Cloud control platform-based control system and motion planning method for autonomous road roller

Abstract

A cloud control platform-based control system and motion planning method for an autonomous road roller. The control system comprises a road end module (1), a paver terminal module (2) and an autonomous-road-roller terminal module (3), wherein the road end module (1) comprises a first temperature sensing module (11), a signal base station (12) and a satellite (13); the paver terminal module (2) comprises a second temperature sensing module (21), a cloud sub-module (22) and a first positioning module; and the autonomous-road-roller terminal module (13) comprises a second 5G communication module (31), a computing unit and execution module (32), a vehicle-borne sensing module (33) and a second positioning module. Therefore, it can be ensured that pavement compaction work is carried out within the optimal temperature range, thereby significantly improving the compaction efficiency and road quality, and effectively avoiding uneven compaction which may be caused by manual operations.

Description

Unmanned road roller control system and motion planning method based on cloud control platform TECHNICAL FIELD

[0001] The present application relates to the technical field of pavement engineering, and in particular to an unmanned road roller control system and motion planning method based on a cloud control platform. BACKGROUND

[0002] In a construction site, good road roller motion planning not only needs to meet the requirements of road compaction degree and compaction efficiency, ensure that the road compaction is within a certain temperature threshold, but also needs to avoid vehicle collision when multiple road rollers are running at the same time. However, for a road roller driver, the perception ability of the construction site environment and the decision-making ability of road roller motion planning are limited, and he can usually only perform a relatively simple motion trajectory and uniform speed running, which is difficult to guarantee the quality of road compaction. Therefore, it is of great research significance and engineering application value to use advanced environment perception, motion planning and vehicle control means to accurately plan and control the trajectory and speed of the road roller.

[0003] In recent years, automatic driving technology has been applied to road rollers. In the single-vehicle intelligent stage, motion planning of a single vehicle is mainly based on single-vehicle perception data, and advanced vehicle control technology is used to complete the automatic driving of unmanned road rollers. However, in actual application, multiple road rollers are usually required to complete the compaction work. In the later multi-vehicle cooperative stage, the environment information and targets required for motion planning of unmanned road roller clusters are more. In order to meet the data requirements and computing power requirements, the calculation of the unmanned road roller system mostly depends on the remote cloud data processing center. However, the communication delay, packet loss and other problems increase the uncertainty of the control system, making it difficult to guarantee the real-time and accurate control requirements of the unmanned road roller. At the same time, the existing motion planning method of the unmanned road roller cluster usually adopts a rule-based method, which cannot adapt to complex and variable actual construction environments. In particular, when some road rollers lose control or human drivers join, the motion planning scheme of the unmanned road roller cannot be updated adaptively, and the adaptability of the unmanned road roller control system to dynamic environments is poor. SUMMARY

[0004] In view of the above, the present application provides an unmanned road roller control system and motion planning method based on a cloud control platform.

[0005] An unmanned road roller control system based on a cloud control platform comprises a road end module, a paver terminal module and an unmanned road roller terminal module 3.

[0006] The road end module comprises a first temperature perception module, a signal base station and a satellite.

[0007] The paver terminal module comprises a second temperature sensing module, a cloud terminal module and a first positioning module; the cloud terminal module comprises a first 5G communication module and an edge computing server; the first positioning module receives positioning information of the satellite and performs accuracy enhancement of the positioning information in combination with a signal base station;

[0008] The unmanned road roller terminal module comprises a second 5G communication module, a computing unit and an execution module, a vehicle-mounted sensing module and a second positioning module; the vehicle-mounted sensing module comprises a third temperature sensing module, a road surface compaction degree sensing module and a road roller motion state sensing module; the second positioning module receives positioning information of the satellite and performs accuracy enhancement of the positioning information in combination with a signal base station.

[0009] Further, the cloud terminal module is arranged on the paver and is used for vehicle-road information fusion and motion planning of the unmanned road roller; the first positioning module is arranged on the paver and is used for determining a position of the paver;

[0010] The second positioning module is arranged on the unmanned road roller; the computing unit and the execution module are arranged in a cab of the unmanned road roller and are mainly used for adjustment and execution of a motion planning result of the unmanned road roller.

[0011] Further, the first positioning module and the second positioning module each comprise a vehicle-mounted RTK positioning device, a vehicle-mounted GPS positioning device, an inertial navigation system and a vehicle-mounted vehicle-end wheel speed meter.

[0012] Further, in the driving process of the paver, the second temperature sensing module is used to acquire a road surface temperature behind the paver;

[0013] In the driving process of the unmanned road roller, the third temperature sensing module is used to acquire a road surface temperature in front of the unmanned road roller; the road surface compaction degree sensing module is used to acquire road surface compaction degree information at a current position of the unmanned road roller; and the road roller motion state sensing module is used to acquire a current motion state of the unmanned road roller, wherein the motion state comprises acceleration, speed and a steering angle;

[0014] The first temperature sensing module is arranged on a road side of the road surface to be compacted and is used to sense a wide-area road surface temperature to be compacted;

[0015] Meanwhile, the first positioning module on the paver and the second positioning module on the unmanned road roller are used to acquire real-time positioning to determine specific positions of the collected information.

