An autonomous driving method for trackless rubber-tyred vehicles in underground roadways

Abstract

The present invention belongs to the field of driverless technology, and discloses a driverless method for trackless rubber-tyred vehicles facing underground roadways. The method includes: accurately determining tasks and roadway perception in the early stage; receiving the perceived and detected information through the state decision-making layer, and determining the driving state according to the detected information; the control decision-making layer selects corresponding control decisions according to the determined driving state; and the tracking control layer selects corresponding tracking control methods according to the selected control decisions. This method can adapt to special underground environmental conditions, perform different state decision-making judgments and control execution switching based on the perception results, ensure control accuracy, flexibly handle various dynamic scenarios such as meeting vehicles, bypassing obstacles, and stopping obstacles, reduce the number of manual interventions, improve the operation efficiency of driverless driving, and reduce system costs.

Description

Technical Field

[0001] The present invention belongs to the technical field of driverless technology, and particularly relates to a driverless method for a trackless rubber-tyred vehicle for underground roadways. Background Art

[0002] As an important part of the construction of national green mines, like open-pit mines, the development of driverless technology for underground mines is also particularly crucial. Compared with open-pit mines, underground mines have limited communication and high positioning costs, which pose high requirements for decision-making control and are difficult to achieve driverless operation. Currently, for driverless operation in underground roadways, there are control methods that rely on a combination of vehicle-mounted sensors and road surface sensors, which can achieve horizontal and longitudinal control in static scenarios, but the installation cost of road surface sensors is high; in addition, there is also a driverless planning control method for trackless rubber-tyred vehicles based on high-precision maps. Due to accurate map information and real-time positioning information, relatively accurate control can be achieved, but for mapping in long and narrow roadways, the data volume is large and the cycle is long. The existing research results can initially meet the needs of driverless operation of underground trackless rubber-tyred vehicles, but there is still room for improvement in terms of cost, preparation cycle, processing ability in dynamic environments, etc. Therefore, driverless operation of underground trackless rubber-tyred vehicles has not been widely applied at present. Summary of the Invention

[0003] In view of the problems existing in the above-mentioned prior art, the present invention proposes a driverless method for a trackless rubber-tyred vehicle for underground roadways.

[0004] The technical solution adopted by the present invention is specifically as follows:

[0005] A driverless method for a trackless rubber-tyred vehicle for underground roadways, comprising the following steps:

[0006] S1, preliminary preparation: Set up identification signs at the fork in the roadway. The identification signs carry characteristic information for distinguishing different forks and their corresponding curvature information that can be recognized by sensors;

[0007] Place curvature identification signs at large curvature bends, where the large curvature bends are bends with a curvature exceeding the curvature range that can be recognized by the sensors;

[0008] Place a task end identification sign at the end of the path;

[0009] Install sensors for sensing information in the roadway on the vehicle end;

[0010] S2, determine the task: According to the actual task requirements underground, determine the overall driving path; according to the determined overall driving path, determine the characteristic information on the identification signs that need to be recognized at each fork passed by the driving path;

[0011] S3, Roadway Sensing: Real-time sense the roadway boundary information, perform quadratic fitting, and send the boundary abnormal state to the state decision layer for state decision-making;

[0012] Real-time sense and output the obstacle information to the state decision layer for state decision-making;

[0013] Real-time sense and output the detected curvature sign information to the state decision layer for state decision-making;

[0014] The inertial navigation outputs the ramp information to the tracking control layer for tracking control;

[0015] S4, State Decision-making: Based on the obtained sensing information, make a driving state decision;

[0016] During the state decision-making process, judge different driving states through the output results of roadway sensing, and output the judged driving states to the control decision layer for control decision-making;

[0017] S5, Control Decision-making: The control decision layer processes the boundary information differently according to different driving states of the state decision layer, and sends the final boundary information to the tracking control layer;

