computational bandwidth supply deficit , ,like This indicates that even if there is no current lag, future network supply will be severely insufficient, leading to playback stuttering or interruptions. In this case, it is necessary to reduce the bitrate or increase FEC redundancy in advance.

[0074] Therefore, if If a future network supply shortage is determined, a sender degradation is immediately triggered, specifically: based on... Amplitude calculation bitrate reduction ratio coefficient Where max() is the function to find the maximum value. The on-board terminal encoder adjusts the bitrate based on the target output bitrate currently used by the transmitting end's video encoder. The target encoding bitrate is lowered. If the lowered target bitrate is lower than the preset resolution switching threshold, which is set to 1.5Mbps in this embodiment, the video resolution is simultaneously lowered from the current level to the next level, thereby eliminating the bandwidth supply deficit by reducing the amount of source data generated.

[0075] And if , determine the future If the network has sufficient cumulative capacity within the specified time, no degradation operation will be performed on the sending end, and the sending end will maintain the current high bitrate and high resolution video encoding parameters.

[0076] Step S006: Adjust the playback rate of the video receiver in the vehicle and control the playback strategy of the player.

[0077] buffer data volume to be decoded This represents the existing backlog of debt at the receiving end that must be played and processed. Therefore, the buffer backlog duration is calculated. ,in, ,in, The average bitrate of the current video stream. The buffer backlog duration. With safe braking duration Compare and calculate the transmission delay deficit. The specific expression is:

[0078]

[0079] like This indicates that the current video delay is within a safe range. The current state is deemed safe, and no intervention is required.

[0080] like This indicates a transmission deficit. This means there's excessive backlog, and the video latency has exceeded the physical braking limit, posing a safety hazard. To ensure safety, this excess latency must be eliminated immediately by either accelerating playback or discarding the video.

[0081] To translate the calculated transmission deficit into specific playback control commands, the risk is proactively eliminated before the actual delay exceeds the safety threshold through speed adjustment or frame skipping operations at the receiving end.

[0082] Based on transmission delay deficit The numerical value is used to determine the risk level and calculate the target playback rate, specifically:

[0083] like This indicates that the current image is delayed within the safe braking duration. Internally, set the target playback rate multiplier for the receiving player. Maintain at 1.0 (normal speed) without intervention.

[0084] like ,in, The preset critical threshold, specifically 0.5 seconds in this embodiment, indicates a slight transmission deficit that can be recovered by accelerating playback.

[0085] The target playback speed is calculated using an acceleration formula based on time deficit, specifically the target playback rate multiplier. The expression is:

[0086] ;in, To adjust the sensitivity coefficient, this embodiment sets it to 1.0. A higher value requires a more aggressive response to deficits and faster acceleration; a lower value results in smoother playback speed changes, a better user experience, but slower latency. This is the maximum allowed playback speed for the player; in this embodiment, it's set to 2x to ensure the catching-up process doesn't affect the user's viewing experience. `min()` is the function that takes the minimum value. This target playback speed multiplier is applied by calling the player's speed adjustment interface. It also incorporates the audio time stretching algorithm (WSOLA) to ensure that the pitch remains unchanged.

[0087] when or calculated When the playback speed exceeds the maximum allowed by the player, it indicates severe backlog, and linear acceleration can no longer solve the problem, triggering a frame skipping instruction. Specifically:

[0088] The player received a frame skipping instruction. Then, immediately clear all P-frames and B-frames in the current video buffer queue, and request the most recent keyframe (I-frame) from the streaming media server, directly jumping to the latest live feed (Seek-to-Live). This operation can instantly eliminate... This ensures that the driver sees the latest real-time footage.

[0089] Secondly, to prevent the system from falling into congestion during periods of continuous network jitter. Frame skipping Clear Instantaneous backlog The infinite loop of frame skipping causes frequent screen flickering, making the video unwatchable. This embodiment introduces a cooling protection mechanism. The system maintains a timer. Record the last time a frame skipping instruction was executed. Before generating a new frame skipping instruction, check the current time. Does it meet the requirements? , Cooling time, denoted as the duration threshold, is set to 2 seconds in this embodiment;

[0090] If the conditions are met, then frame skipping is performed normally, and updates are made. ;

[0091] If the conditions are not met (during the cooling-off period), frame skipping will be temporarily suspended. Instead, playback will be forced to maintain the maximum allowed speed, prioritizing the continuity of the image over real-time performance, until the cooling-off period ends.

[0092] This completes the control of the video receiver and video transmitter of the remote control system for the unmanned delivery vehicle, reducing the control delay effect of the unmanned delivery vehicle under complex road conditions.