[0016] Furthermore, by utilizing the second 5G communication module, the information of the road surface to be compacted detected by the roadside module, paver terminal module, and unmanned roller terminal module is uploaded to the cloud sub-module to complete the fusion of multi-source sensing data, forming spatiotemporal data of road surface temperature and compaction degree during the compaction process, and storing the fused spatiotemporal data in the edge computing server.

[0017] Furthermore, the edge computing server outputs the motion plan of the unmanned road roller based on the current road surface temperature and compaction information and the current motion status of the unmanned road roller. The motion plan includes trajectory and speed planning information, and then sends it to each unmanned road roller using the first 5G communication module.

[0018] Furthermore, the unmanned road roller obtains trajectory and speed planning information through the second 5G communication module. The computing unit and execution module, based on the vehicle-mounted perception module, obtain the current motion status of the unmanned road roller and surrounding road rollers, adjust the trajectory and motion planning information, and use the computing unit and execution module to calculate the steering wheel and throttle control quantities. At the same time, the computing unit and execution module are used to control the steering wheel and throttle. After the vehicle position and motion status change, the updated motion status information is sent to the cloud sub-module of the paver through the second 5G communication module.

[0019] This invention also provides a motion planning method for an unmanned road roller based on a cloud control platform, implemented using the aforementioned cloud control platform-based unmanned road roller control system. The motion planning method includes the following steps:

[0020] S1. Based on the width and length of the unmanned road roller and the width of the overlapping area, the road surface to be compacted is gridded with a rectangular grid of a certain size;

[0021] S2. Based on the gridded road surface to be compacted in step S1, construct the information interaction and data processing architecture of vehicle, road and cloud modules, including information interaction between unmanned road roller and paver, multi-source road detection data fusion and unmanned road roller motion planning implemented on the cloud sub-module of paver;

[0022] S3. The cloud submodule plans the speed of the paver's advance based on the current gridded pavement compaction and temperature information and the movement status of the unmanned road roller after integration in step S2, controls the progress of paving and compaction of the entire area, and establishes an electronic boundary for the operation of the unmanned road roller.

[0023] S4. Based on the current gridded road surface compaction and temperature information, the cloud submodule aims to maximize the working efficiency of the unmanned road roller within the electronic boundary provided in step S3, while ensuring safety and compaction.

[0024] S5. Building an unmanned road roller simulation platform in the cloud sub-module, establishing a road construction operation area simulation model, including a road surface temperature evolution model, a road roller dynamics model, and a paver dynamics model, displaying the compaction degree and temperature in a grid, simulating the road compaction process, the paver and the unmanned road roller driving conditions, adjusting and determining the unmanned road roller trajectory and speed planning model parameters in step S4 in combination with the feedback control idea;

[0025] S6. In the actual planning and control process, using the digital twin mode, outputting the planning results and the real-time measured perception data and motion data of the unmanned road roller in the simulation platform provided in step S5, completing the online update of the motion planning model parameters, improving the adaptability of the model to the real working conditions, and completing the migration of the trajectory and speed planning model from the virtual to the real environment.

[0026] Further, step S2 is as follows:

[0027] S21. Using the first temperature perception module on the roadside of the road surface to be compacted, the second temperature perception module on the paver, and the vehicle-mounted perception module on the unmanned road roller, cooperatively sensing the road surface compaction degree and temperature, and sending the detection data to the cloud sub-module arranged on the paver;

[0028] S22. The cloud sub-module completes the spatio-temporal fusion of the road information and the unmanned road roller information provided in step S21, and uses different colors to represent the compaction degree and temperature of each grid in the grid road surface;

[0029] S23. The edge computing server on the paver formulates the motion planning of the unmanned road roller according to the road surface compaction degree and temperature information provided in step S22 and the current motion state of the unmanned road roller, and sends the motion planning results to each unmanned road roller through the first 5G communication module;

[0030] S24. After the unmanned road roller completes the vehicle control according to the motion planning results provided in step S23, it sends the vehicle motion information and road detection information to the cloud sub-module in real time to integrate the vehicle-road data;

[0031] S25. As the construction continuously proceeds, the paver edge computing server continuously integrates the vehicle-road data provided in step S24, gradually forming a spatio-temporal graph of the road surface and the road roller state, which can be used for extracting the grid road surface compaction degree and temperature information and the motion state of the unmanned road roller.