[0018] S6, Tracking Control: According to the received ramp information and boundary information, select the corresponding control method; among them, the ramp information is used for longitudinal control, and the throttle / brake pedal opening is output longitudinally based on the fuzzy calibration table, and a ramp supplement is added; the boundary information is used for lateral control, including bilateral control, unilateral control, center control, and parking control;

[0019] S7, Task End: The infrared camera recognizes the task end sign, decelerates and stops, and clears the steering angle, and ends; and / or, when the parking control exceeds the given time, send an information of unable to continue passing to the control center, request manual takeover, and the current automatic driving task ends;

[0020] Otherwise, according to the real-time sensing information of the sensor, maintain the state switching control in steps S3 - S6.

[0021] Furthermore, during the state decision-making process, the driving states include:

[0022] Normal driving state, when the bilateral boundaries of the roadway are detected to be normal and no obstacles are detected;

[0023] Obstacle avoidance driving state, when the bilateral boundaries of the roadway are detected to be normal, an obstacle is detected, and the obstacle can be bypassed;

[0024] Abnormal driving state, when one side or both sides of the roadway boundary are missing and no obstacles are detected;

[0025] Curve driving state, when the roadway curvature identification information is detected and no obstacle is detected;

[0026] Takeover parking state, when the vehicle breaks down, the roadway boundary is normal and the obstacle cannot be bypassed, the roadway boundary is missing and an obstacle is detected, the roadway curvature identification information is detected and an obstacle is detected;

[0027] Among them, the curve driving state takes precedence over the abnormal driving state and the normal driving state.

[0028] Furthermore, for roadway information perception, when both sides of the roadway boundary are detected to be normal and no obstacle is detected, and the state is determined to be the normal driving state, the control decision is normal tracking, and the tracking control is bilateral control: determine the preview point P(x p ,y p ) through the preview distance, substitute x p into the left and right side boundary fitting curves to obtain y1 and y2 and calculate the target point Finally, convert the current coordinates to the rear axle center, and calculate the front wheel steering angle control amount using the tracking algorithm; the left side boundary fitting curve is y = a1x 2 + b1x + c1, and the right side boundary fitting curve is y = a2x 2 + b2x + c2.

[0029] Furthermore, for roadway information perception, when both sides of the roadway boundary are detected to be normal and an obstacle is detected, and the state is determined to be the obstacle avoidance driving state, the control decision is abnormal tracking, and the tracking control is unilateral control: determine the preview point P(x p ,y p ) through the preview distance, substitute x p into the side boundary fitting curve on the side away from the obstacle to obtain y1 and calculate the target point Q(x p ,y1 - d), where d is the boundary offset safety distance, and finally convert the current coordinates to the rear axle center, and calculate the front wheel steering angle control amount using the tracking algorithm; among them, if the obstacle is in the middle of the roadway, any side boundary fitting curve can be selected.

[0030] Furthermore, for roadway information perception, when one side of the roadway boundary is normal and the other side is abnormal and no obstacle is detected, and the state is determined to be the abnormal driving state, the control decision is abnormal tracking, and the tracking control is unilateral control: determine the preview point P(x p ,y p ) through the preview distance, substitute x p into the side boundary fitting curve on the normal side to obtain y1 and calculate the target point Q(x p ,y1 - d), where d is the boundary offset safety distance, and finally convert the current coordinates to the rear axle center, and calculate the front wheel steering angle control amount using the pure tracking algorithm.

[0031] Furthermore, for roadway information perception, when it is detected that both sides of the roadway boundary are abnormal and no obstacles are detected, the state is determined as an abnormal driving state, the control decision is abnormal tracking, and the tracking control is central control: based on the central curve y = a0x 2 +b0x + c0 fitted with the normal boundary parameters at the previous moment, the preview point P(x p , y p ) is determined through the preview distance, and x p is substituted into the central fitting curve to calculate the target point Q(x0, y0). Finally, the current coordinates are converted to the center of the rear axle, and the front-wheel steering angle control amount is calculated using the pure tracking algorithm.

