Mechanical-Data Dual-Driven Unmanned Road Fleet Cooperative Operation Method and System

Abstract

The present invention relates to a method and system for unmanned fleet collaborative operation on roads driven by both machinery and data, which establishes a path planning combining dynamic and static methods. According to different construction scenarios, the switching between dynamic and static path planning is realized. For the soil subgrade, static path planning is adopted for operation, and for the asphalt surface course, dynamic path planning based on the position interaction between the paver and the roller fleet is used. The dynamic path planning ensures the synchronization of construction, solves the problem that the filler cools too quickly to be rolled in time under winter construction conditions, and realizes "rolling immediately after paving" and synchronous operation. The compaction quality after the compaction operation of the roller fleet is analyzed and evaluated, and precise compaction quality optimization is carried out for the areas where the compaction quality does not meet the standard, avoiding the disadvantages such as easy omission of rolling, under-rolling, and over-rolling in the previous roller operation, and realizing automatic "quality control" of the compaction operation.

Description

Technical Field

[0001] The present invention relates to the technical field of driverless and intelligent compaction control of a fleet of rollers. Specifically, it is a mechanical-data dual-driven method and system for unmanned fleet collaborative operation of roads, which can achieve the full coverage compaction of the compaction area and the purpose of optimizing the compaction quality. Background Art

[0002] In recent years, the scale of expressways in China has developed rapidly. The rapid development of its scale is inseparable from the guarantee of compaction machinery. As one of the typical types of pavement machinery, rollers are widely used in the subgrade and pavement compaction operations of expressways. In the compaction operation, the compaction quality of the subgrade and pavement determines the quality and service life of the expressway. However, the traditional manual operation of rollers for compaction mainly relies on the work experience of drivers. The compaction quality of the subgrade and pavement is affected by the driver's experience, and problems such as over-compaction, missed compaction, and under-compaction are likely to occur, making it difficult to control the overall compaction quality. Moreover, if there is missed compaction or the compaction quality does not meet the standard during the compaction process, it is time-consuming and laborious to re-roll. With the rapid development of technology, technologies such as artificial intelligence and driverless have begun to be widely used in the construction field. Currently, driverless rollers have begun to be used in the subgrade and pavement compaction of expressways, and through various sensors installed on the roller body, such as GPS, millimeter-wave radar, cameras, etc., path planning, tracking control, and obstacle avoidance functions are realized.

[0003] Currently, many construction machinery manufacturers have combined rollers with positioning technology, sensing technology, and computer technology to achieve driverless operation of single rollers or a fleet of rollers, which has improved the construction quality and construction efficiency to a certain extent. However, the above single rollers or a fleet of rollers have the following deficiencies: (1) The applicable scenarios are few, mostly applicable to the compaction operation of soil subgrades, not applicable to asphalt surface layer operations. The asphalt surface layer operation is greatly affected by temperature, and paving and rolling need to be carried out simultaneously. The existing rollers cannot interact with the paver to achieve simultaneous paving and compaction. (2) The operation of rollers mostly constructs through a pre-determined path, and the compaction quality detection often uses the method of post-event sampling inspection, which is not only time-consuming and laborious but also cannot obtain the compaction quality of the entire subgrade and pavement, and cannot perform real-time and comprehensive detection of the compaction quality of the subgrade and pavement during the operation of the roller, making it difficult to optimize the compaction quality.

[0004] Therefore, the present invention proposes a mechanical-data dual-driven road unmanned machine group collaborative operation method and system. The method and system can switch between static path planning and dynamic path planning in the face of different construction scenarios. During the operation, rollers and rollers, as well as rollers and pavers, interact and work collaboratively to meet actual engineering needs. During the operation, the intelligent compaction equipment and system installed on the machine group can obtain compaction data in real time, realize the whole process monitoring of compaction quality, and achieve precise compaction quality optimization in areas that do not meet quality standards afterwards. Summary of the invention

[0005] In view of the deficiencies in the prior art, the technical problem that the present invention intends to solve is: to propose a mechanical-data dual-driven unmanned road machine group collaborative operation method and system with more applicable scenarios, interactive and collaborative operation between construction machines, full-process monitoring of operation data, higher operation accuracy, and less time consumption, so as to realize multi-machine collaborative construction operation of unmanned road rollers in multiple construction scenarios, and to monitor the overall compaction quality and subsequently optimize the precise compaction quality of areas that do not meet the quality standards, thereby improving construction efficiency and compaction quality.

[0006] The technical solution adopted by the present invention to solve the technical problem is:

[0007] In a first aspect, the present invention provides a mechanical-data dual-driven road unmanned machine group collaborative operation method, the operation method comprising the following contents:

[0008] The model of the roller group is selected according to the characteristics of the compacted material and the boundary information of the compacted area, the plane rectangular coordinates of the center line of the compacted area are obtained by using the longitude and latitude grid distance calculation method, and the compacted area map is generated according to the plane rectangular coordinates of the center line of the compacted area and the road width;

[0009] According to different construction scenarios, a path planning combining dynamic and static methods is established to realize the switching between dynamic and static path planning. Specifically: for subgrade construction, static path planning is used. The method of static path planning is as follows: several equal-length working sections are pre-divided for the compaction area, and the paths of all working sections in the compaction area are automatically generated when generating the working path. The machine fleet sequentially compacts all working sections; for asphalt surface layer construction, dynamic path planning is used, and the path of each working section is dynamically generated during work. The process of dynamic path planning is as follows: the machine fleet does not have the end point information of the first working section. Taking the starting point of the paver as the starting point for compaction of the first working section, when starting work, the machine fleet follows the paver for compaction. Taking the parking position of the paver each time and reserving a certain safety distance as the compaction end point of the current working section of the machine fleet, the path of the machine fleet for the current working section is generated based on the compaction starting point and compaction end point of the current working section and sent to the machine fleet. The machine fleet reciprocally compacts in the current working section according to this path. After all compactions are completed, the paver continues to move forward. The machine fleet then sets a working section overlap distance from the compaction end point of the current working section to the already-compacted part to determine the compaction starting point of the next working section. The paver stops again, and taking the parking position of the paver and reserving a certain safety distance as the compaction end point of the next working section of the machine fleet, the path of the machine fleet for the next working section is generated based on the compaction starting point and compaction end point of the next working section. The machine fleet reciprocally compacts in the next working section according to this path; the paver continues to move forward, and so on until the entire compaction area is compacted;

[0010] The path planning algorithm is executed in both static path planning and dynamic path planning;

[0011] During the compaction process, the compaction degree of the whole process is obtained in real time. After the compaction operation of the machine fleet is completed, the global compaction cloud map is displayed. According to this compaction cloud map, the quality of the non-compliant compaction area is optimized in a precise area;

[0012] The compaction cloud map is divided into three categories: uncompacted area, area that has been compacted but with unqualified quality, and fully compacted area according to the compaction quality, and they are respectively marked. Then, the uncompacted area data set and the data set of areas with unqualified compaction quality are respectively established. The uncompacted area target detection model is trained using the uncompacted area data set, and the unqualified area target detection model is trained using the data set of areas with unqualified compaction quality;

[0013] During construction, the leakage area is identified from the compaction cloud map using the leakage area target detection model, and the coordinate information of the four corner points of the boxed position of the leakage area is obtained. The path planning algorithm is used to perform supplementary compaction on the leakage areas of the soil subgrade construction and the asphalt surface course construction respectively. Then, the unqualified area target detection model is used to identify the areas that have been compacted but do not meet the quality standards from the compaction cloud map after supplementary compaction, and the coordinate information of the four corner points of the boxed position of the areas that have been compacted but do not meet the quality standards is obtained. The recompaction area is determined for recompaction, and a recompaction distance threshold is set. During the recompaction process, when passing through the fully compacted area, the roller is set to the static pressure mode. At the same time, the distance from the roller to the area to be recompacted is calculated in real time. When the distance is within the set recompaction distance threshold range, the roller automatically turns on the vibration mode for recompaction. After the roller exits the recompaction area, the vibration mode is turned off, and the recompaction ends. The compaction quality is checked again, and the unqualified areas are recompacted twice until the global compaction quality meets the standards, and then the compaction operation ends.

[0014] Further, the calculation method of the graticule distance is as follows: There are multiple positioning points on the surface of the ellipsoid. The longitude and latitude coordinates of the i-th positioning point are denoted as ( , ). Taking ( , ) as the longitude and latitude coordinates of the reference point, the relative coordinate position between the positioning point and the reference point is calculated. The calculation formula is:

[0015]

[0016]

[0017] where L e is the difference in the longitude direction, and B e is the difference in the latitude direction;

[0018] By multiplying the difference in the longitude direction and the difference in the latitude direction by the conversion coefficients for converting to the plane rectangular coordinates respectively, the points on the plane relative rectangular coordinate system of the compaction area are obtained. The calculation formula is:

[0019]

[0020]

[0021] where k L is the conversion coefficient in the longitude direction, k B is the conversion coefficient in the latitude direction, x is the x coordinate of the plane relative rectangular coordinate system, and y is the y coordinate of the plane relative rectangular coordinate system.