[0032] Further, step S4 is as follows:

[0033] S41. Characterize the pavement compaction degree, temperature and the motion state of the road roller; the current state of the pavement compaction degree and temperature is recorded in the form of two matrices, the size of each matrix MxN corresponds to the result of the pavement gridding, where M is the number of horizontal grids and N is the number of vertical grids, each number in the matrix represents the compaction degree or temperature of the pavement in the corresponding grid; correspondingly, the position of the road roller is recorded in a matrix with the size of MxN using 01 coding, where 1 is used to represent the grid occupied by the road roller; the motion state of the road roller is represented by a vector, including speed, acceleration, steering angle, and the range of the occupied grid, and the state S is represented as:

[0034] Where T, P and L are the temperature, compaction degree and position matrix respectively; a i , v i , are the acceleration, speed and steering angle of the ith road roller respectively; are the row and column numbers of the edge of the grid range occupied by the ith road roller; and i={1, 2, 3, …, u}; u is the total number of unmanned road rollers;

[0035] S42. Design the reward function related to compaction efficiency, pavement temperature and driving safety; the compaction efficiency is represented by n p , which is the number of grids that meet the compaction degree requirement at the current compaction pass within a certain period of time, the more grids compacted within a certain period of time, the higher the efficiency; the pavement temperature is represented by t, the pavement temperature to be compacted is lower than the specified threshold or the pavement being compacted is higher than the specified threshold, which means that the current motion planning scheme needs to be further optimized; the driving safety is represented by the lateral distance d lat and the longitudinal distance d lon between road rollers, when the lateral or longitudinal distance is less than the safety threshold, it is considered that the current action selection is dangerous; the reward function r is specifically: r=w1r e +w2r t +w3r s ;

[0036] Where ρ t is the penalty term when the pavement temperature does not meet the requirement; ρ s is the penalty term when the distance between road rollers does not meet the requirement; t1 is the specified threshold for the pavement temperature to be compacted; t2 is the specified threshold for the pavement being compacted; D lat is the safety threshold for the lateral distance between road rollers; D lon is the safety threshold for the longitudinal distance between road rollers; r e is the reward function term corresponding to the compaction efficiency; r t is the reward function term corresponding to the pavement temperature; rs w1 is the weight of the reward function item corresponding to the compaction efficiency; w2 is the weight of the reward function item corresponding to the road surface temperature; w3 is the weight of the reward function item corresponding to the roller spacing;

[0037] S43. The action of designing an unmanned roller motion planning model; the decision of the unmanned roller motion planning is the lateral acceleration al at and the longitudinal motion acceleration al on , expressed as:

[0038] When the vehicle spacing is less than the specified safety threshold, emergency braking measures are taken:

[0039] wherein, is the maximum acceleration when the unmanned roller suddenly decelerates;

[0040] S44. Solve the multi-objective unmanned roller motion planning problem;

[0041] S45. Adjust the parameters of the unmanned roller motion planning model, train the model to obtain the convergence of the reward through interaction between the simulation platform and the environment, and evaluate the effect of the model by using the efficiency and safety indicators.

[0042] The beneficial effects of the present application are:

[0043] The present application, by adopting the advanced environment perception technology and complex data processing capability integrated by the cloud control platform, the system can accurately master the environmental information of the construction site in real time, including the key parameters such as road surface temperature and compaction degree, combined with advanced motion planning algorithm, can provide accurate travel trajectory and speed control instruction for each unmanned roller, ensure that the road compaction work is carried out in the best temperature range, significantly improve the compaction efficiency and road quality, and effectively avoid the uneven compaction caused by manual operation.

[0044] The present application, through the powerful computing and data processing capability of the cloud platform and the edge computing server, realizes efficient information interaction and data fusion, which not only overcomes the data island problem existing in the traditional single vehicle intelligentization scheme, but also optimizes the cooperation and scheduling strategy among multiple vehicles, effectively reduces the influence of technical obstacles such as communication delay and packet loss on work efficiency and safety, especially in complex construction environment or with human drivers participating, the system can flexibly respond and adaptively update the motion planning scheme, ensure the safety and smooth progress of construction.

[0045] The unmanned road roller can effectively respond to the dynamic changes of the construction site through the advanced motion planning method and vehicle control strategy, including avoiding collision with other road rollers or obstacles, the system considers the complexity and variability of the actual construction environment, and continuously receives updated environment and operation data from the cloud to adjust the travel route and speed of the road roller in real time to respond to unexpected situations, the high adaptability and flexibility ensure that the unmanned road roller can efficiently and safely complete the operation task under various construction conditions. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the present application or prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only a part of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0047] Fig. 1 is a schematic diagram of the control system of the unmanned road roller based on the cloud control platform according to the embodiment of the present application;

[0048] Fig. 2 is a schematic diagram of the motion planning method of the unmanned road roller based on the cloud control platform according to the embodiment of the present application;

[0049] Fig. 3 is a schematic diagram of the row number of the edge of the grid range occupied by the unmanned road roller according to the embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail below in combination with specific embodiments.