[0032] Furthermore, for roadway information perception, when the roadway perception module detects the curvature identification information in the roadway information signboard and the state determination changes to driving on a curve, the control decision is curve tracking. At this time, the two sides of the boundary are fitted into a central curve, and the central curve is spliced with the circular trajectory according to the best entry point fitted by the curvature and the identified curvature; the central control is selected for the tracking control.

[0033] Furthermore, for roadway information perception, when the vehicle breaks down, the obstacle cannot be bypassed, the roadway boundary is missing and an obstacle is detected, or the roadway curvature identification information is detected and an obstacle is detected, and the state is determined as the takeover parking state, the control decision is to wait for parking, and the tracking control enters the parking control;

[0034] Parking control: The longitudinal throttle opening gradually decays to zero, then the braking opening gradually increases, the parking enable is activated when the vehicle speed is zero, and the pedal opening and steering wheel angle are initialized.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] 1. The driverless method for a trackless rubber-tyred vehicle for underground roadways provided by the present invention can adapt to the special underground environmental conditions, switch different state decision-making judgments and control executions based on the perception results, ensure the control accuracy, flexibly handle various dynamic scenarios such as meeting vehicles, bypassing obstacles, and stopping obstacles, reduce the number of manual interventions, improve the driverless operation efficiency, and reduce the system cost.

[0037] In addition, the driverless method for underground roadways of the present invention that does not rely on road surface sensors and roadway sensors relies on on-vehicle multi-sensors for environmental perception, and the control unit switches the control mode according to the environmental information to complete driverless driving.

[0038] 2. The present invention places digital identification signs through the pre-measured curvature information, identifies the identification signs with different digital features according to the task requirements during the actual operation process, and performs boundary fitting prediction to achieve stable driverless driving of vehicles at the fork, improving the adaptability of the system to complex environments.

[0039] 3. In view of abnormal situations such as unilateral boundary loss, bilateral boundary loss, and inability to perceive large curvature, the present invention solves the problem of the stability of unmanned driving control in abnormal perception scenarios and improves the operation efficiency and safety of the unmanned driving system through methods such as unilateral boundary optimization, center line fitting, and curvature fitting. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be construed as imposing any limitations on the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them:

[0041] Figure 1 is a schematic diagram of the control logic of the present invention;

[0042] Figure 2 is a schematic diagram of decision-making control during the process of the vehicle of the present invention entering a curve and a fork (where part a is entering the curve by sensing the sign information at the curve of the roadway and adopting central control; part b is identifying the sign information at the fork position of the roadway, selecting the curved path according to the task path, and entering the curve with central control);

[0043] Figure 3 is a schematic diagram of tracking path optimization in the state of boundary loss during the driving process of the vehicle of the present invention (where part a is the state where both bilateral boundaries are normal and there are obstacles ahead; part b is the state where there is a unilateral boundary and there are no obstacles ahead; c is the state where both bilateral boundaries are abnormal and there are no obstacles ahead);

[0044] Figure 4 is a schematic diagram of different boundary control principles;

[0045] Description of the reference numerals: P is the preview point, Q is the target point, d is the safety distance of boundary offset, and point O is the steering center of the vehicle. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] In order to more clearly understand the above-mentioned objects, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other.

[0047] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0048] Embodiment 1

[0049] In this embodiment, an unmanned driving method for a trackless rubber-tyred vehicle for underground roadways is disclosed, which includes roadway perception, state decision-making, control decision-making, and tracking control. Roadway perception is based on on-vehicle lidar and infrared cameras, and outputs driving boundary information, road curvature information, and obstacle information. State decision-making determines different driving states according to the perception information. Control decision-making outputs different tracking information according to different driving states. Tracking control calculates the control quantity based on the tracking information and finally controls the vehicle to drive along the roadway path.

