CN-122027671-A - Vehicle data edge calculation processing method

CN122027671ACN 122027671 ACN122027671 ACN 122027671ACN-122027671-A

Abstract

The invention relates to the technical field of intelligent transportation and Internet of vehicles, in particular to a vehicle data edge calculation processing method, which comprises a state acquisition, differential construction, cloud edge cooperation and prediction control module, wherein a system acquires real-time data to compare with a reference model to construct a state differential tensor, a cloud global environment field model is updated through uploading tensor, field gradient parameters are received, the core is that the field gradient parameters are mapped into cost function constraint terms of a model prediction controller, an optimal control sequence is solved to adjust equipment, the transition from full data uploading to deviation triggering on-demand reporting is realized, communication load is greatly reduced by semantic compression, and the problem of channel congestion under a high concurrency scene is effectively solved.

Inventors

YU SHUANGSHUANG
YAN ZHIXIANG
XIE BENJU

Assignees

青岛维特海科科技有限公司
中恒诺森(青岛)科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (8)

1. The vehicle data edge calculation processing method is characterized by comprising the following specific steps: Step one, collecting real-time running state data of controlled equipment, and calling a preset reference dynamic behavior model; Step two, uploading the state difference tensor to a cloud analysis server to update a global environment field model in the cloud analysis server, and receiving field gradient parameters issued by the cloud analysis server, wherein the field gradient parameters are calculated based on the global environment field model and represent field distribution gradients of local environments where controlled equipment is located; And thirdly, constructing a model prediction controller integrating field coupling items, mapping field gradient parameters into field constraint penalty items of a cost function, solving an optimal control sequence by using the controller, generating a device control instruction, and inputting the device control instruction into an executing mechanism of the device to adjust the running state of the device.
2. The vehicle data edge calculation processing method according to claim 1, wherein the step one includes: S11, acquiring real-time running state data of controlled equipment by using an equipment body sensor and a measurement and control unit, wherein the real-time running state data comprises space position coordinates, speed vectors, attitude angles, angular rate data and internal state variable data; S12, calling a reference dynamic behavior model stored in a local edge computing unit, wherein the reference dynamic behavior model describes an ideal dynamic process and state response of the equipment under the condition of no external field interference; S13, inputting real-time running state data into a reference dynamic behavior model, calculating characteristic difference measurement between an actual state and an ideal state, extracting characteristic data only containing significant deviation information, and generating a state difference tensor.
3. The vehicle data edge calculation processing method according to claim 2, characterized in that the step two includes: S21, establishing communication connection with a cloud analysis server, and uploading a state difference tensor and current absolute space position coordinates of controlled equipment; S22, receiving field gradient parameters fed back by a cloud analysis server in response to aggregation and fusion processing of state difference tensors uploaded by the cloud analysis server on a plurality of controlled devices, wherein the field gradient parameters are gradient vectors calculated by the cloud analysis server according to an aggregate updated global environment field model and aiming at the current state of the controlled devices and field distribution in a preset prediction time domain; S23, storing the received field gradient parameters in a local buffer area for subsequent control calculation.
4. A vehicle data edge calculation processing method according to claim 3, wherein step three includes: S31, initializing a model prediction controller of an integrated field coupling item, and setting a prediction time domain and a control time domain; s32, constructing a cost function, wherein the cost function comprises a state set point tracking error term, a control input change rate penalty term and a field constraint penalty term; S33, extracting field gradient parameters in the local buffer area, and substituting the field gradient parameters serving as weighting factors of field constraint penalty items into a cost function; and S34, solving an optimization problem for minimizing a cost function on the premise of meeting the constraints of equipment dynamics and processes, and obtaining an optimal control sequence containing control quantity set values at a plurality of moments in the future.
5. The vehicle data edge computing method according to claim 4, wherein the third step further includes a network anomaly handling mechanism: s35, monitoring the connection state with the cloud analysis server in real time; s36, when the connection interruption is monitored and the interruption time is smaller than a preset safe time delay threshold, calling the field gradient parameter updated last time in the local buffer area; S37, continuously deducting the predicted motion trail of the equipment in the current local field environment by using the controller based on the called field gradient parameters and the current real-time running state data of the equipment, and generating a retentivity control instruction.
6. The vehicle data edge calculation processing method according to claim 4, characterized in that said step three further includes: s38, presetting a field gradient safety threshold, and comparing the modular length of the current field gradient parameter with the field gradient safety threshold to obtain a field gradient risk assessment result; S39, constructing a gradient risk grade label according to a field gradient risk assessment result, wherein the step comprises the following steps: When the module length of the field gradient parameter is smaller than or equal to the field gradient safety threshold value, the field gradient risk assessment result is normal, wherein the module length indicates that the equipment is in a low gradient stable field region; When the module length of the field gradient parameter is larger than the field gradient safety threshold and smaller than or equal to a preset high risk early warning threshold, the module length of the field gradient parameter indicates that the equipment is close to a high gradient field intensity change area, and the field gradient risk assessment result is a primary early warning; When the module length of the field gradient parameter is larger than the high-risk early warning threshold value, the field gradient risk assessment result is a secondary alarm, and the secondary alarm is stronger than the control constraint corresponding to the primary early warning.
7. The vehicle data edge computing method according to claim 6, wherein the third step further includes policy execution based on a gradient risk level tag: s391, when the field gradient risk assessment result is normal, the controller takes the reference set point tracking as a main optimization target and outputs a conventional adjustment control instruction; S392, when the field gradient risk assessment result is primary early warning, extracting corresponding field gradient parameters, dynamically increasing the weight coefficient of a field constraint penalty term in the cost function, and generating a preventive suppression control instruction; S393, when the field gradient risk assessment result is a secondary alarm, extracting corresponding field gradient parameters, mapping the field gradient parameters into virtual constraint boundaries, compacting a feasible solution search space of the controller, and generating evasive or protective emergency control instructions.
8. The vehicle data edge calculation processing method according to claim 1, wherein the generation of the state difference tensor follows the following rule: generating and marking a state difference tensor at the moment as effective uploading data only when the deviation between the real-time running state data and the reference dynamic behavior model exceeds a preset trigger threshold; When the deviation does not exceed a preset trigger threshold, the equipment is judged to be in a reference steady state or a small disturbance range, and a state difference tensor is not generated, so that the communication load between the edge side and the cloud analysis server is reduced.