[0051] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present application should be understood as the usual meaning understood by those skilled in the art to which the present application belongs. The "first", "second" and similar words used in the present application do not represent any order, quantity or importance, but are only used to distinguish different components. "Include" or "contain" and similar words mean that the elements or objects before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connected" or "connected" and similar words are not limited to physical or mechanical connection, but can include electrical connection, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to represent relative positional relationship, when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0052] As shown in Figure 1, a cloud control platform-based unmanned road roller control system comprises a road end module 1, a paver terminal module 2 and an unmanned road roller terminal module 3;

[0053] The road end module 1 comprises a first temperature sensing module 11, a signal base station 12 and a satellite 13;

[0054] The paver terminal module 2 comprises a second temperature sensing module 21, a cloud terminal module 22 and a first positioning module; the cloud terminal module 22 comprises a first 5G communication module 221 and an edge computing server 222; the first positioning module receives positioning information of the satellite 13 and performs accuracy enhancement of the positioning information in combination with the signal base station 12;

[0055] The unmanned road roller terminal module 3 comprises a second 5G communication module 31, a computing unit and an execution module 32, a vehicle-mounted sensing module 33 and a second positioning module; the vehicle-mounted sensing module 33 comprises a third temperature sensing module, a road surface compactness sensing module and a road roller motion state sensing module; the second positioning module receives positioning information of the satellite 13 and performs accuracy enhancement of the positioning information in combination with the signal base station 12.

[0056] The cloud terminal module 22 is arranged on the paver 5 and is used for vehicle-road information fusion and motion planning of the unmanned road roller 6; the first positioning module is arranged on the paver 5 and is used for determining a position of the paver;

[0057] The second positioning module is arranged on the unmanned road roller 6; the computing unit and the execution module 32 are arranged in a cab of the unmanned road roller 6 and are mainly used for adjustment and execution of a motion planning result of the unmanned road roller 6;

[0058] The first positioning module and the second positioning module both comprise a vehicle-mounted RTK positioning device, a vehicle-mounted GPS positioning device, an inertial navigation system and a vehicle-mounted vehicle-end wheel speed meter.

[0059] The information of the road surface to be compacted 4 is acquired in a vehicle-road cooperative sensing manner, specifically as follows:

[0060] In a driving process of the paver 4, the second temperature sensing module 21 is used to acquire a road surface temperature behind the paver 4;

[0061] In a driving process of the unmanned road roller 6, the third temperature sensing module is used to acquire a road surface temperature in front of the unmanned road roller 6; the road surface compactness sensing module is used to acquire road surface compactness information at a current position of the unmanned road roller 6; the road roller motion state sensing module is used to acquire a current motion state of the unmanned road roller 6, and the motion state comprises acceleration, speed and a steering angle;

[0062] The first temperature sensing module 11 is arranged at the road side of the to-be- compacted road surface 4, and is used for sensing the temperature of the wide area of the to-be- compacted road surface, so as to make up for the sensing information of the blind area and the dead angle area which may exist in the vehicle end sensing;

[0063] Meanwhile, the first positioning module on the paver 5 and the second positioning module on the unmanned road roller 6 are used to obtain real-time positioning, so as to determine the specific position of the collected information.

[0064] The second 5G communication module 31 is used to upload the information of the to-be- compacted road surface 4 detected by the road end module 1, the paver terminal module 2 and the unmanned road roller terminal module 3 to the cloud terminal module 22, to complete the fusion of the multi-source sensing data, form the space-time data of the road surface temperature and the compactness in the compaction process, and store the fused space-time data in the edge computing server 222.

[0065] The edge computing server 222 outputs the motion planning of the unmanned road roller 6 based on the motion planning algorithm of the unmanned road roller, according to the current road surface temperature and compactness information and the current motion state of the unmanned road roller, the motion planning including trajectory and speed planning information, and then sends the motion planning to each unmanned road roller 6 by using the first 5G communication module 221.

[0066] The unmanned road roller 6 obtains the trajectory and speed planning information by using the second 5G communication module 31, the calculation unit and the execution module 32 obtain the current motion state of the unmanned road roller and the surrounding road roller based on the vehicle-mounted sensing module 33, adjust the trajectory and motion planning information, complete the calculation of the steering wheel and throttle control amount by using the calculation unit and the execution module 32, and complete the control of the steering wheel and the throttle by using the calculation unit and the execution module 32. After the position and motion state of the vehicle change, the second 5G communication module 31 is used to send the updated motion state information to the cloud terminal module 22 of the paver.