[0050] Refer to the attached Figure 1 As shown, an unmanned driving method for a trackless rubber-tyred vehicle for underground roadways includes the following steps:

[0051] S1, Preliminary preparation: Set identification signs at the roadway fork. The identification signs carry infrared features that can be recognized by sensors to distinguish different forks and their corresponding curvature information;

[0052] Place curvature identification signs at large-curvature bends; In this embodiment, the determination basis for the "large curvature" is: during the driving process, it is determined by whether the radar in the perception module can normally detect the boundary. As those skilled in the art should know, when the curvature of the bend exceeds the detection range of the configured radar, that is, the radar cannot normally perceive, that is, the curvature is too large, it is called a large curvature; in this embodiment, the large curvature is curvature > 0.03. That is, the "large curvature" is the curvature beyond the radar detection range.

[0053] Place a task end identification sign at the path end;

[0054] Install sensors for perceiving information in the roadway on the vehicle end.

[0055] S2, Determine the task: According to the actual underground task requirements, determine the overall driving path; According to the determined overall driving path, determine the characteristic information on the identification signs that need to be recognized at each fork passed by the overall driving path. In this embodiment, the characteristic information is digital characteristics. That is, according to the overall driving path, determine the digital characteristics (infrared characteristics) on the identification sign that need to be recognized at the nth fork passed, that is, drive into the fork in the preset direction after recognizing the specific infrared characteristic at the nth intersection.

[0056] S3, Roadway perception: The lidar perceives the roadway boundary information in real time, performs quadratic fitting, and sends the boundary abnormal state to the state decision-making layer for state decision-making;

[0057] The lidar and infrared camera fuse to output obstacle information to the state decision layer for state decision-making. The infrared camera outputs the detected curvature sign information in real time to the state decision layer for state decision-making. The inertial navigation outputs ramp information to the tracking control layer for control decision-making;

[0058] S4. State decision: Make a driving state decision. The driving states include normal driving, abnormal driving, obstacle avoidance driving, curve driving, and takeover parking. Normal driving means the boundary is normal and no obstacle is detected. Obstacle avoidance driving means the boundary is normal but an obstacle is detected and can be bypassed. Curve driving means the infrared camera detects curvature sign information and no obstacle is detected. Takeover parking means any of the following situations: the vehicle has a fault, the roadway boundary is normal but the obstacle cannot be bypassed, the roadway boundary is missing and an obstacle is detected, or curvature sign information of the roadway is detected and an obstacle is detected;

[0059] During the state decision process, judge different driving states through the output result of roadway perception, and output the judged driving state to the control decision layer;

[0060] S5. Control decision: The control decision includes normal tracking, abnormal tracking, curve tracking, and parking waiting. Normal tracking corresponds to the normal driving state. Abnormal tracking corresponds to the obstacle avoidance driving and abnormal driving states. Curve tracking corresponds to the curve driving state. During curve tracking, curvature fitting is performed based on the curvature information and the current boundary information. Parking waiting corresponds to the upper-layer takeover parking state, and the control output is initialized during parking waiting;

[0061] The control decision layer processes the boundary information differently according to different decision states of the state decision layer, and sends the final boundary information to the tracking control layer.

[0062] S6. Tracking control: Select the corresponding control method according to the received ramp information and boundary information. Among them, the ramp information is used for longitudinal control. Longitudinally, the throttle / brake pedal opening is output based on the fuzzy calibration table, and a ramp supplement is added. The boundary information is used for lateral control. The lateral control is divided into bilateral control, unilateral control, center control, and parking control according to different states. Bilateral control calculates the control amount according to the issued bilateral boundary. Unilateral control tracks according to the processed unilateral information. Center control tracks according to the issued center line. Parking control is to decelerate and stop and reset the steering angle after receiving the control decision parking waiting command;

[0063] S7. Task end: The infrared camera recognizes the task end sign, decelerates and stops and resets the steering angle to end; and / or, if the parking control exceeds the given time, send an information of unable to continue passing to the control center, request manual takeover, and the current automatic driving task ends;

[0064] Otherwise, maintain the state switching control in steps S3 - S6 according to the real - time sensing information of the sensor.