[0022] Further, the process of generating the compaction area map based on the plane rectangular coordinates of the center line of the compaction area and the road width is as follows: During on-site construction, with the north as the positive direction of the Y-axis, the east as the positive direction of the X-axis, and the origin set at the southwest side of the construction area, a plane relative rectangular coordinate system is constructed;

[0023] Set the road width as W, and the coordinates of the center point on the center line of the compaction area as (x j , y j ), where j = 1, 2...m, and m is the number of points on the center line of the compaction area; the coordinates of the four corner points of the compaction area are (x oe , y oe ), where e = 1, 2, 3, 4. Define the acute angle between the center line of the compaction area and the X-axis in the plane relative rectangular coordinate system as the azimuth angle α of the center line of the compaction area, and define the azimuth angle formed with the positive direction of the X-axis as positive and the azimuth angle formed with the negative direction of the X-axis as negative;

[0024] When 0 < α < 90°, the point closest to the Y-axis among the four corner points of the compaction area is denoted as o1, and the other corner points are determined in a clockwise order as o2, o3, o4; when -90° < α < 0, the point closest to the X-axis among the four corner points of the compaction area is denoted as o1, and the other corner points are determined in a clockwise order as o2, o3, o4;

[0025] According to the coordinates of the starting point and the ending point on the center line of the compaction area, calculate the azimuth angle α of the center line of the compaction area according to the following formula,

[0026]

[0027] Then calculate the coordinates of the four corner points of the compaction area according to the following formula:

[0028]

[0029]

[0030]

[0031]

[0032] Among them, α′ is the azimuth angle of the boundary o1o4 or o2o3 of the compaction area. When 0 < α < 90°, α′ = α - 90°; when -90° < α < 0, α′ = α + 90°;

[0033] Thus, the coordinates of the four corner points of the entire compaction area are obtained, and the rectangular area formed by connecting the four corner points is the entire compaction area.

[0034] Furthermore, the path planning algorithm includes rolling area path planning and lane-changing area path planning. According to the number of selected rollers and the ratio of wheel widths, operation zones are divided for each roller in the compaction area, and the total wheel width of the roller fleet is b i , where i = 1, 2...n, n represents the number of rollers, and the corresponding operation zones S responsible for each roller are generated i ; among which, the process of rolling area path planning is as follows:

[0035] For the operation zone S 1 , let the width of the operation zone S 1 be W 1 , the wheel width of the roller responsible for the operation zone S 1 be b 1 , and the overlapping distance between adjacent rolling lanes be L. Then the actual effective rolling width of one rolling lane is D 1 = b 1 - L. Then the number of rolling lanes N in this operation zone = roundup(W 1 / D 1 ), where roundup() is the ceiling function;

[0036] According to the coordinates of the four corner points of the compaction area, the starting and ending coordinate information of the midline of each rolling lane is calculated using the following formula

[0037] The starting coordinate (S 1 x 1 , S k1 y 1 ) of the midline of each rolling lane in the operation zone S k1 is:

[0038]

[0039]

[0040] The ending coordinate (S 1 x 1 , S kE y 1 ) of the midline of each rolling lane in the operation zone S kE is:

[0041]

[0042]

[0043] where k represents the k-th rolling pass, k = 1, 2,..., N;

[0044] When the subsequent rollers perform rolling, they operate along the midline of each rolling lane, and the rollers reciprocate between each rolling lane;

[0045] The process of path planning for the lane-changing area is as follows:

[0046] When the roller completes the preset number of rolling passes on the current rolling lane, a lane-changing operation is performed. After lane-changing, the roller will immediately carry out a new compaction operation. After lane-changing, the roller will straighten its body and the articulated steering angle will be zero.

[0047] A cubic polynomial curve is used for lane-changing path planning: For the first operation zone and the roller responsible for the operation of the first operation zone, assuming the wheel width of the roller is b 1 , the overlapping distance between adjacent rolling lanes is L, then the lateral lane-changing distance D * 1 = b 1 - L, the longitudinal lane-changing distance is s, and the lane-changing end point is used as the local sampling point.

[0048] A relative coordinate system for the lane-changing path is established with the driving direction of the roller after lane-changing as the x-axis direction. Then the starting point coordinates are (0, D * 1 ), the lane-changing end point coordinates are (s, 0), the lane-changing duration is t, and the roller's body is straightened when reaching the lane-changing end point.

[0049] Let the lane-changing curve equation be:

[0050]

[0051] Substitute the coordinate constraints (0, D * 1 ), (s, 0) and the curvature constraints ρ(0) = 0, ρ(t) = 0, and obtain the values of the parameters a, b, c, d in the lane-changing curve equation, and then obtain the lane-changing curve equation of the roller and the coordinate information of the points on the curve.

[0052] Perform coordinate transformation on the obtained lane-changing curve equation and the coordinate information of the points on the curve to the plane relative rectangular coordinate system of the compaction area.

[0053] Furthermore, in the actual operation of the roller, the body azimuth angle is θ 2 obtained by the GNSS device installed on the vehicle body, and the relative rotation angle φ of the roller body is obtained by the angle sensor installed at the articulated joint of the roller. Then the azimuth angle θ of the front wheel of the roller 1 is: θ 1 = θ 2 + φ;

[0054] During the operation of the machine fleet, path tracking control of the roller is achieved through the path tracking control algorithm and the sensing devices installed on the roller body.

[0055] The coordinates of the roller are (x B, y B ), the target point coordinates are (x A , y A ), and the angle between the line connecting the roller and the target point and the Y-axis is:

[0056]

[0057] According to calculate the expected angle β of the current front wheel. It is stipulated that if β is negative, turn the steering wheel to the left, and if β is positive, turn the steering wheel to the right. The specific β is calculated by the following formula:

[0058] 1) 0 < θ 1 < 180°

[0059] ① If y A < y B , then β = -θ 1 + 180°,

[0060] ② If y A > y B , then β = -θ 1 ;

[0061] 2) 180° < θ 1 < 360°

[0062] ① If y A < y B , then β = -θ 1 + 180°,

[0063] ② If y A > y B , then β = -θ 1 + 360°;

[0064] When the roller is operating, it calculates the angle deviation between the current position and the target position in real time, and drives the electric steering wheel of the roller to correct the angle deviation to achieve the path tracking control of the roller.

[0065] Furthermore, when the fleet is operating, a dynamic obstacle avoidance strategy is adopted for obstacle avoidance. The dynamic obstacle avoidance strategy is: dynamically adjust the detection range L0 of the front and rear millimeter-wave radars of the roller according to the speed of the roller. The detection range meets the requirement of L0 = 2·v, where the unit of L0 is m, and v is the real-time speed of the roller, with the unit of m / s;

[0066] The heading obstacle detection logic is: when a heading obstacle is detected during the operation of the fleet, stop the operation and give a warning, and continue to work after the heading obstacle has moved away;

[0067] The lateral obstacle detection logic is as follows: when the roller detects a lateral obstacle, it stops and gives a warning, and continues to work after the lateral obstacle is removed; if a roller vehicle in other areas except the current rolling lane is detected, both roller vehicles stop, quickly judge the lateral distance, relative position information between the two roller vehicles, and the deviation amount from the preset route, and then the roller vehicle in the front adjusts the route first and continues to operate. After traveling a certain distance, the roller vehicle in the rear adjusts the route.

[0068] During the compaction operation of the machine fleet, the monocular cameras on both sides of the rollers in the compaction area collect the curb picture information in real time to obtain the lateral distance between the side of the roller body and the curb. When the lateral distance from the curb is less than 15 cm, the protection mechanism is triggered, the roller stops, and then it is adjusted according to the error between the position of the roller and the planned route, so that the roller can work again.

[0069] Furthermore, intelligent compaction equipment is installed on each roller to monitor the compaction degree in real time; the parameters of the roller are dynamically optimized according to the change of the compaction degree, and a fitness function of the compaction degree, the speed and the vibration frequency of the roller is established, and the value ranges of the speed and the vibration frequency are set as constraints: the speed range of the roller is set to 0.5 - 8 km / h, and the accuracy is 0.1 km / h; the vibration mode is divided into two modes: large vibration and small vibration.

[0070] Set the maximum number of iterations, and through the iterative process of the intelligent algorithm, obtain the optimal speed and the optimal vibration frequency, and control the roller to perform the compaction operation at the optimal speed and the optimal vibration frequency.

[0071] In the second aspect, the present invention provides a mechanical-data dual-driven unmanned machine fleet collaborative operation system for roads, including a roller machine fleet, a perception subsystem, a decision-making subsystem, a control subsystem, a transmission subsystem, and a management subsystem. The perception subsystem and the decision-making subsystem are both installed on the roller, and the perception subsystem is electrically connected to the decision-making subsystem; the control subsystem is used to control the gear shifting, steering, throttle supply, and working mode of the roller, is connected to the decision-making subsystem, and is directly controlled by the decision-making subsystem; the transmission subsystem is used to transmit data and instructions; the management subsystem receives the operation data of the roller through the transmission subsystem, sends instructions to the roller machine fleet, and obtains the operation data of the paver.

[0072] The roller machine fleet includes several single-drum rollers and / or double-drum rollers of different models and different weights. The roller includes a front frame and a rear frame, and the front frame and the rear frame are connected by a hinge shaft. The cab is located on the rear frame, and the compaction drum is located on the front frame.

[0073] The perception subsystem includes: a GNSS rover for collecting the real-time position information of the roller, a monocular camera for detecting the roadside curbs on both sides, four millimeter-wave radars, an angle sensor, and a speed encoder;

[0074] The GNSS rover is a GNSS dual-antenna, fixed on the top of the cab without obstacle occlusion; the GNSS rover communicates with the GNSS base station and at the same time communicates with the GNSS board installed on the decision-making subsystem; the monocular camera is fixed on both sides of the roller cab; the speed encoder is installed on the tire of the roller;

[0075] The four millimeter-wave radars are respectively installed at the front end of the front frame, the rear end of the rear frame, and both sides of the cab, for detecting surrounding obstacles and transmitting the obstacle information to the decision-making subsystem through the CAN protocol;

[0076] The angle sensor is implemented by an angle encoder and installed at the articulated joint between the front frame and the rear frame of the roller for obtaining the relative rotation angle of the roller body;

[0077] An angle sensor, millimeter-wave radars, monocular cameras, and GNSS rovers are installed on each roller in the roller fleet, and the entire roller fleet shares a GNSS base station;

[0078] The decision-making subsystem is a roller control board, and the main control of the roller control board uses a single-chip microcomputer. The GNSS board, angle sensor, millimeter-wave radar, and monocular camera all communicate with the single-chip microcomputer;

[0079] The management subsystem is a host computer, including a PC-side page, a host computer interface for controlling the roller fleet, and a real-time operation status display interface. The path planning algorithm, longitude and latitude network distance calculation method, missed compaction area target detection model, and unqualified area target detection model are loaded in the management subsystem. The path planning algorithm includes the path planning of the compaction area and the path planning of the lane-changing area; the operator remotely controls the roller through the host computer and views the operation conditions and compaction quality of the roller fleet.