[0067] As shown in FIG. 2, a motion planning method of an unmanned road roller based on a cloud control platform is realized based on the unmanned road roller control system of the cloud control platform, and includes the following steps:

[0068] S1. According to the width and length of the unmanned road roller 6 and the width of the overlapping area, the to-be- compacted road surface 4 is rasterized into a rectangular grid with a certain size;

[0069] S2. Based on the rasterized to-be- compacted road surface 4 in step S1, an information interaction and data processing architecture of the vehicle, the road and the cloud module is constructed, including the information interaction between the unmanned road roller 6 and the paver 5, the multi-source road detection data fusion and the motion planning of the unmanned road roller realized on the cloud terminal module 22 of the paver 5. The specific steps can be subdivided as follows:

[0070] S21. Utilize the first temperature sensing module 11 on the side of the road to be compacted 4, the second temperature sensing module 21 on the paver 5, and the on-board sensing module 33 on the unmanned road roller 6 to collaboratively perceive the road surface compactness and temperature, and send the detection data to the cloud terminal module 22 provided on the paver 5;

[0071] S22. The cloud terminal module 22 completes the spatio-temporal fusion of the road information and unmanned road roller information provided in step S21, and uses different colors to represent the compactness and temperature of each grid in the grid road surface;

[0072] S23. The edge computing server 222 on the paver 5 formulates a motion plan for the unmanned road roller based on the road surface compactness and temperature information provided in step S22 and the current motion state of the unmanned road roller, and sends the motion planning result to each unmanned road roller 6 through the first 5G communication module 221;

[0073] S24. After the unmanned road roller 6 completes vehicle control based on the motion planning result provided in step S23, it sends the vehicle motion information and road surface detection information to the cloud terminal module 22 in real time to integrate the vehicle-road data;

[0074] S25. As the construction continues, the paver edge computing server 222 continuously integrates the vehicle-road data provided in step S24, gradually forming a spatio-temporal graph of the road surface and the road roller state, which can be used for extracting the grid road surface compactness and temperature information and the motion state of the unmanned road roller.

[0075] S3. The cloud terminal module 22 plans the speed of the paver 5 based on the current grid road surface compactness and temperature information and the motion state of the unmanned road roller integrated in step S2, controls the progress of the overall area paving and compaction, and establishes an electronic fence for the operation of the unmanned road roller;

[0076] S4. The cloud terminal module 22 establishes a speed and trajectory collaborative planning model for the unmanned road roller within the electronic fence provided in step S3 based on the current grid road surface compactness and temperature information, with the goal of maximizing the work efficiency of the unmanned road roller under the premise of ensuring safety and compactness. The trajectory and speed planning model can be but not limited to a model combining deep reinforcement and rule-based motion planning. The specific steps can be further divided into:

[0077] S41. Characterize the pavement compaction degree, temperature and the motion state of the road roller. The current state of the pavement compaction degree and temperature is recorded in the form of two matrices, each matrix has a size of MxN corresponding to the result of the pavement gridding, where M is the number of horizontal grids and N is the number of vertical grids, each number in the matrix represents the compaction degree or temperature of the pavement in the corresponding grid; accordingly, the position of the road roller is recorded in a matrix with a size of MxN using 01 coding, where 1 represents the grid occupied by the road roller; the motion state of the road roller is represented by a vector, including speed, acceleration, steering angle, and the range of the occupied grid (represented by the row and column numbers): in summary, the state S is represented as:

[0078] where T, P, L are the temperature, compaction degree and position matrix respectively; a i , v i , are the acceleration, speed and steering angle of the ith road roller respectively; and are the row and column numbers of the edge of the grid range occupied by the ith road roller; and i={1, 2, 3, …, u}; u is the total number of unmanned road rollers.

[0079] As shown in FIG. 3, the first unmanned road roller is located on the grid of the pavement to be compacted 4, and the row and column numbers of the edge of the grid range are specifically:

[0080] S42. Design a reward function related to compaction efficiency, pavement temperature and driving safety. The compaction efficiency is represented by n p , which is the number of grids that meet the compaction degree requirement of the current compaction pass within a certain period of time, the more grids compacted within a certain period of time, the higher the efficiency; the pavement temperature is represented by t, if the temperature of the pavement to be compacted is lower than the specified threshold or the temperature of the pavement being compacted is higher than the specified threshold, it means that the current motion planning scheme needs to be further optimized; the driving safety is represented by the lateral distance d lat and longitudinal distance d lon between road rollers, when the lateral or longitudinal distance is less than the safety threshold, it is considered that the current action selection is dangerous; the design of the reward function considers the compaction efficiency, pavement temperature and safety guarantee ability comprehensively, and the reward function r is specifically: r=w1r e +w2r t +w3r s ;