[0065] Specifically:

[0066] The roadway perception conducts boundary detection, obstacle detection, and curvature recognition of bends and turnouts, and outputs the results to the state decision - making part. The state decision includes normal driving, abnormal driving, obstacle - avoidance driving, bend driving, and takeover parking states, and jumps correspondingly according to the sensing information; the control decision includes normal tracking, abnormal tracking, bend tracking, and parking waiting states, and jumps correspondingly according to the state decision - making information; the tracking control adopts corresponding control methods according to the upper - layer decision, including bilateral control, unilateral control, center control, and parking control, and its specific implementation process is as follows:

[0067] Curvature fitting, referring to Appendix Figure 2 , in this embodiment, for each underground turnout and large - curvature bends where lidar cannot sense, collect the steering wheel angle during manual driving and convert it into the front - wheel angle, calculate the average curvature according to the curvature - angle formula, and determine the best entry point of the bend (i.e., the starting point S of the fitting path in Appendix Figure 2 ).

[0068] Sign - board placement, in this embodiment, different infrared - reflection - characteristic curvature sign - boards are placed at the left, middle, and right of the turnout. The first part of the numbers on the sign - board represents the direction, "1" for left, "2" for middle, "3" for right, and the second - part numbers range from 0 to 999, which is 1000 times the corresponding curvature (i.e., the resolution is 0.001). For example, "1080" means the left - turn bend curvature is 0.08 (the specific installation position can be seen in the schematic diagram of Figure 2 ), and curvature sign - boards are also placed at large - curvature bends. The first part of the numbers is fixed as "0", and the second - part numbers are defined in the same way as at the turnout; a sign - board with the number "9999" is placed at the task end.

[0069] Task determination, according to the actual underground operation requirements, select a suitable driving path and configure the digital - feature sign - board parameters that the roadway perception part needs to identify at the turnout. This parameter consists of a series of numbers, and each number corresponds to the first - part number of the sign - board that needs to be identified. For example, "2 - 1 - 3 - 1 - 2" means going straight, turning left, turning right, going straight, and turning right respectively at the turnouts in the whole process.

[0070] Longitudinal control, longitudinally output the throttle / brake pedal opening based on the fuzzy calibration table and add a slope compensation amount. Since a mature algorithm is used, it will not be elaborated here.

[0071] The following controls all correspond to the lateral control in different states:

[0072] Boundary detection control. When the task starts to execute, the roadway perception first performs boundary detection, extracts the boundaries within 10m ahead based on the lidar point cloud data, and fits them into a quadratic curve in the radar coordinate system (with the radar as the origin, the direction along the vehicle axle as the x-axis, and the direction perpendicular to the x-axis and to the right as the y-axis). And consider the basic characteristics of the boundary to judge the abnormality of the fitted curve. If it is abnormal, the flag value of the curve is set to 0, and if it is normal, it is 1. Its format is as follows in the table:

[0073] Table 1 Fitted curves of both sides of the boundary

[0074] Direction Flag bit Fitting curve Left side 1 <![CDATA[y = a1x 2 + b1x + c1]]> Right side 1 <![CDATA[y = a2x 2 + b2x + c2]]>

[0075] When both sides of the boundary are normal and no obstacle is detected, the state decision is normal driving; the control decision is normal tracking; the tracking control is bilateral control: as Figure 4 shown in a below, determine the preview point P(x p ,y p ) through the preview distance, substitute x p into the fitted curves of the left and right side boundaries, obtain y1 and y2 and calculate the target point Finally, convert the current coordinates to the center of the rear axle and calculate the front wheel steering angle control amount using the pure tracking algorithm.