[0080] Furthermore, the control subsystem mainly includes: modification of the gear execution mechanism. The gears of the roller include forward gear, reverse gear, and parking gear. A stepper motor is used as an additional power assist device. The gear shift lever is rotated by the torque of the motor to achieve the gear shifting function. The rotation position of the motor is controlled by a pulse signal; modification of the steering execution mechanism. The electric steering wheel is used to replace the original steering wheel in the steering execution mechanism; modification of the switches, including modification of the ignition switch, vibration switch, and high / low vibration switches, which are used to control the static pressure operation of the roller or the adjustment of the vibration mode of the roller. The vibration mode of the roller is divided into high vibration mode and low vibration mode, and a relay is used to control the switches; modification of the throttle execution mechanism. An electric motor power assist device is used to drive the throttle lever to achieve acceleration or deceleration.

[0081] The transmission subsystem uses a wireless transmission module. The antenna is placed on the top of the roller cab, and the decision-making subsystem and the management subsystem are communicated through the virtual serial port.

[0082] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0083] 1. The present invention establishes a path planning combining dynamic and static methods, which realizes the switching between dynamic and static path planning according to different construction scenarios. For the soil subgrade, static path planning is adopted for operation, that is, according to the collected compaction area information, the compaction area is pre-divided into several working sections with fixed lengths, and the machine fleet rolls over the working sections in sequence; for the asphalt surface course, due to its material characteristics, the roller fleet needs to alternately roll behind the paver in a certain order, with higher real-time requirements. At this time, dynamic path planning based on the position interaction between the paver and the roller fleet is adopted. The GNSS device is installed on the paver body, and the driving distance of the paver is preset. When the paver detects that the body has traveled to the preset distance, it stops, and the final stop position of the paver is used as the end point of paving. During actual construction, the paver first paves the filler in the compaction area. After the paver has traveled a certain safe distance, the roller fleet enters the site and real-time collects and judges the distance between the fleet and the paver, and presets the distance between the fleet and the paver. When the distance between the roller and the paver is too small, the roller will decelerate to ensure the safety of construction. When the distance between the roller and the paver is too large, the roller will accelerate. After the paver stops, the body coordinates of the paver are sent to the management subsystem, and the coordinates after leaving a safe distance are used as the rolling end point of the current working section of the fleet. The path of the fleet for this working section is generated and sent to the fleet. After the fleet arrives at the rolling end point of this working section, the fleet performs reciprocating rolling of this working section. After the compaction of this working section is completed, the paver continues to travel, and the fleet sets a working section overlap distance from the rolling end point of the current working section to the already rolled part to determine the rolling starting point of the next working section. The paver stops again, and the stopping position of the paver with a safe distance reserved is used as the rolling end point of the next working section of the fleet. The path of the fleet for the next working section is generated based on the rolling starting point and rolling end point of the next working section, and the fleet performs reciprocating rolling in the next working section along this path; return to the paver and continue to travel, and so on until the entire compaction area is rolled. A lap distance is set at the junction of two working sections to ensure that there is no missed rolling at the junction. The dynamic path planning ensures the synchronization of construction, solves the problem that the filler cools down too quickly to be rolled in time during winter construction, and realizes "paving and rolling immediately" and synchronous operation.

[0084] 2. The unmanned road machine fleet collaborative operation system of the present invention will select suitable rollers for corresponding operation zones according to the collected and uploaded compaction area boundary information and compaction material characteristics. According to the vehicle body parameters of each roller, such as model and wheel width, operation zones equivalent to the number of rollers are divided; according to the construction scenario, a suitable roller model and path planning method are selected. For example, when compacting soil subgrade, a single-drum roller is selected for static path planning, and when compacting asphalt, a double-drum roller is selected for dynamic path planning. An operation path is generated, and the operation path information includes several parallel rolling paths and turning paths. The machine fleet will carry out compaction operations on each operation zone. It realizes the "local conditions" machine fleet operation, shortens the construction period and improves the utilization efficiency of compaction machinery.

[0085] 3. The dynamic obstacle avoidance strategy in the present invention determines the detection range based on the real-time speed of the roller and conducts omnidirectional safety detection. Through devices such as millimeter-wave radars, monocular cameras, and speed encoders installed on the roller machine fleet, the obstacle avoidance and safety detection functions of the rollers are realized. Among them, the speed encoder is used to detect the speed of the roller and determine the safety threshold during the operation of the machine fleet according to the speed; the millimeter-wave radars are installed on the front, rear, left, and right of the vehicle body to detect surrounding obstacles and realize mutual avoidance between multiple rollers in special cases; since the working area is mostly roads and the roadside curbs on both sides are relatively low, which is a detection blind area for the millimeter-wave radar, the monocular cameras are installed on both sides of the roller vehicle body to identify the roadside curbs on both sides to prevent the roller from driving out of the working area and causing danger.

[0086] 4. The present invention analyzes and evaluates the compaction quality after the compaction operation of a fleet of rollers, and makes precise compaction quality optimization for the areas where the compaction quality does not meet the standard. After the first compaction operation of the fleet is completed, the management subsystem can view the global compaction cloud map of this operation, and the compaction quality of each point in each operation area can be viewed. First, the uncompacted areas in the compaction area are recompacted to prevent "missing compaction"; after full coverage compaction in the compaction area, select the areas that have been rolled but whose quality does not meet the requirements and whose predicted compaction degree does not meet the requirements in the management subsystem, and resend the coordinate information of these areas to the main controller of the roller for recompaction. During the recompaction process, in order to avoid repeatedly rolling the areas where the compaction degree has met the requirements many times, resulting in "over compaction", when passing through the fully compacted areas that have reached the standard, the "static pressure" method is selected to reduce the impact on the compacted areas. At the same time, the distance between the roller and the area to be recompacted is calculated in real time. When the roller is within the set recompaction distance threshold range from the recompaction area, the roller automatically turns on the vibration mode for recompaction. After the roller drives out of the recompaction area, the vibration mode is turned off, and the recompaction ends. At this time, the compaction quality is viewed again, and the unqualified areas are recompacted twice. This cycle continues until the global compaction quality meets the standard, and then the compaction operation ends, and the fleet of rollers exits. This avoids the disadvantages such as easy occurrence of missing compaction, under compaction, and over compaction in the past operation of rollers, and realizes the automatic "quality control" of the compaction operation.

[0087] 5. The present invention realizes the high-precision unmanned operation of a fleet of rollers through GNSS devices, electric steering wheels, angle sensors, etc. installed on the fleet of rollers and path tracking control algorithms. And the intelligent compaction equipment installed in the cabs of the fleet of rollers monitors the compaction quality of the operation of the fleet of rollers throughout the process. The intelligent compaction equipment installed on the roller collects the vibration signals of the steel wheel in real time through an acceleration sensor, processes the vibration signals and establishes a compaction degree evaluation model based on artificial intelligence algorithms, realizes real-time monitoring of the compaction degree throughout the process, and combines with the on-board GNSS equipment to provide the whole process and all-round monitoring and guidance for the entire cross-section of the highway rolling, forming a real-time, continuous and intelligent "quality feedback surface". The data of the compaction operation is transmitted through a 4G wireless network card, and the data is transmitted and stored in the management subsystem, which can realize real-time monitoring and display of the compaction data on the web page. The main stored data includes: the speed of the roller, vibration frequency, compaction passes, predicted compaction degree, longitude and latitude coordinates, etc. At the same time, it can realize real-time monitoring of information such as the running track, speed, compaction quality, and position of the vehicle, achieve full-process monitoring of the compaction area, and realize the "positive evaluation" of the compaction operation.

[0088] 6. The present invention establishes a more accurate plane relative rectangular coordinate system of the compaction area through the longitude and latitude network distance calculation method and GNSS equipment. The plane relative rectangular coordinate system is used for the path planning of the roller fleet and to achieve the tracking control of the rollers. During tracking control, it can automatically control the electric steering wheel and perform tracking adjustments one by one according to continuous target points. It has high real-time performance and simple calculation. This method avoids problems such as excessive and inaccurate conversion values in the absolute coordinate system of the earth, establishes a high-precision plane relative rectangular coordinate system of the compaction area, and improves the accuracy and precision during the operation of the rollers.

[0089] 7. The present invention dynamically optimizes the parameters of the rollers during the compaction operation of the roller fleet. When the rollers perform compaction operations, various factors may affect the compaction quality. The main influencing parameters of the rollers include: the speed of the rollers, vibration frequency, etc. When traditional manually driven rollers are used to adjust the parameters of the rollers, more often than not, the driver adjusts the speed and vibration frequency of the rollers based on experience. However, this method has large errors and uncertainties. Therefore, the present invention establishes a fitness function between the compaction degree and the speed and vibration frequency of the rollers, and through iteration, obtains the optimized optimal speed and optimal vibration frequency. The optimized roller parameters are sent to the main controller, and the main controller controls the throttle, brake actuators, and large and small vibration buttons to achieve changes in the speed and vibration frequency of the rollers. Among them, the speed is calculated by a speed encoder installed on the roller tires, and acceleration and deceleration operations are performed in real time according to the difference between the optimized optimal speed and the current speed to achieve feedback regulation and control of the speed. The present invention solves the uncertainty of determining the speed and vibration frequency relying on the experience of the driver, sets parameter constraints according to the actual construction situation, obtains the optimal roller parameters that conform to the current compaction situation, keeps the compaction quality in the optimal state, and controls the rollers to operate with the optimal roller parameters, realizing the "reverse optimization" of the compaction operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 It is a schematic structural diagram of the mechanical-data dual-driven road unmanned fleet collaborative operation system of the present invention.