[0081] where ρ t is the penalty term when the pavement temperature does not meet the requirement; ρ sis a penalty term for the distance between the road rollers not meeting the requirements; t1 is a threshold value of the temperature of the road surface to be compacted; t2 is a threshold value of the road surface being compacted; D lat is a safety threshold value of the transverse distance between the road rollers; D lon is a safety threshold value of the longitudinal distance between the road rollers; r e is a reward function subterm corresponding to the compaction efficiency; r t is a reward function subterm corresponding to the temperature of the road surface; r s is a reward function subterm corresponding to the distance between the road rollers; w1 is the weight of the reward function subterm corresponding to the compaction efficiency; w2 is the weight of the reward function subterm corresponding to the temperature of the road surface; w3 is the weight of the reward function subterm corresponding to the distance between the road rollers;

[0082] S43. Action of designing a motion planning model of an unmanned road roller. The decision-making action of the motion planning of the unmanned road roller is the transverse acceleration al at and the longitudinal motion acceleration al on , expressed as:

[0083] When the distance between the vehicles is less than the specified safety threshold value, an emergency braking measure is adopted:

[0084] wherein, is the maximum acceleration when the unmanned road roller is rapidly decelerated.

[0085] S44. Solving a multi-objective motion planning problem of an unmanned road roller.

[0086] The multi-objective motion planning problem of the unmanned road roller is solved by using a deep reinforcement learning algorithm. In this embodiment, the deep reinforcement learning algorithm is taken as an example, and the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm can be used, but is not limited to the MADDPG algorithm. The forms of the Critic and Actor networks can adopt convolutional neural networks and fully connected networks, and the convolutional neural networks used for understanding the state can be shared.

[0087] S45. Adjusting the parameters of the motion planning model of the unmanned road roller. Through interaction between the simulation platform and the environment, the reward obtained by the model is trained to converge, and the model effect is evaluated by using the efficiency, safety and other indicators. If the related indicators meet the requirements, it can be considered that the MADDPG strategy network obtained is feasible.

[0088] S5. In the cloud sub-module 22, an unmanned road roller simulation platform is built, and a road construction operation area simulation model is established, including but not limited to a road surface temperature evolution model, a road roller dynamics model, and a paver dynamics model, a grid display compaction degree and temperature, a simulation of a road surface compaction process, a paver and an unmanned road roller driving working condition, a combination of a feedback control idea, adjustment and determination of the unmanned road roller trajectory and speed planning model parameter in step S4.

[0089] S6. In the actual planning and control process, a digital twin mode is adopted, a planning result is output by using a motion planning model in the simulation platform provided in step S5, and measured perception data and motion data returned by the unmanned road roller are fed back, online update of the motion planning model parameter is completed, adaptability of the model to a real working condition is improved, and migration of the trajectory and speed planning model from a virtual environment to a real environment is completed. The migration mode adopted can include but is not limited to model parameter reuse and fine-tuning, and reuse of training samples.

[0090] Another aspect of the present application provides a computer readable storage medium, which stores a computer program, and a vehicle-mounted perception module program is executed by a processor to implement the steps of the unmanned road roller motion planning method based on the cloud control platform.

[0091] Another aspect of the present application provides an apparatus, which includes a processor and a memory, the vehicle-mounted perception module memory is used to store a computer program, and the vehicle-mounted perception module processor is used to execute the computer program stored in the vehicle-mounted perception module memory, so that the vehicle-mounted perception module apparatus executes the steps of the unmanned road roller motion planning method based on the cloud control platform.

[0092] Those skilled in the art should understand that the above discussion of any embodiment is only exemplary, and is not intended to imply that the scope of the present application is limited to these examples; under the idea of the present application, the above embodiments or technical features in different embodiments can also be combined, the steps can be implemented in any order, and there are many other changes of different aspects of the present application as described above. In order to be brief, they are not provided in details.