[0076] When one side of the detected roadway boundary is normal and the other side is abnormal, and no obstacle is detected, the state decision is abnormal driving; the control decision is abnormal tracking, as Figure 3 shown in b below; the tracking control is unilateral control: as Figure 4 shown in b below, determine the preview point P(x p ,y p ) through the preview distance, substitute x p into the fitted curve of the normal boundary, obtain y1 and calculate the target point Q(x p ,y1 - d), where d is the boundary offset safety distance. Finally, convert the current coordinates to the center of the rear axle and calculate the front wheel steering angle control amount using the pure tracking algorithm.

[0077] When both sides of the detected roadway boundary are abnormal and no obstacle is detected, the state decision is abnormal driving; the control decision is abnormal tracking, as Figure 3 shown in c below, the tracking control is center control: as Figure 4 shown in c below, based on the central curve y = a0x 2 + b0x + c0 fitted with the normal boundary parameters at the previous moment, determine the preview point P(x p ,y p ) through the preview distance, substitute x p into the central fitted curve to calculate the target point Q(x0, y0). Finally, convert the current coordinates to the center of the rear axle and calculate the front wheel steering angle control amount using the pure tracking algorithm.

[0078] Obstacle perception control. In this embodiment, since the underground mainly relies on lidar to perceive the boundary for positioning, the obstacle avoidance driving state is triggered only when both side boundaries exist and the obstacles can be bypassed. During the obstacle avoidance driving process, the roadway perception part will output the fusion perception result of the lidar and the infrared camera, output the obstacle information, and realize working conditions such as obstacle bypassing and passing vehicles.

[0079] After detecting an obstacle, it is determined whether it can pass based on the position of the obstacle and the boundary. In the actual driverless operation environment, especially on the driverless operation route, to ensure the transportation efficiency, the situation of parking control caused by obstacles will be avoided as much as possible. Therefore, taking obstacle bypassing as an example (if it is determined that it can pass, then the obstacle is bypassed):

[0080] As Figure 3 shown in a of, when an obstacle is detected on the right side, at this time the two side boundaries of the roadway are normal, the state decision is obstacle avoidance driving, and the control decision is abnormal tracking, as Figure 3 shown in b of; the tracking control is unilateral control, as Figure 4 shown in b of, and the subsequent control principle adopts the unilateral control in the above-mentioned boundary detection control.

[0081] Curve and turnout identification control. In the curvature fitting work, the large curvature boundaries that cannot be recognized by the lidar have been calibrated (the corresponding identification information signs are preset in the roadway). When the vehicle recognizes the curvature identification sign at the curve or the curvature identification sign at the turnout during driving, the state decision changes to curve driving, and the control decision changes to curve tracking, as Figure 2 shown in a of, at this time the two side boundaries are fitted into the central curve y = a0x 2 + b0x + c0, and according to the best entry point of the curvature fitting and the recognized curvature, the central curve and the circular trajectory are spliced; the tracking control is central control, and the control principle adopts the above-mentioned central control.

[0082] When the vehicle travels to the turnout, the roadway perception accurately recognizes the corresponding entrance identification sign and curvature according to the initially input task information, the state decision changes to curve driving, and the control decision changes to curve tracking, as Figure 2 shown in b of, and the central curve splicing and central control in this process are the same as those in the curve driving process.

[0083] Other state control. When the vehicle has a fault, the obstacle cannot be bypassed, the roadway boundary is missing and an obstacle is detected, or the roadway curvature identification information is detected and an obstacle is detected, and the state is determined to be the takeover parking state, the control decision is to wait for parking, the tracking control directly enters the parking control, the longitudinal throttle opening gradually decays to zero, then the braking opening gradually increases, the vehicle speed is zero when the parking is enabled, and the pedal opening and steering wheel angle are initialized.