[0091] Figure 2 It is a schematic diagram of the installation positions of each module of the driverless roller fleet.

[0092] Figure 3 It is a schematic diagram of the positions of the four corner points of the compaction area in the plane relative rectangular coordinate system.

[0093] Figure 4 It is a schematic diagram of the operation area division of the fleet.

[0094] Figure 5It is a schematic diagram of the rolling path, lane-changing path of a road roller, and the overlapping distance between adjacent rolling lanes within an operation section.

[0095] Figure 6 It is a schematic diagram of the lane-changing path curve.

[0096] Figure 7 It is a schematic diagram of the conversion between the plane relative rectangular coordinate system of the compaction area and the relative coordinate system of the lane-changing path.

[0097] Figure 8 It is a schematic diagram of the process of the pre-set static path planning for subgrade construction.

[0098] Figure 9 It is a schematic diagram of the process of the dynamic path planning based on the interaction between the paver and the road roller fleet for asphalt surface layer construction.

[0099] Figure 10 It is a logic judgment diagram for the optimization and adjustment of road roller parameters during the operation of the fleet.

[0100] Figure 11 It is a schematic diagram of the path tracking control based on target point tracking during the operation of the fleet.

[0101] Figure 12 It is a schematic diagram of the compaction quality optimization process after the first compaction operation of the fleet.

[0102] In the figure, 1. Sensing subsystem, 2. Decision-making subsystem, 4. Transmission subsystem, 3. Control subsystem, 5. Management subsystem, 11. GNSS rover, 12. Millimeter-wave radar, 13. Speed encoder, 14. Monocular camera, 15. Angle sensor, 16. GNSS base station, 17. Acceleration sensor. Specific implementation manners

[0103] The following gives specific embodiments of the present invention. The specific embodiments are only used to further elaborate the present invention in detail and are not limited to the protection scope of this application. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0104] Embodiment 1:

[0105] The mechanical-data dual-driven unmanned road roller fleet collaborative operation system in this embodiment (see Figure 1), mainly including a fleet of rollers, a sensing subsystem 1, a decision-making subsystem 2, a control subsystem 3, a transmission subsystem 4, and a management subsystem 5. The sensing subsystem and the decision-making subsystem are both installed on the roller. The sensing subsystem is electrically connected to the decision-making subsystem; the control subsystem is used to control the gear shifting, steering, throttle supply, working mode, etc. of the roller, is connected to the decision-making subsystem, and is directly controlled by the decision-making subsystem; the transmission subsystem is used to transmit data and instructions, and connects the decision-making subsystem and the management subsystem; the management subsystem receives the operation data of the roller through the transmission subsystem, sends instructions to the fleet of rollers, and obtains the operation data of the paver.

[0106] The fleet of rollers includes several single-drum rollers or double-drum rollers of different models and weights. The roller includes a front frame and a rear frame, and the front frame and the rear frame are connected by a hinge shaft. The cab is located on the rear frame, and the rolling drum is located on the front frame.

[0107] The sensing subsystem 1 mainly includes:

[0108] 1. RTK GNSS. It is used to collect the real-time position information of the roller, including a GNSS base station 16 and a GNSS rover 11, and a GNSS board (not shown in the figure), to achieve differential solution. The GNSS base station is placed in an open environment with high sky visibility; the GNSS rover 11 is a GNSS dual antenna. Two GNSS rovers 11 are installed on an aluminum alloy bracket, and then the aluminum alloy bracket is horizontally fixed on the top of the cab through a strong magnetic chuck without any obstacles blocking. The GNSS board is installed on the decision-making subsystem, and transmits the positioning data to the decision-making subsystem through a serial port. The data protocol adopts the GPGGA form in the NEMA0183 protocol. The GNSS board obtains the positioning data by parsing the message and is used for subsequent path calculation.

[0109] 2. Monocular camera. The monocular camera 14 is fixed on both sides of the roller cab and is used to detect the road edges on both sides to prevent the roller from getting into danger due to driving out of the working area. The monocular camera needs to be calibrated and single-target calibrated before use. After a series of image processing, the lateral distance between the roller and the road edge is obtained. When the lateral distance from the road edge is less than 15 cm, parking and route adjustment will be carried out. If the lateral distance from the road edge is not less than 15 cm, return to the camera detection step for continuous monitoring.

[0110] 3. Millimeter-wave radar. Four millimeter-wave radars 12 are respectively installed at the front end of the front frame, the rear end of the rear frame, and both sides of the cab, used to detect surrounding obstacles, and transmit the obstacle information to the decision-making subsystem through the CAN protocol. The decision-making subsystem makes judgments and corresponding operations. At the same time, the obstacle information will also be uploaded to the management subsystem by the transmission subsystem and displayed. Compared with both sides, the possibility of "rear-end collision" at the front and rear ends is greater and the danger is also greater, and it increases with the increase of speed. Therefore, it is necessary to increase the safety threshold of the roller. However, an excessive safety threshold may cause the detection of the millimeter-wave radar to be too sensitive, resulting in the roller braking repeatedly and affecting the construction efficiency. Therefore, for the speed of the roller, the detection range of the front and rear millimeter-wave radars is adjusted in real time to ensure the safety and efficiency of the roller operation; the millimeter-wave radars on both sides are mainly used to detect rollers in other areas and lateral obstacles, that is, it is divided into two situations: obstacle detection and roller detection. The process of obstacle detection is: when it is judged that a lateral obstacle is detected, a parking operation is performed, and a warning is given in the management subsystem, and the azimuth and distance information of the lateral obstacle are displayed. The lateral obstacle is processed after parking, and work continues after the lateral obstacle is cleared; if no lateral obstacle is detected, the obstacle detection continues.

[0111] The process of roller detection is: when the millimeter-wave radars on both sides detect a roller vehicle in other areas except the current rolling lane, it means that the distance between the two vehicles is too close. At this time, a parking operation is performed on the two roller vehicles, a warning is given in the management subsystem, and the relative position information between the two roller vehicles and the deviation from the planned route are quickly judged. Subsequently, the two vehicles adjust their routes successively, that is, the route of the vehicle in front is corrected first. After a certain distance is vacated, the route of the vehicle behind is corrected; if no roller vehicle in other areas is detected, the roller detection continues.

[0112] 4. Angle sensor. The angle sensor 15 is implemented by an angle encoder and is installed at the articulated connection between the front frame and the rear frame of the roller, used to obtain the relative rotation angle of the roller body and transmit the data to the decision-making subsystem for analysis and calculation, for the path tracking control of the roller, and the communication protocol is the RS485 protocol.

[0113] Angle sensors, millimeter-wave radars, monocular cameras, and GNSS rovers 11 are installed on each roller in the roller fleet in the above manner, and the entire roller fleet shares a GNSS base station.

[0114] The decision-making subsystem is the control board of the roller. The main control of the control board of the roller uses an STM32 single-chip microcomputer. The GNSS board, angle sensor, millimeter-wave radar, and monocular camera all communicate with the STM32 single-chip microcomputer, and the GNSS board communicates with the GNSS rover. This control board of the roller is the "brain" of the entire driverless roller, and all judgments and corresponding decisions are made on this control board of the roller.

[0115] The control subsystem 3 is mainly used to control the gear shifting, steering, throttle supply, working mode, etc. of the roller, and mainly includes: modification of the gear execution mechanism, modification of the steering execution mechanism, modification of the switch, and modification of the throttle execution mechanism. For the modification of the gear execution mechanism, the gears of the roller include forward gear, reverse gear, and parking gear. A stepper motor is used as an external assisting device, and the gear push rod is rotated by the torque of the motor to achieve the gear shifting function. The rotation position of the motor is controlled by a pulse signal. For the modification of the steering execution mechanism, an electric steering wheel is used to replace the original steering wheel. For the modification of the switch, it includes the modification of the ignition switch, vibration switch, and large / small vibration switch, which are used to control the static pressure work of the roller or the adjustment of the vibration mode of the roller. The vibration mode of the roller is divided into large vibration mode and small vibration mode, and a relay is used to control the switch. For the modification of the throttle execution mechanism, a motor and other assisting devices are used to drive the throttle push rod to achieve acceleration or deceleration. The modification of this part can be achieved according to the existing technology.

[0116] The transmission subsystem 4 uses a wireless transmission module, and the antenna is placed on the top of the cab of the roller. The decision-making subsystem and the management subsystem are communicated through the virtual serial port.

[0117] The management subsystem 5 is a host computer, including a PC-side page, a host computer interface for controlling a fleet of rollers, a real-time operation status display interface, etc. Operators can remotely control the rollers through the host computer and view the operation conditions and compaction quality of the fleet of rollers.

[0118] Embodiment 2:

[0119] The mechanical-data dual-driven road unmanned fleet collaborative operation method in this embodiment can realize the collaborative operation of a fleet of driverless rollers with full coverage of the compaction area and can control the compaction quality. The specific steps are as follows:

[0120] Step 1, construction boundary information collection and operation path generation. Taking the north as the positive direction of the Y-axis and the east as the positive direction of the X-axis, an origin is set on the southwest side of the construction area to construct a plane relative rectangular coordinate system.

[0121] In this embodiment, a single GNSS rover is placed 50 to 100 m southwest of the area to be compacted to collect the longitude and latitude information of the origin of the longitude and latitude coordinate system, and the collected longitude and latitude information of the origin of the longitude and latitude coordinate system is input into the management subsystem. The origin of the longitude and latitude coordinate system is used as the origin of the plane relative rectangular coordinate system of the compacted area.