[0093] The present application is intended to cover all such alternatives, modifications and variations as fall within the broad scope of the claims. Therefore, any omission, modification, equivalent replacement, improvement, etc. made in the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A cloud control platform-based motion planning method for an unmanned road roller, characterized in that, Comprise the following steps: S1. According to the width and length of the unmanned road roller (6), the width of the overlapping area, the to-be-compacted road surface (4) is rasterized into a rectangular grid of a certain size; S2. Based on the rasterized to-be-compacted road surface (4) in step S1, an information interaction and data processing architecture of vehicle, road, and cloud modules is constructed, including information interaction between the unmanned road roller (6) and the paver (5), multi-source road surface detection data fusion implemented on the cloud terminal module (22) of the paver (5), and unmanned road roller motion planning; S3. The cloud terminal module (22) plans the speed of the paver (5) advancing, controls the progress of overall area paving and compaction, and establishes an electronic fence for the operation of the unmanned road roller according to the current rasterized road surface compaction degree and temperature information integrated in step S2 and the motion state of the unmanned road roller; S4. The cloud terminal module (22) establishes a speed and trajectory coordination planning model of the unmanned road roller within the scope of the electronic fence provided in step S3, with the goal of maximizing the work efficiency of the unmanned road roller under the premise of ensuring safety and compaction degree, based on the current rasterized road surface compaction degree and temperature information; step S4 is specifically as follows: S41. Characterize the compaction degree of the road surface, temperature and the motion state of the road roller, the current state of the compaction degree of the road surface and temperature is recorded in the form of two matrices, the size of each matrix MxN corresponds to the result of the road surface gridding, where M is the number of horizontal grids, N is the number of vertical grids, each number in the matrix represents the compaction degree or temperature of the road surface in the corresponding grid; accordingly, the position of the road roller is recorded in a matrix of size MxN using 01 coding, where 1 represents the grid occupied by the road roller; the motion state of the road roller is represented by a vector, including speed, acceleration, steering angle, and the range of occupied grids, the state S is represented as: Wherein, T, P, L are temperature, compaction degree, position matrix respectively; a i , v i , respectively the acceleration, the speed, the steering angle of the i-th road roller; The serial numbers of the rows and columns where the edges of the grid range occupied by the i-th road roller are located; and i={1, 2, 3, …, u}; u is the total number of unmanned road rollers; S42. Design a reward function related to compaction efficiency, road surface temperature, and driving safety, where the compaction efficiency is represented by the number of grids n that reach the compaction degree requirement at the current compaction pass within a certain time period p The more grids compacted within a certain time period, the higher the efficiency. The road surface temperature is represented by t, and if the temperature of the road surface to be compacted is lower than a specified threshold or the temperature of the road surface being compacted is higher than a specified threshold, it means that the current motion planning scheme needs further optimization. The driving safety is represented by the lateral distance d lat and the longitudinal distance d lon between the road rollers, and when the lateral or longitudinal distance is less than a safety threshold, it is considered that the current action selection is dangerous. The reward function r is specifically: r = w1r e +w2r t +w3r s ; wherein, ρ t is a penalty term when the road surface temperature does not meet the requirement; ρ s is a penalty term when the distance between the road rollers does not meet the requirement; t1 is a prescribed threshold value of the road surface temperature to be compacted; t2 is a prescribed threshold value of the road surface being compacted; D lat is a safety threshold value of the lateral distance between the road rollers; D lon is a safety threshold value of the longitudinal distance between the road rollers; r e is a reward function subterm corresponding to the compaction efficiency; r t is a reward function subterm corresponding to the road surface temperature; r s is a reward function subterm corresponding to the distance between the road rollers; w1 is the weight of the reward function subterm corresponding to the compaction efficiency; w2 is the weight of the reward function subterm corresponding to the road surface temperature; and w3 is the weight of the reward function subterm corresponding to the distance between the road rollers. S43. An action of designing a motion planning model of the unmanned road roller; the decision action of the motion planning of the unmanned road roller is lateral acceleration al at and longitudinal motion acceleration al on , expressed as: Emergency braking measures are taken when the vehicle distance is less than a prescribed safety threshold: wherein The maximum acceleration when the unmanned road roller suddenly decelerates; S44. Solve the multi-objective unmanned road roller motion planning problem; S45. Adjust the motion planning model parameters of the unmanned road roller, train the model to obtain reward convergence through interaction between the simulation platform and the environment, and evaluate the model effect using efficiency and safety indicators; S5. Build an unmanned road roller simulation platform on the cloud terminal module (22), establish a road construction work area simulation model, including a road surface temperature evolution model, a road roller dynamics model, and a paver dynamics model, display the compaction degree and temperature in a grid, simulate the road surface compaction process, the driving conditions of the paver and the unmanned road roller, adjust and determine the trajectory and speed planning model parameters of the unmanned road roller in step S4 based on the feedback control idea; S6. In the actual planning and control process, use the digital twin mode to output the planning results of the motion planning model and the real-time perception data and motion data returned by the unmanned road roller in the simulation platform provided in step S5, complete the online update of the motion planning model parameters, improve the adaptability of the model to real working conditions, and complete the migration of the trajectory and speed planning model from the virtual to the real environment.

2. The cloud control platform based motion planning method for unmanned road roller according to claim 1, wherein, Step S2 is specifically as follows: S21. Use the first temperature perception module (11) on the roadside of the to-be-compacted road surface (4), the second temperature perception module (21) on the paver (5), and the vehicle-mounted perception module (33) on the multiple unmanned road rollers (6) to cooperatively perceive the road surface compaction degree and temperature, and send the detection data to the cloud terminal module (22) arranged on the paver (5); S22. The cloud sub-module (22) completes the spatio-temporal fusion of the road information and the unmanned road roller information provided in step S21, and uses different colors to represent the compaction degree and temperature of each grid in the grid road surface; S23. The edge computing server (222) on the paver (5) formulates the motion planning of the unmanned road roller according to the road surface compaction degree and temperature information provided in step S22 and the current motion state of the unmanned road roller, and sends the motion planning result to each unmanned road roller (6) through the first 5G communication module (221); S24. After the unmanned road roller (6) completes the vehicle control according to the motion planning result provided in step S23, the vehicle motion information and the road surface detection information are sent to the cloud sub-module (22) in real time to integrate the vehicle-road data; S25. As the construction continuously proceeds, the paver edge computing server (222) continuously integrates the vehicle-road data provided in step S24, and gradually forms the spatio-temporal graph of the road surface and the road roller state, which can be used for extracting the grid road surface compaction degree and temperature information and the motion state of the unmanned road roller.