[0084] In addition, only some embodiments are described above, and changes, modifications, additions, and / or variations can be made without departing from the scope and essence of the disclosed embodiments. The embodiments are illustrative rather than restrictive. In addition, the described embodiments relate to the currently considered most practical and preferred embodiments, and it should be understood that the embodiments should not be limited to the disclosed embodiments. On the contrary, it is intended to cover different modifications and equivalent arrangements included in the essence and scope of the embodiments. In addition, the various embodiments described above can be applied in combination with other embodiments. For example, aspects of one embodiment can be combined with aspects of another embodiment to achieve yet another embodiment. Additionally, the individual features or components of any given component can constitute additional embodiments.

[0085] The foregoing description of the embodiments is provided for purposes of illustration and description, and is not intended to be exhaustive or limiting of the present disclosure. Each element or feature of the specific embodiments is generally not limited to that specific embodiment, but where applicable, each element or feature is interchangeable and can be used in alternative embodiments, even if not specifically shown or described, and can be varied in many ways. Such variations are not regarded as a departure from the present disclosure, and all such variations are included within the scope of the present disclosure.

[0086] Therefore, it should be understood that the drawings and the description are provided herein by way of example to facilitate understanding of the present invention and should not constitute a limitation on its scope.

Claims

1. An unmanned driving method for trackless rubber-tyred vehicles in underground roadways, characterized in that, It includes the following steps: S1, Preliminary preparation: Set identification signs at the fork in the roadway. The identification signs carry characteristic information for distinguishing different branch roads and their corresponding curvature information that can be recognized by sensors; Place curvature identification signs at large curvature bends, where the large curvature bends are bends with a curvature exceeding the curvature range that the sensor can recognize; Place a task end identification sign at the end of the path; Install sensors at the vehicle end for sensing information in the roadway; S2, Determine the task: Determine the overall driving path according to the actual underground task requirements; according to the determined overall driving path, determine the characteristic information on the identification signs that need to be recognized at each fork in the driving path; S3, Roadway perception: Continuously sense the roadway boundary information and perform quadratic fitting, and send the boundary abnormal state to the state decision layer for state decision-making; Continuously sense and output obstacle information to the state decision layer for state decision-making; Continuously sense and output the detected curvature identification sign information to the state decision layer for state decision-making; The inertial navigation outputs the ramp information to the tracking control layer for tracking control; S4, State decision-making: Make a driving state decision according to the obtained perception information; During the state decision-making process, judge different driving states through the output results of roadway perception, and output the judged driving states to the control decision layer for control decision-making; S5, Control decision-making, the control decision layer processes the boundary information differently according to different driving states of the state decision layer, and sends the final boundary information to the tracking control layer; S6, Tracking control, select the corresponding control method according to the received ramp information and boundary information; among them, the ramp information is used for longitudinal control, and the throttle / brake pedal opening is output longitudinally based on a fuzzy calibration table, and a ramp supplement amount is added; the boundary information is used for lateral control, including bilateral control, unilateral control, center control, and parking control; S7, Task end, the infrared camera recognizes the task end identification sign, decelerates and stops, and the steering angle is cleared to end; and / or, if the parking control exceeds the given time, send an information of unable to continue passing to the control center, request manual takeover, and the current autonomous driving task ends; Otherwise, according to the real-time perception information of the sensor, maintain the state switching control in steps S3 - S6.

2. The driverless method for a trackless rubber-tyred vehicle for underground roadways according to claim 1, wherein During the state decision-making process, the driving states include: Normal driving state, when the bilateral boundaries of the roadway are detected to be normal and no obstacles are detected; Obstacle avoidance driving state, when the bilateral boundaries of the roadway are detected to be normal, obstacles are detected, and obstacle bypass can be achieved; Abnormal driving state, when the absence of one side boundary or the absence of bilateral boundaries of the roadway is detected and no obstacles are detected; Bend driving state, when the roadway curvature identification information is detected and no obstacles are detected; Takeover parking state, when the vehicle breaks down, the roadway boundary is normal and the obstacle cannot be bypassed, the roadway boundary is missing and an obstacle is detected, the roadway curvature identification information is detected and an obstacle is detected; Among them, the bend driving state takes precedence over the abnormal driving state and the normal driving state.