[0122] The management subsystem incorporates a longitude and latitude grid distance calculation method and a path planning algorithm. The longitude and latitude grid distance calculation method is used to convert longitude and latitude coordinates into plane rectangular coordinates, and the plane relative rectangular coordinate system can reflect the actual distance of the compacted area.

[0123] Select a suitable roller model according to the characteristics of the compacted material and the boundary information of the compacted area. Use the longitude and latitude grid distance calculation method to obtain the plane rectangular coordinates of the center line of the compacted area, and generate a map of the compacted area based on the plane rectangular coordinates of the center line of the compacted area and the road width.

[0124] According to the different construction sites, a dynamic and static combined path planning is established to realize the switching between dynamic and static path planning. According to the wheel width of each type of roller, the operation area is divided according to the proportion of the roller wheel width and the operation path is generated. Specifically: for subgrade construction, use static path planning, pre-divide several equal-length operation sections for the compacted area, and the path of all operation sections in the compacted area will be automatically generated when generating the operation path, and the machine fleet will roll over all operation sections in sequence; for asphalt surface layer construction, use dynamic path planning, and the path of each operation section will be dynamically generated during work. The machine fleet does not have the end point information of the first operation section. When starting work, the machine fleet follows the paver for rolling, and uses the position where the paver stops each time and reserves a certain safety distance as the rolling end point of the current operation section of the machine fleet.

[0125] The longitude and latitude grid distance calculation method means that there are multiple positioning points on the surface of the ellipsoid. The longitude and latitude coordinates of the i-th positioning point are recorded as ( , ). Taking ( , ) as the longitude and latitude coordinates of the reference point, calculate the relative coordinate position of the positioning point and the reference point. In actual use later, the origin collected by the GNSS rover is used as the reference point, and the calculation formula is as follows:

[0126]

[0127]

[0128] Among them, L e is the difference in the longitude direction, and B e is the difference in the latitude direction.

[0129] By multiplying the difference in the longitude direction and the difference in the latitude direction by the conversion coefficients for converting to a plane rectangular coordinate system, the points on the plane relative rectangular coordinate system of the compaction area can be obtained. The calculation formula is as follows:

[0130]

[0131]

[0132] Among them, k L is the conversion coefficient in the longitude direction; k B is the conversion coefficient in the latitude direction; x is the actual difference distance in longitude, that is, the x coordinate of the plane relative rectangular coordinate system; y is the actual difference distance in latitude, that is, the y coordinate of the plane relative rectangular coordinate system.

[0133] For two points on the same meridian, the actual distance corresponding to a 1-degree difference in latitude is 111.2018 kilometers, that is, the value of k B is 111.2018 kilometers.

[0134] For two points on the equator, the actual distance corresponding to a 1-degree difference in longitude is 111.3195 kilometers, that is, the value of k L is 111.3195 kilometers.

[0135] On other latitudes except the equator, the actual distance corresponding to a 1-degree difference in longitude between two points is 111.3195·cosB kilometers. At this time, the value of k L is 111.3195·cosB kilometers, and B is the latitude value of the positioning point.

[0136] The path planning algorithm is executed in both static path planning and dynamic path planning;

[0137] In on-site construction, the longitude and latitude coordinates of the center line of the compaction area and the road width are often given. First, the longitude and latitude coordinates of the center line of the compaction area are converted into the plane rectangular coordinates of the plane relative rectangular coordinate system through the calculation method of the distance on the graticule. Subsequently, the entire compaction area map is generated based on the plane rectangular coordinates of the center line of the compaction area and the road width.

[0138] Compaction area calculation: Set the road width as W, and the wheel width of the i-th roller in the roller fleet is b i (i = 1, 2...n), where n represents the number of rollers, and the corresponding operation zones S i (i = 1, 2...n) responsible for each roller are generated; in the plane relative rectangular coordinate system, the north is the positive direction of the Y-axis, and the east is the positive direction of the X-axis. The center point coordinates on the center line of the compaction area are (x j , y j ), (j = 1, 2...m), where m is the number of points on the center line; the coordinates of the four corner points of the compaction area are (xoe , y oe ), where \(e = 1, 2, 3, 4\). Define the azimuth angle \(\alpha\) of the center line of the compaction area as the acute angle between the center line of the compaction area and the \(X\)-axis in the plane relative to the rectangular coordinate system. Define the azimuth angle formed with the positive direction of the \(X\)-axis as positive and the azimuth angle formed with the negative direction of the \(X\)-axis as negative, and the absolute value of the azimuth angle is less than \(90^{\circ}\). When \(0 < \alpha < 90^{\circ}\), the point closest to the \(Y\)-axis among the four corner points of the compaction area is denoted as \(o1\), and the other corner points are determined in a clockwise order as \(o2\), \(o3\), \(o4\) in sequence; when \(-90^{\circ} < \alpha < 0\), the point closest to the \(X\)-axis among the four corner points of the compaction area is denoted as \(o1\), and the other corner points are determined in a clockwise order as \(o2\), \(o3\), \(o4\) in sequence. For details, see Figure 3 .

[0139] By default, the compaction area is a regular rectangle. In the embodiment, the starting coordinates \((x 1 ,y 1 ) and the ending coordinates \((x m , y m ) on the center line of the compaction area can be used to calculate the slope \(k^*\), and then according to \(k^*=\tan\alpha\), the azimuth angle \(\alpha\) of the center line of the compaction area can be solved inversely. The specific formula is:

[0140]

[0141] Then, the coordinates of the four corner points of the entire compaction area are calculated according to the following formula:

[0142]

[0143]

[0144]

[0145]

[0146] Among them, \(W\) is the road width; \(\alpha'\) is the azimuth angle of the boundary \(o1o4\) or \(o2o3\) of the compaction area. When \(0 < \alpha < 90^{\circ}\), \(\alpha'=\alpha - 90^{\circ}\), and when \(-90^{\circ} < \alpha < 0\), \(\alpha'=\alpha + 90^{\circ}\).

[0147] The rectangular area formed by connecting the four corner points (coordinates are \((x oe , y oe )) is the entire compaction area. Next, the path planning of the compaction area is carried out.

[0148] Path planning of the compaction area: According to the number of each selected roller and the ratio of the wheel widths, divide the operation area for each roller in the compaction area. As Figure 4 shown, the fleet consists of 3 rollers, and the ratio of the wheel widths is 1:2:1. Then, the entire compaction area is divided into three operation areas \(S\)1 , S 2 , S 3 , corresponding to the working zone widths of W / 4, W / 2, and W / 4, Figure 4 in which rollers 1, 2, and 3 are respectively responsible for the compaction work in the working zones S 1 , S 2 , S 3 . Taking the working zone S 1 as an example for path planning, the same applies to other working zones. The path planning algorithm includes the path planning in the rolling area and the path planning in the lane-changing area. The specific implementation of the path planning in the rolling area is as follows:

[0149] Figure 5 in which, let the width of the working zone S 1 be W 1 , the wheel width of the roller responsible for the working zone S 1 be b 1 , and the overlapping distance between adjacent rolling lanes be L. Then the actual effective rolling width of one rolling lane is D 1 = b 1 - L. The number of rolling lanes N in this working zone = roundup (W 1 / D 1 ), where N is the integer obtained by rounding up W 1 / D 1 , and roundup() is the rounding-up function. According to the coordinates of the four corner points of the compaction area, the starting and ending coordinate information of the midline of each rolling lane is calculated using the following formula:

[0150] The starting coordinate (S 1 x 1 , S k1 y 1 ) of the midline of each rolling lane in the working zone S k1 is:

[0151]

[0152]

[0153] The ending coordinate (S 1 x 1 , S kE y 1 ) of the midline of each rolling lane in the working zone S kE is:

[0154]

[0155]

[0156] where k represents the k-th pass (k = 1, 2... N), D 1 is the effective rolling width of the roller in the operation area S 1 and the value of the effective rolling width D is different for different operation areas.

[0157] When the subsequent roller performs rolling, it will operate along the median line of each rolling pass. The roller reciprocates between each rolling pass, and the rolling path is the median line of each rolling pass.

[0158] The specific implementation of the path planning in the lane-changing area is as follows:

[0159] When the roller completes the preset number of rolling passes in the current rolling pass, it will perform a lane-changing operation. After the lane change, the roller will immediately perform the compaction operation on a new pass. In order to shorten the lane-changing time and the position error after the lane change, it is necessary for the roller to straighten the body and have a zero articulated steering angle after the lane change, and the entire lane-changing path needs to be continuous and smooth, meeting the kinematic and dynamic constraints of the roller. The present invention uses a polynomial curve as the lane-changing path. The polynomial curve should meet the requirement that the curvature is zero at the starting point and the ending point, and the route is smooth and stable during the lane-changing process, and the calculation is more convenient, and there will be no problem that the body of the roller cannot be straightened when it reaches the starting point of the next rolling pass and it is difficult for the roller to adjust the route.

[0160] The present invention uses a cubic polynomial curve for lane-changing path planning. First, a relative coordinate system for the lane-changing path is established to facilitate obtaining the curve equation of the lane-changing path and the coordinate information of the points on the curve. Subsequently, the relative coordinate system of the lane-changing path is converted to the plane relative rectangular coordinate system of the compaction area.

[0161] Taking the first operation area and the roller responsible for the operation of the first operation area as an example, the lane-changing paths of the subsequent operation areas are the same. Assume the wheel width of the roller is b 1 , and the overlapping distance between adjacent rolling passes is L, then the lateral lane-changing distance is D * 1 =b 1 -L, and the longitudinal lane-changing distance is s. Taking the ending point of the lane change as the local sampling point, the value of s is about 6m;

[0162] Taking the driving direction of the roller after the lane change as the x-axis direction to establish a coordinate system, then the starting point coordinates are (0, D * 1 ), the ending point coordinates of the lane change are (s, 0), and the lane-changing duration is t. When reaching the ending point of the lane change, the body of the roller is straightened. From the coordinate constraints of the starting point and the ending point, we can obtain:

[0163] x(0)=0, y(0)=D * 1 ;

[0164] (0)=0, (0)=0;

[0165] (0)=0, (0)=0;

[0166] x(t)=s, y(t)=0;

[0167] (t)=0, (t)=0;

[0168] (t)=0, (t)=0;

[0169] From the curvature constraints at the starting point and the ending point, we can obtain: ρ(0)=0, ρ(t)=0. The superscript "'" represents the first derivative, and the superscript "" represents the second derivative.