3. A cloud platform based unmanned road roller control system for performing the method of motion planning of the unmanned road roller as claimed in any one of claims 1-2, wherein, The system comprises a road end module (1), a paver terminal module (2), and an unmanned road roller terminal module (3); The road end module (1) comprises a first temperature sensing module (11), a signal base station (12), and a satellite (13); The paver terminal module (2) comprises a second temperature sensing module (21), a cloud sub-module (22), and a first positioning module; the cloud sub-module (22) comprises a first 5G communication module (221) and an edge computing server (222); the first positioning module receives the positioning information of the satellite (13) and enhances the accuracy of the positioning information in combination with the signal base station (12); The unmanned road roller terminal module (3) comprises a second 5G communication module (31), a computing unit and an execution module (32), a vehicle-mounted sensing module (33), and a second positioning module; the vehicle-mounted sensing module (33) comprises a third temperature sensing module, a road surface compaction degree sensing module, and a road roller motion state sensing module; the second positioning module receives the positioning information of the satellite (13) and enhances the accuracy of the positioning information in combination with the signal base station (12).

4. The cloud-based platform controlled unmanned road roller control system according to claim 3, wherein, The cloud sub-module (22) is arranged on the paver (5) and is used for vehicle-road information fusion and motion planning of the unmanned road roller (6); the first positioning module is arranged on the paver (5) and is used for determining the position of the paver; The second positioning module is arranged on the unmanned road roller (6); the computing unit and the execution module (32) are arranged in the cab of the unmanned road roller (6), and are mainly used for adjusting and executing the motion planning result of the unmanned road roller (6).

5. The cloud-based platform controlled unmanned road roller control system according to claim 4, wherein, The first positioning module and the second positioning module each comprise a vehicle-mounted RTK positioning device, a vehicle-mounted GPS positioning device, an inertial navigation system, and a vehicle-mounted vehicle-end wheel speed sensor.

6. The cloud-based platform controlled unmanned road roller control system according to claim 5, wherein, In the driving process of the paver (4), the second temperature sensing module (21) is used to obtain the road surface temperature behind the paver (4). The unmanned road roller (6) obtains the road surface temperature in front of the unmanned road roller (6) by using the third temperature sensing module during driving; The road surface compaction degree sensing module is used to obtain the road surface compaction degree information of the current position of the unmanned road roller (6), and the road roller motion state sensing module is used to obtain the current motion state of the unmanned road roller (6), including acceleration, speed and steering angle; The first temperature sensing module (11) is arranged on the road side of the road surface to be compacted (4) and is used to sense the temperature of the wide-area road surface to be compacted; At the same time, the first positioning module on the paver (5) and the second positioning module on the unmanned road roller (6) are used to obtain real-time positioning to determine the specific position of the collected information.

7. The cloud-based platform controlled unmanned road roller control system according to claim 6, wherein, The information of the road surface to be compacted (4) detected by the road end module (1), the paver terminal module (2) and the unmanned road roller terminal module (3) is uploaded to the cloud terminal module (22) by using the second 5G communication module (31), multi-source sensing data fusion is completed, the space-time data of the road surface temperature and the compaction degree in the compaction process are formed, and the fused space-time data is stored in the edge computing server (222).

8. The cloud-based platform controlled unmanned road roller control system according to claim 7, wherein, The edge computing server (222) outputs the motion planning of the unmanned road roller (6) according to the current road surface temperature and compaction degree information and the current motion state of the unmanned road roller, the motion planning includes trajectory and speed planning information, and then the first 5G communication module (221) is used to send the motion planning to each unmanned road roller (6).

9. The cloud-based platform controlled unmanned road roller control system according to claim 8, wherein, The unmanned road roller (6) obtains the trajectory and speed planning information by using the second 5G communication module (31), the calculation unit and the execution module (32) obtain the current motion state of the unmanned road roller and the surrounding road roller based on the vehicle-mounted sensing module (33), adjust the trajectory and motion planning information, calculate the steering wheel and throttle control amount by using the calculation unit and the execution module (32), control the steering wheel and the throttle by using the calculation unit and the execution module (32), and send the updated motion state information to the cloud terminal module (22) of the paver by using the second 5G communication module (31) after the vehicle position and the motion state are changed.

Images