3. The driverless method for a trackless rubber-tyred vehicle facing an underground roadway according to claim 1 or 2, characterized in that Roadway information perception. When it is detected that both sides of the roadway boundary are normal and no obstacles are detected, and the state is determined to be the normal driving state, the control decision is normal tracking, and the tracking control is bilateral control: Determine the preview point P(x p ,y p ) through the preview distance, substitute x p into the left and right boundary fitting curves to obtain y1 and y2 and calculate the target point . Finally, convert the current coordinates to the center of the rear axle and calculate the front wheel steering angle control amount using the tracking algorithm; the left boundary fitting curve is y = a1x 2 + b1x + c1, and the right boundary fitting curve is y = a2x 2 + b2x + c2.

4. A driverless method for a trackless rubber-tyred vehicle for underground roadways according to claim 1 or 2, characterized in that Lane information perception: when the lane boundaries on both sides are normal and obstacles are detected, the state is determined to be obstacle avoidance driving state, the control decision is abnormal tracking, and the tracking control is unilateral control: the preview point P(x p ,y p ), and x p Substitute the boundary fitting curve on the side away from the obstacle to obtain y1 and calculate the target point Q(x p ,y1-d), d is the boundary offset safety distance, and finally the current coordinate is converted to the center of the rear axle, and the front wheel steering angle control amount is calculated using the tracking algorithm; among them, if the obstacle is in the middle of the lane, the boundary fitting curve on one side is selected.

5. A driverless method for a trackless rubber-tyred vehicle for underground roadways according to claim 1 or 2, characterized in that, Roadway information perception. When one side boundary of the roadway is normal, the other side boundary of the roadway is abnormal, no obstacle is detected, and the state is determined to be an abnormal driving state, the control decision is abnormal tracking, and the tracking control is unilateral control: Determine the preview point P(x p ,y p ) through the preview distance, substitute x p into the fitting curve of the side boundary in the normal state to obtain y1 and calculate the target point Q(x p , y1 - d), where d is the boundary offset safety distance. Finally, convert the current coordinates to the rear axle center and calculate the front wheel steering angle control amount using the pure tracking algorithm.

6. The driverless method for a trackless rubber-tyred vehicle facing an underground roadway according to claim 1 or 2, characterized in that Roadway information perception: when both sides of the roadway boundary are abnormal and no obstacles are detected, the state is determined as an abnormal driving state, the control decision is abnormal tracking, and the tracking control is central control: based on the central curve y = a0x 2 +b0x+c0 fitted with the normal boundary parameters at the previous moment, the preview point P(x p ,y p ) is determined through the preview distance, and x p is substituted into the central fitting curve to calculate the target point Q(x0,y0). Finally, the current coordinates are converted to the center of the rear axle, and the front wheel steering angle control amount is calculated using the pure tracking algorithm.

7. The driverless method for a trackless rubber-tyred vehicle facing an underground roadway according to claim 1 or 2, characterized in that Roadway information perception: When the roadway perception module detects the curvature identification information in the roadway information sign and the state determination changes to curve driving, the control decision is curve tracking. At this time, the two sides' boundaries are fitted into a center curve, and the center curve and the circular trajectory are spliced according to the best curve entry point fitted by the curvature and the identified curvature; the center control is selected for the tracking control.

8. The driverless method for a trackless rubber-tyred vehicle facing an underground roadway according to claim 2, characterized in that Roadway information perception: When the vehicle has a fault, the obstacle cannot be bypassed, the roadway boundary is missing and an obstacle is detected, or the roadway curvature identification information is detected and an obstacle is detected, and the state determination is the takeover parking state, the control decision is to stop and wait, and the tracking control enters the parking control; Parking control: The longitudinal throttle opening gradually decays to zero, then the braking opening gradually increases, the parking enable is activated when the vehicle speed is zero, and the pedal opening and the steering wheel angle are initialized.

Images