[0170] The curvature calculation formula is:

[0171]

[0172] Let the lane-changing curve equation be:

[0173]

[0174] Substitute the coordinate constraints (0, D * 1 ), (s, 0) and the curvature constraints ρ(0)=0, ρ(t)=0 into it, and obtain the values of the parameters a, b, c, and d in the lane-changing curve equation, and then get the lane-changing curve equation of the roller (see Figure 6 ), and obtain the coordinate information of several points on the curve.

[0175] For the sake of simplicity in calculation and a neat equation form, the above calculation established a coordinate system for the lane-changing curve separately, rather than in the plane relative rectangular coordinate system of the compaction area established before. And since during construction, the machine group operates in the plane relative rectangular coordinate system of the entire compaction area, therefore, it is necessary to perform coordinate transformation on the above-obtained lane-changing curve equation and the coordinate information of the points on the curve. After coordinate transformation, the lane-changing path is obtained.

[0176] Assume that in the plane, the relative coordinate system of the lane-changing path is , and the coordinate of a point on the curve is ( , ), the plane relative rectangular coordinate system of the compaction area is XOY, and the corresponding coordinate of a point on the curve is (X, Y). The rotation angle of the coordinate system relative to XOY is , and the rotation angle The positive direction of the Y-axis of the XOY coordinate system and the included angle with the positive direction of the axis of the coordinate system of the axis, with counterclockwise being positive. The relative coordinate system origin of the lane-changing path The translation displacements of the relative rectangular coordinate system origin O of the plane relative to the compaction area in the X-axis and Y-axis directions are DX and DY respectively. DX and DY are positive along the square of the XOY coordinate system (see Figure 7 ). It can be deduced that the coordinate transformation formula from to XOY is:

[0177]

[0178]

[0179] All the rolling paths and lane-changing paths during construction operations are obtained therefrom. Subsequently, through target tracking of points, path tracking control of the roller will be achieved. Figure 5 In [Figure], the blue dashed line is the midline of the rolling lane, the blue solid curve is the lane-changing path, and the gray area is the overlapping distance between adjacent rolling lanes. Figure 5 In it is the operation partition S 1 The rolling area and lane-changing area marking schematic diagram, where the area of the lane-changing area is W 1 ×s. The lane-changing areas and rolling areas of all operation partitions are collectively called the compaction area.

[0180] Step 2, the compaction operation of the driverless roller fleet. The roller fleet drives into the operation starting point and waits for the start of construction.

[0181] For subgrade construction, static path planning is used, and the path information of all generated operation segments is sent to the rollers responsible for each operation partition in the format of a txt document. At the start of the operation, given the end information of the first operation segment, first the path of the first operation segment is sent to the fleet, and the fleet performs the rolling operation on this operation segment. The compaction steps in the operation segment are as follows: The fleet rolls according to the pre-divided rolling lanes, starting from the first rolling lane. After reaching the end of the first rolling lane, it reverses for rolling. After reaching the starting point of the first rolling lane again, it stops, changes lanes, and starts the compaction of the next rolling lane, and then performs the above operations, and so on, until the roller completes the rolling of this operation segment, which is regarded as the end of the operation of this operation segment. The compaction passes and the initial parameters of the roller can be preset. The end point and the starting point of the rolling lane in the same operation segment are the same, both being the starting point and the end point of this operation segment. After the first operation segment is rolled, the path of the second operation segment is sent. There is a certain overlapping distance between the two operation segments, and so on, until the compaction is completed (see Figure 8 ).

[0182] For asphalt surface layer construction, dynamic path planning is adopted, such as Figure 9As shown in the figure, the yellow area in the figure is the end position of each working section of the paver. The rolling end of each working section of the machine group is the end position - safety distance of the paver's working section. The rolling start point of the next working section is the rolling end - overlap distance of the previous working section. The machine group does not have the end information of the first working section. Therefore, when starting work, the machine group will first follow the paver for rolling, using the position where the paver stops each time and leaving a safety distance as the rolling end of the machine group in the current working section. When the paver reaches the working end of the paver in each working section, it will obtain the coordinates installed on the paver body at this time and leave a safety distance to prevent the machine group from rear-ending. The coordinates on the paver body are corrected using the safety distance, and the corrected coordinates are used as the rolling end of the machine group in the current working section. The path of the machine group in this working section is generated and sent to the machine group. The machine group rolls back and forth in this working section according to this path. After all rolling is completed, the paver continues to drive, and the machine group sets an overlap distance for the rolled part from the rolling end of the current working section to ensure that there is no missed rolling between the two working sections. Then, it continues to work in the next working section, and so on until the entire compacted area is rolled.

[0183] 1. Before the machine group starts working, the monocular camera needs to be calibrated. The monocular camera is calibrated through MATLAB software to obtain the camera internal parameters for subsequent distance measurement. During the compaction operation of the machine group, the monocular cameras of the rollers on both sides of the compacted area will collect the road edge picture information in real time. After a series of processing on the pictures, the lateral distance between the side of the roller body and the road edge is obtained. When the lateral distance from the road edge is less than 15 cm, the protection mechanism is triggered, and the roller stops. Then, it is adjusted according to the error between the roller position and the planned route to make the roller work again.

[0184] 2. When the machine group is working, a dynamic obstacle avoidance strategy is adopted for obstacle avoidance. The millimeter-wave radars in the front and rear of the vehicle body are mainly used to detect obstacles in the driving direction of the roller, and the danger increases with the increase of speed. The dynamic obstacle avoidance strategy is: the detection range of the front and rear millimeter-wave radars is dynamically adjusted according to the speed of the roller, that is, it meets the requirement of L0 = 2·v, where L0 is the detection range in meters, and v is the real-time speed of the roller in m / s.

[0185] The detection logic for obstacles in the driving direction is: when the machine group detects an obstacle in the driving direction during work, it stops and gives a warning, and continues to work after the obstacle in the driving direction has moved away.

[0186] The lateral obstacle detection logic is as follows: when the roller detects a lateral obstacle, it stops and gives a warning, and continues to work after the lateral obstacle is removed; if a roller vehicle in other areas is detected, both roller vehicles stop, quickly judge the lateral distance, relative position information between the two roller vehicles, and the deviation amount from the preset route, and then the roller vehicle in the front adjusts the route first and continues to operate. After traveling a certain distance, the roller vehicle at the back adjusts the route.

[0187] 3. During the operation of the fleet, the intelligent compaction equipment installed on the roller will collect the vibration signals of the steel wheel in real time through the acceleration sensor 17, process the vibration signals and establish a compaction degree evaluation model based on the artificial intelligence algorithm to realize real-time monitoring of the compaction degree throughout the process, and the compaction path and compaction data can be visually displayed; in addition, the parameters of the roller will be dynamically optimized according to the change of the compaction degree, establish a fitness function between the compaction degree and the speed and vibration frequency of the roller, and set the value range of the speed and vibration frequency of the roller as a constraint. In actual construction, the roller will not operate at too fast or too slow a speed, so the speed range of the roller is set to 0.5~8 km / h, and the accuracy is 0.1 km / h; the vibration mode is divided into two modes: large vibration and small vibration.

[0188] The specific process of optimizing and adjusting the roller parameters (see Figure 10 ) is as follows: set the speed range to 0.5~8 km / h, and the accuracy is 0.1 km / h; the vibration mode is divided into two modes: large vibration and small vibration. The vibration frequency is automatically matched within each mode system. Set the maximum number of iterations. Through iteration, obtain the result with the highest fitness, that is, obtain the optimal speed and the optimal vibration frequency. Send the optimized roller parameters to the main controller of the decision-making subsystem, and the main controller controls the execution components of the control subsystem such as the throttle, brake actuator, and large and small vibration buttons to realize the change of speed and vibration frequency. Among them, the roller will calculate the current speed of the roller in real time through the speed encoder installed on the tire, use the optimized optimal speed as the target speed, subtract the current speed from the target speed. If the current speed of the roller reaches the target vehicle speed, continue to drive at the current vehicle speed. If the target speed is not reached, drive the throttle actuator of the roller according to the speed error to adjust the error until the current speed of the roller is consistent with the optimal speed; the vibration frequency is controlled by the level signal. When switching between large and small vibrations, use the optimized optimal vibration frequency as the target vibration frequency, judge the current vibration frequency state according to the current level signal. If the current vibration frequency is consistent with the target vibration frequency, continue to maintain without operation. If the current vibration frequency is inconsistent with the target vibration frequency, reverse the level signal to realize the switching of the vibration frequency and realize the dynamic optimization of the roller parameters during the compaction operation.

[0189] 4. During fleet operation, path tracking control of the roller will be achieved through the path tracking control algorithm and the sensing devices installed on the roller body. Set the azimuth angle of the front wheel of the roller as θ 1 , the azimuth angle of the vehicle body as θ 2 , the relative angle of the vehicle body as φ, and it is stipulated that the relative angle φ of the vehicle body is negative when it is deflected to the left relative to θ 2 and positive when it is deflected to the right; the coordinates of the roller are (x B , y B ), the coordinates of the target point are (x A , y A ), and the formula for the angle between the line connecting the roller and the target point and the Y-axis is,

[0190]

[0191] According to , the current expected steering angle β of the front wheel is obtained according to the following formula. It is stipulated that if β is negative, the steering wheel is turned to the left, and if β is positive, the steering wheel is turned to the right:

[0192] (1) 0 < θ 1 <180°

[0193] ① If y A <y B , then β = - θ 1 + 180°.