[0093] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0094] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0095] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some of the technical features, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A remote low-latency control method for unmanned delivery vehicles in complex road environments, characterized in that, The method includes the following steps: By calculating the longitudinal speed and maximum braking deceleration of the unmanned delivery vehicle at the current moment, the physical time required for the vehicle to come to a complete stop from the current speed is obtained, thus obtaining the safe braking duration; Within the time period corresponding to the future safe braking duration, the network quality of the unmanned delivery vehicle at each predetermined driving path point is predicted by using the historical network quality of each predetermined driving path point and the network quality at the current moment. This prediction is then used to calculate the target coding redundancy ratio of the collected video stream and dynamically adjust the generation frequency of FEC packets. During the time period, the total amount of effective data that the communication link can carry is quantified by using the predicted network quality of the vehicle at each predetermined driving path point and the characteristic of the confidence level of the predicted value decaying over time. This is then compared with the total amount of data required for the video receiver player of the vehicle to maintain the current playback, to assess the bandwidth supply deficit, and to adjust the video encoding parameters of the video transmitter.

2. The remote low-latency control method for unmanned delivery vehicles in complex road environments as described in claim 1, characterized in that, The process for obtaining the safe braking duration is as follows: Calculate the ratio of the longitudinal vehicle speed to the maximum braking deceleration, amplify the ratio using a preset safety redundancy coefficient, and determine the sum of the amplified result and the inherent response delay constant of the vehicle control. The safe braking duration is obtained by applying boundary constraints to the sum.

3. The remote low-latency control method for unmanned delivery vehicles in complex road environments as described in claim 1, characterized in that, The network quality assessment metrics include network bandwidth and packet loss rate.

4. The remote low-latency control method for unmanned delivery vehicles in complex road environments as described in claim 3, characterized in that, The prediction of network quality for unmanned delivery vehicles at each predetermined driving path point includes: The system queries the historical network quality of the unmanned delivery vehicle at each predetermined driving route point. If the query is successful, the system obtains the historical average bandwidth and historical average packet loss rate for the corresponding location. If the query is unsuccessful, the system presets the bandwidth and packet loss rate for the corresponding location based on the measured bandwidth at the current moment. The bandwidth obtained by the unmanned delivery vehicle at each predetermined driving path point is weighted and summed with the measured bandwidth at the current moment to obtain the bandwidth prediction result of the unmanned delivery vehicle at each predetermined driving path point. Correspondingly, the packet loss rate prediction result is obtained.

5. The remote low-latency control method for unmanned delivery vehicles in complex road environments as described in claim 4, characterized in that, The calculation of the target coding redundancy ratio of the acquired video stream includes: Extract the maximum value of the predicted packet loss rate of the unmanned delivery vehicle among all predetermined driving path points, calculate the product of the maximum value and the preset redundancy multiple coefficient, and apply boundary constraints to the product to obtain the target coding redundancy ratio.

6. The remote low-latency control method for unmanned delivery vehicles in complex road environments as described in claim 1, characterized in that, The total amount of effective data that the quantized communication link can carry includes: The time period is divided into time units, and the corresponding time of the unmanned delivery vehicle at each predetermined driving path point is obtained and recorded as the path time. The bandwidth prediction value and packet loss rate prediction value corresponding to the path time that is closest to the center time of each time unit are used as the unit prediction bandwidth and unit prediction packet loss rate of each time unit. Based on the predicted bandwidth and predicted packet loss rate of each time unit, the effective transmission volume of each time unit is calculated, and after summing them, the total effective data that the communication link can carry is obtained.

7. The remote low-latency control method for unmanned delivery vehicles in complex road environments as described in claim 6, characterized in that, The calculation of the effective transmission amount of each time unit includes: A nonlinear positive mapping is performed on the predicted packet loss rate of each time unit to obtain the dynamic loss resistance cost coefficient. The product of the dynamic loss resistance cost coefficient and the predicted packet loss rate of the corresponding time unit is calculated and denoted as the first product. The difference between the natural number 1 and the first product is determined. The product of the difference and the predicted bandwidth of the corresponding time unit is calculated and denoted as the second product. The second product is subject to a non-negative constraint. Based on the result of the non-negative constraint, the corresponding length of the time unit, and the preset time confidence attenuation weight, the effective transmission amount of each time unit is determined.

8. The remote low-latency control method for unmanned delivery vehicles in complex road environments as described in claim 1, characterized in that, The bandwidth supply deficit is the difference between the total amount of data and the total amount of effective data.

9. The remote low-latency control method for unmanned delivery vehicles in complex road environments as described in claim 1, characterized in that, The adjustment of video encoding parameters at the video transmitter includes: If the bandwidth supply deficit is greater than 0, the video transmitter degradation strategy is immediately triggered, and the target encoding bitrate of the transmitter's video encoder is reduced. If the bandwidth supply deficit is less than or equal to 0, the transmitter maintains the current video encoding parameters.

10. The remote low-latency control method for unmanned delivery vehicles in complex road environments as described in claim 1, characterized in that, The buffer backlog duration is calculated based on the amount of data to be decoded in the buffer of the video receiver player in the vehicle, and compared with the safe braking duration to obtain the transmission delay deficit. Based on the magnitude of the transmission delay deficit, the playback rate of the video receiver in the vehicle is dynamically adjusted to control the playback strategy of the player.

Images