Description

Vehicle data edge calculation processing method Technical Field The invention relates to the technical field of intelligent transportation and Internet of vehicles, in particular to a vehicle data edge computing and processing method. Background In the current intelligent networking vehicle and intelligent traffic running environment, the controlled equipment needs to process massive sensor data in real time to maintain safe running, the existing scheme generally adopts a centralized cloud control or full data uploading architecture, namely, the vehicle uploads all position, speed and state data to the cloud, and calculates and issues specific control instructions, although the scheme has global planning capability, the scheme has high dependence on a high-bandwidth low-delay communication link, so that the contradiction between global optimal and real-time response is faced when the high-frequency dynamic data are processed, particularly under the network fluctuation or weak network condition, the transmission of the full data not only causes the excessive occupation of uplink bandwidth, but also causes delayed instruction issuing due to the increase of control loop delay, the deadlock or vehicle collision risk is extremely easy to be caused, and the control continuity and safety under complex traffic are difficult to be ensured. Disclosure of Invention In order to solve the above technical problems, the present invention provides a vehicle data edge computing method, and specifically, the technical solution of the present invention includes: Step one, collecting real-time running state data of controlled equipment, and calling a preset reference dynamic behavior model; Step two, uploading the state difference tensor to a cloud analysis server to update a global environment field model in the cloud analysis server, and receiving field gradient parameters issued by the cloud analysis server, wherein the field gradient parameters are calculated based on the global environment field model and represent field distribution gradients of local environments where controlled equipment is located; And thirdly, constructing a model prediction controller integrating field coupling items, mapping field gradient parameters into field constraint penalty items of a cost function, solving an optimal control sequence by using the controller, generating a device control instruction, and inputting the device control instruction into an executing mechanism of the device to adjust the running state of the device. Preferably, the first step includes: S11, acquiring real-time running state data of controlled equipment by using an equipment body sensor and a measurement and control unit, wherein the real-time running state data comprises space position coordinates, speed vectors, attitude angles, angular rate data and internal state variable data; S12, calling a reference dynamic behavior model stored in a local edge computing unit, wherein the reference dynamic behavior model describes an ideal dynamic process and state response of the equipment under the condition of no external field interference; S13, inputting real-time running state data into a reference dynamic behavior model, calculating characteristic difference measurement between an actual state and an ideal state, extracting characteristic data only containing significant deviation information, and generating a state difference tensor. Preferably, the second step includes: S21, establishing communication connection with a cloud analysis server, and uploading a state difference tensor and current absolute space position coordinates of controlled equipment; S22, receiving field gradient parameters fed back by a cloud analysis server in response to aggregation and fusion processing of state difference tensors uploaded by the cloud analysis server on a plurality of controlled devices, wherein the field gradient parameters are gradient vectors calculated by the cloud analysis server according to an aggregate updated global environment field model and aiming at the current state of the controlled devices and field distribution in a preset prediction time domain; S23, storing the received field gradient parameters in a local buffer area for subsequent control calculation. Preferably, the third step includes: S31, initializing a model prediction controller of an integrated field coupling item, and setting a prediction time domain and a control time domain; s32, constructing a cost function, wherein the cost function comprises a state set point tracking error term, a control input change rate penalty term and a field constraint penalty term; S33, extracting field gradient parameters in the local buffer area, and substituting the field gradient parameters serving as weighting factors of field constraint penalty items into a cost function; and S34, solving an optimization problem for minimizing a cost function on the premise of meeting the constraints of equipment dynamics and pr