[0194] ② If y A >y B , then β = - θ 1 .

[0195] (2) 180° < θ 1 <360°

[0196] ① If y A <y B , then β = - θ 1 + 180°.

[0197] ② If y A >y B , then β = - θ 1 + 360°.

[0198] During the actual operation of the roller, the azimuth angle θ 2 of the vehicle body is obtained by the GNSS device (GNSS rover) installed on the vehicle body, and the relative angle φ of the roller vehicle body is obtained by the angle sensor installed at the hinge of the roller. The azimuth angle θ 1 of the front wheel of the roller = θ 2+φ. When the roller is operating, it will calculate the angular deviation between the current position and the target position in real time, and drive the electric steering wheel to correct this angular deviation to achieve the path tracking control of the roller (see Figure 11 ). During operation, whether it is the path tracking control on a straight line in the rolling area or the path tracking control on a curve in the lane-changing area, it is to track several points on the straight line or curve and calculate the distance from the target point in real time. After detecting that the target point has been reached, it will automatically track the next position.

[0199] Step 3, Real-time evaluation of compaction quality and precise optimization of the compaction area. After the first compaction operation of the fleet is completed, a global compaction cloud map will be displayed in the management subsystem. Different compaction degrees in the compaction area will be displayed in different colors, and the position information of each point in the whole area can be displayed. According to this compaction cloud map, quality optimization of the non-compliant compaction area will be carried out precisely. During optimization, first perform supplementary compaction on the missed compaction area, collect information on the missed compaction area, generate a path, and the roller performs supplementary compaction. After the entire compaction area is fully covered and compacted, then perform recompaction on the area where the quality does not meet the standard. When performing recompaction, first collect information on the area where the compaction quality does not meet the standard and generate an operation path, and the roller performs recompaction. After recompaction, judge again whether the global compaction degree meets the standard. If it meets the standard, the operation is completed; if it does not meet the standard, perform secondary recompaction until the global compaction degree of the compaction area meets the requirements.

[0200] Use the target detection algorithm to identify and detect the places where the compaction quality is unqualified. First, obtain and organize the data set, organize and manually annotate the compaction cloud map obtained by the roller equipped with intelligent compaction equipment. The compaction quality of the compaction area is divided into three categories. One is the basically unrolled area, that is, the missed compaction area; the second is the rolled area but with unqualified quality; the third is the fully compacted area. Mark and frame the missed compaction area and the rolled area but with unqualified quality. When marking, pay attention to the framing range. Since the compaction cloud map may be uneven, the target detection area should be completely framed within the detection frame during marking. Subsequently, construct, train and verify the model.

[0201] In this embodiment, the official model of YOLOv8 is used, and some adjustments are made to the model using the Python language. The samples are replaced with the compaction quality classifications to be recognized, and two YOLOv8 target detection models are established. The first YOLOv8 target detection model is used to detect the missed compaction area, and the second YOLOv8 target detection model is used to detect the rolled area but with unqualified quality. Divide the data set used for training each YOLOv8 target detection model into a training set, a validation set and a test set according to 7:2:1. Input the training set into the model to train the model. Finally, use the test set to test the model, and embed the two trained YOLOv8 target detection models into the management subsystem.

[0202] First, the areas where compaction is missed in the compaction area will be recompacted. The compaction degree of the areas not rolled by the machine fleet is significantly low. Click "Detect Missed Compaction Areas" in the management subsystem, and the target detection model for the missed compaction areas will be called to perform target detection and bounding on the missed compaction areas. After the missed compaction areas are bounded, the detection box will display the coordinate information of the four corner points of the bounded area. Then, the above path planning algorithm is used to plan the path for the missed compaction areas. The system will automatically generate the recompaction path and instructions, and then the management subsystem will send the instructions to the single roller responsible for compacting the missed compaction areas to control it to recompact the missed compaction areas.

[0203] Further, after full-coverage compaction in the compaction area, click "Detect Areas with Unqualified Quality" to detect the areas that have been rolled but have unqualified quality. Similarly, after the detection box bounds the areas that have been rolled but have unqualified quality, the coordinate information of the four corner points of this area will be generated (see Figure 12 ), determine the recompaction area, then generate the path information and instructions and resend them to the main controller of the roller to perform recompaction. Set the recompaction distance threshold. During the recompaction process, in order to avoid repeatedly rolling the areas where the compaction degree has met the requirements, resulting in "overcompaction", when passing through the fully compacted areas that have reached the standard, the "static pressure" method is selected to reduce the impact on the fully compacted areas. At the same time, the distance from the roller to the area to be recompacted is calculated in real time. When the distance is within the set recompaction distance threshold range, the roller automatically turns on the vibration mode to perform recompaction. After the roller exits the recompaction area, the vibration mode is turned off, and the recompaction ends. At this time, check the compaction quality again, and perform secondary recompaction on the unqualified areas. Repeat this cycle until the global compaction quality meets the standard, then the compaction operation ends, and the roller fleet exits.

[0204] Matters not described in this invention apply to the prior art.

Claims

1. A mechanical-data dual-driven road unmanned machine group collaborative operation method, characterized in that: The operation method includes the following contents: The model of the roller group is selected according to the characteristics of the compacted material and the boundary information of the compacted area, the plane rectangular coordinates of the center line of the compacted area are obtained by using the longitude and latitude grid distance calculation method, and the compacted area map is generated according to the plane rectangular coordinates of the center line of the compacted area and the road width; According to different construction scenarios, a path planning combining dynamic and static methods is established. Specifically, for earth roadbed construction, static path planning is used. The static path planning method is as follows: the compaction area is divided into several equal-length work sections in advance, and the paths of all work sections in the compaction area are automatically generated when the work path is generated. The machine group rolls all work sections in turn. For asphalt surface layer construction, dynamic path planning is used. The path of each work section is dynamically generated during work. The process of dynamic path planning is as follows: the machine group does not have the end point information of the first work section, and the starting point of the paver is used as the rolling starting point of the first work section. When the operation starts, the machine group follows the paver for rolling, and the paver stops each time and a safety distance is reserved as the machine group. The group determines the rolling end point of the current working section, generates the path of the group of the working section with the rolling starting point and rolling end point of the current working section and sends it to the group of machines. The group of machines reciprocates and rolls in the current working section using this path. After all rolling, the paver continues to drive, and the group of machines sets a section of overlapping distance from the rolling end point of the current working section to the rolled part, determines the rolling starting point of the next working section, and the paver stops again. The paver parking position and a safe distance are used as the rolling end point of the next working section of the group of machines, and the rolling starting point and rolling end point of the next working section are used to generate the path of the group of machines for the next working section. The group of machines reciprocates and rolls in the next working section using this path; the paver continues to drive, and so on until the compaction area is fully rolled; Path planning algorithms are executed in both static path planning and dynamic path planning; The compaction degree of the whole process is obtained in real time during the rolling process. After the compaction operation of the machine group is completed, the global compaction cloud map is displayed. Based on the compaction cloud map, the quality of the compacted areas that do not meet the standards is precisely optimized. The compaction cloud map is divided into three categories according to the compaction quality: the area that has been rolled but the quality does not meet the standard, and the area that is completely compacted. They are marked separately, and then a data set of the area that has been leaked and a data set of the area with unqualified compaction quality are established. The leaking area data set is used to train the target detection model of the area that has been leaked, and the data set of the area with unqualified compaction quality is used to train the target detection model of the unqualified area. During construction, the missed compaction area is identified by the target detection model of the missed compaction area in the compaction cloud map, the coordinate information of the four corner points of the framed position of the missed compaction area is obtained, and the path planning algorithm is used to perform recompacting on the missed compaction areas of the earth roadbed construction and the asphalt surface layer construction respectively; then, the unqualified area target detection model is used to identify the area that has been rolled but does not meet the quality standard in the compaction cloud map after recompacting, the coordinate information of the four corner points of the framed position of the area that has been rolled but does not meet the quality standard is obtained, the recompacting area is determined for recompacting, and a recompacting distance threshold is set. During the recompacting process, when passing through the fully compacted area, the roller is set to the static compaction mode, and at the same time, the distance from the roller to the area to be recompacted is calculated in real time. When the distance is within the set recompacting distance threshold range, the roller automatically turns on the vibration mode for recompacting. After the roller leaves the recompacting area, the vibration mode is turned off, the recompacting is completed, the compaction quality is checked again, and the unqualified area is recompacted for the second time until the global compaction quality meets the standard, and the compaction operation is completed; The latitude and longitude grid distance calculation method is: there are multiple positioning points on the surface of the ellipsoid, and the longitude and latitude coordinates of the i-th positioning point are marked as (L i ,B i ), taking (L1, B1) as the latitude and longitude coordinates of the reference point, calculate the relative coordinate position between the positioning point and the reference point. The calculation formula is: L e =L i -L1 B e =B i -B1 Among them, L e is the longitude difference, B e is the latitude difference; The longitude direction difference and the latitude direction difference are multiplied by the conversion coefficient converted to the plane rectangular coordinate to obtain the point on the plane relative rectangular coordinate system of the compacted area. The calculation formula is: x=L e ·k L y=B e ·k B Among them, k L is the conversion coefficient in the longitude direction, k B is the conversion coefficient in the latitude direction, x is the x-coordinate of the plane relative to the rectangular coordinate system, and y is the y-coordinate of the plane relative to the rectangular coordinate system.

2. The operation method according to claim 1, characterized in that: The process of generating the compaction area map according to the plane rectangular coordinates of the center line of the compaction area and the road width is: during the on-site construction, the north is the positive direction of the Y axis, the east is the positive direction of the X axis, the origin is set at the southwest side of the construction area, and a plane relative rectangular coordinate system is constructed; Set the road width to W, and the coordinates of the center point on the center line of the compaction area to (x j ,y j ), j = 1, 2...m, m is the number of points on the center line of the compacted area; the coordinates of the four corner points of the compacted area are (x oe ,y oe ), e = 1, 2, 3, 4, define the acute angle between the center line of the compacted area and the X-axis in the plane relative rectangular coordinate system as the azimuth angle α of the center line of the compacted area, and define the azimuth formed with the positive direction of the X-axis as positive, and the azimuth formed with the negative direction of the X-axis as negative; When 0<α<90°, the point closest to the Y axis among the four corner points of the compaction area is recorded as o1, and the other corner points are determined by analogy clockwise as o2, o3, and o4; when -90°<α<0, the point closest to the X axis among the four corner points of the compaction area is recorded as o1, and the other corner points are determined by analogy clockwise as o2, o3, and o4; According to the coordinates of the starting point and the end point on the center line of the compacted area, the azimuth angle α of the center line of the compacted area is calculated according to the following formula: Then calculate the coordinates of the four corner points of the compacted area according to the following formula: Wherein, α′ is the azimuth angle of the compaction area boundary o1o4 or o2o3, when 0<α<90°, α′=α-90°, when -90°<α<0, α′=α+90°; At this point, the coordinates of the four corner points of the entire compacted area are obtained, and the rectangular area formed by the lines connecting the four corner points is the entire compacted area.

3. The operation method according to claim 2, characterized in that: The path planning algorithm includes path planning for the rolling area and path planning for the lane change area. According to the number of selected rollers and the ratio of wheel width, each roller is divided into operation zones according to the compaction area. The wheel width of the roller group is b i , i = 1, 2...n, n represents the number of rollers, and the corresponding operation partition S is generated for each roller i ; The path planning process of the rolling area is: For the work area S1, let the width of the work area S1 be W1, the wheel width of the roller responsible for the work area S1 be b1, and the overlapping distance between adjacent rolling tracks be L. Then the actual effective rolling width of a rolling track is D1=b1-L, and the number of rolling tracks of the work area N=roundup(W1 / D1), where roundup() is an upward rounding function; According to the coordinates of the four corner points of the compaction area, the starting and ending coordinates of the median line of each rolling track are calculated using the following formula: The starting coordinates of the median line of each rolling track in the operation area S1 (S1x k1 , S1y k1 )for: The end point coordinates of each rolling track median line in the operation area S1 (S1x kE , S1y kE )for: Wherein, k represents the kth rolling, k = 1, 2, ..., N; When the subsequent roller is rolling, it operates along the center line of each rolling path, and the roller rolls back and forth between each rolling path; The process of lane change area path planning is: When the roller completes the preset number of rolling passes on the current rolling track, it changes lanes. After changing lanes, the roller immediately starts compacting the next track. After changing lanes, the roller straightens the vehicle body and sets the articulated steering angle to zero. The lane-changing path planning is performed using a cubic polynomial curve: For the first working area and the roller responsible for the first working area, assuming that the roller wheel width is b1 and the overlapping distance between adjacent rolling tracks is L, the lane-changing lateral distance D * 1 = b1-L, the longitudinal distance of lane change is s, and the end point of lane change is used as the local sampling point; The relative coordinate system of the lane changing path is established with the driving direction of the roller after lane changing as the x-axis direction, and the starting point coordinates are (0, D * 1), the coordinates of the lane change end point are (s, 0), the lane change duration is t, and the roller body is straightened when it reaches the lane change end point. Assume the lane change curve equation is: y=ax 3 +bx 2 +cx+d Constrain the coordinates (0,D * 1) (s, 0) and curvature constraints ρ(0) = 0, ρ(t) = 0 are substituted to obtain the values ​​of parameters a, b, c, d in the lane-changing curve equation, and then the lane-changing curve equation of the roller and the coordinate information of the points on the curve are obtained; The lane change curve equation obtained above and the coordinate information of the points on the curve are converted into a plane relative rectangular coordinate system of the compaction area.

4. The operation method according to claim 3, characterized in that: In actual operation of the roller, the body azimuth angle θ2 is obtained by the GNSS equipment installed on the body, and the relative rotation angle of the roller body The angle sensor installed at the hinge of the roller gives the azimuth angle θ1 of the front wheel of the roller: When the group is working, the path tracking control of the roller is realized through the path tracking control algorithm and the sensor equipment installed on the roller body; The coordinates of the roller are (x B ,y B ), the target point coordinates are (x A ,y A ), the angle Ψ between the line connecting the roller and the target point and the Y axis is: The current expected front wheel turning angle β is calculated based on Ψ. If β is a negative number, the steering wheel is turned left, and if β is a positive number, the steering wheel is turned right. The specific β is calculated by the following formula: 1)0<θ1<180° ① If y A < y B , then β = Ψ - θ1 + 180°, ② If y A > y B , then β = Ψ - θ1; 2)180°<θ1<360° ① If y A < y B , then β = Ψ - θ1 + 180°, ② If y A > y B , then β = Ψ - θ1 + 360°; When the roller is operating, it calculates the angle deviation between the current position and the target position in real time, and drives the electric steering wheel of the roller to correct the angle deviation, thereby realizing the path tracking control of the roller.

5. The operation method according to claim 1, characterized in that: When the group is operating, a dynamic obstacle avoidance strategy is used to avoid obstacles. The dynamic obstacle avoidance strategy is: dynamically adjust the detection range L0 of the millimeter-wave radars in front and behind the roller according to the speed of the roller. The detection range meets the requirement of L0=2·v, where the unit of L0 is m, and v is the real-time speed of the roller in m / s; The logic of heading obstacle detection is: when a heading obstacle is detected during the operation of the fleet, the aircraft will stop and issue a warning, and continue to work after the heading obstacle is removed; The logic of lateral obstacle detection is as follows: when a roller detects a lateral obstacle, it stops and issues a warning, and continues to work after the lateral obstacle is removed; if a roller vehicle is detected in an area other than the current rolling track, both roller vehicles stop, and the lateral distance, relative position information, and deviation from the preset route between the two roller vehicles are quickly determined. Then, the roller vehicle in the front position adjusts its route first and continues to work, and the roller vehicle in the rear position adjusts its route after driving a certain distance. During the compaction operation of the machine group, the monocular cameras of the rollers on both sides of the compaction area collect curb image information in real time to obtain the lateral distance between the side of the roller body and the curb. When the lateral distance from the curb is less than 15 cm, the protection mechanism is triggered and the roller stops. Then, adjustments are made based on the error between the roller position and the planned route to put the roller back into operation.

6. The operation method according to claim 1, characterized in that: Each roller is equipped with intelligent compaction equipment to monitor the compaction degree in real time. The roller parameters are dynamically optimized according to the compaction degree changes, and the fitness function of compaction degree, speed and vibration frequency of the roller is established. The speed and vibration frequency ranges are set as constraints: the roller speed range is 0.5-8km / h, with an accuracy of 0.1km / h; the vibration mode is divided into two modes: large vibration and small vibration. Set the maximum number of iterations, obtain the optimal speed and optimal vibration frequency through the iterative process of the intelligent algorithm, and control the roller to perform compaction operations at the optimal speed and optimal vibration frequency.

7. A mechanical-data dual-driven unmanned road machine group collaborative operation system, comprising a roller group, a perception subsystem, a decision subsystem, a control subsystem, a transmission subsystem and a management subsystem, wherein the perception subsystem and the decision subsystem are both installed on the roller, and the perception subsystem is electrically connected to the decision subsystem; the control subsystem is used to control the gear switching, steering, throttle supply and working mode of the roller, and is connected to the decision subsystem and directly controlled by the decision subsystem; the transmission subsystem is used to transmit data and instructions; the management subsystem receives the operation data of the roller through the transmission subsystem, sends instructions to the roller group and obtains the operation data of the paver; the characteristics are: The operating system executes the operating method according to claim 1, The roller group includes a number of single-steel-wheel rollers and / or double-steel-wheel rollers of different models and weights, the rollers include a front frame and a rear frame, the front frame and the rear frame are connected by an articulated shaft, the cab is located on the rear frame, and the rolling steel wheels are located on the front frame; The perception subsystem includes: a GNSS rover for collecting real-time position information of the roller, a monocular camera for detecting the curbs on both sides, four millimeter-wave radars, an angle sensor, and a speed encoder; The GNSS mobile station is a GNSS dual antenna, fixed on the top of the cab, and is not blocked by obstacles; the GNSS mobile station communicates with the GNSS base station, and at the same time, the GNSS mobile station communicates with the GNSS board installed on the decision subsystem; the monocular camera is fixed on both sides of the roller cab; the speed encoder is installed on the tire of the roller; Four millimeter-wave radars are installed at the front end of the front frame, the rear end of the rear frame, and both sides of the cab to detect surrounding obstacles and transmit obstacle information to the decision-making subsystem; The angle sensor is implemented by an angle encoder and is installed at the hinged connection between the front frame and the rear frame of the roller to obtain the relative rotation angle of the roller body; Angle sensors, millimeter-wave radars, monocular cameras and GNSS mobile stations are installed on each roller in the roller fleet, and the entire roller fleet shares a GNSS base station; The decision-making subsystem is a roller control board, the main control of the roller control board adopts a single-chip microcomputer, and the GNSS board, angle sensor, millimeter wave radar, and monocular camera all communicate with the single-chip microcomputer; The management subsystem is a host computer, including a PC page, a host computer interface for controlling a group of rollers, and a real-time operation status display interface. The operator remotely controls the roller through the host computer and checks the operation status and compaction quality of the group of rollers.

8. The operating system according to claim 7, characterized in that: The management subsystem is loaded with a path planning algorithm, a latitude and longitude grid distance calculation method, a pressure leakage area target detection model and an unqualified area target detection model. The path planning algorithm includes path planning for the rolling area and path planning for the lane change area.

Images