Automatic driving electric sightseeing vehicle obstacle avoidance system

Abstract

The invention provides an automatic driving electric sightseeing vehicle obstacle avoidance system, which comprises an environment sensing layer, the environment sensing layer comprises multi-sensor synchronous acquisition and dynamic obstacle tracking, the multi-sensor synchronous acquisition is used for acquiring environment data of a sightseeing vehicle, and the dynamic obstacle tracking is used for tracking the environment data of the sightseeing vehicle; the dynamic obstacle tracking is used for environmental data fusion, trajectory prediction and obstacle analysis; the vehicle group communication layer is used for real-time communication of position data between vehicles; and the collaborative decision-making layer comprises a dynamic right gaming and path planning module. According to the scheme, when the sightseeing vehicle recognizes that the vehicle is in the horizontal road environment and meets at the turning, the vehicle with the high score is selected according to the vehicle passing priority score to provide the priority; the selection of vehicle passing priorities in different modes can be conveniently provided for the sightseeing vehicle according to different driving environments, and support is provided for safe driving and avoidance of the automatic driving sightseeing vehicle.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle obstacle avoidance, and particularly to an obstacle avoidance system for an autonomous electric sightseeing vehicle. Background Art

[0002] In places such as tourist attractions, large schools, and hospitals, using an electric sightseeing vehicle for transportation can facilitate people's rides and save physical strength. When using an electric sightseeing vehicle, it can make full use of the surplus power during the low electricity consumption period at night for charging, enabling the power generation equipment of the power plant to be fully utilized day and night, which is beneficial to energy conservation and reducing carbon dioxide emissions. Therefore, it has gradually become a favored means of transportation. With the continuous development of autonomous driving technology, the combination of autonomous driving and sightseeing vehicles has become a development trend. However, during the autonomous driving of sightseeing vehicles, the safe obstacle avoidance of the vehicle is particularly important.

[0003] In the related prior art, the publication number: CN205844895U discloses an obstacle avoidance system for an autonomous electric sightseeing vehicle, including a controller, a positioning module, a deceleration and braking module, and an obstacle monitoring module. The controller is respectively connected to the positioning module, the deceleration and braking module, and the obstacle monitoring module. This obstacle avoidance system for an autonomous electric sightseeing vehicle realizes the autonomous driving, deceleration and braking, and obstacle avoidance of the electric sightseeing vehicle by setting an obstacle monitoring module, etc., and has functions such as decelerating when encountering a speed bump and decelerating and stopping when encountering a pedestrian.

[0004] Using the above-mentioned obstacle avoidance system for an autonomous sightseeing vehicle, in the case of passing through roads with different slopes, there is only one set of sightseeing vehicle avoidance modes. For uphill vehicles, when encountering oncoming vehicles and avoiding obstacles at a turning point, it is not convenient to provide different obstacle avoidance modes according to vehicles with different uphill and downhill degrees.

[0005] Therefore, it is necessary to provide an obstacle avoidance system for an autonomous electric sightseeing vehicle to solve the above technical problems. Summary of the Invention

[0006] The present invention provides an obstacle avoidance system for an autonomous electric sightseeing vehicle, which solves the problem in the related technology that it is not convenient to provide different obstacle avoidance modes according to vehicles with different uphill and downhill degrees.

[0007] To solve the above technical problems, the obstacle avoidance system for an autonomous electric sightseeing vehicle provided by the present invention includes:

[0008] An environmental perception layer, which includes multi-sensor synchronous acquisition and dynamic obstacle tracking. The multi-sensor synchronous acquisition is used to collect environmental data of the sightseeing vehicle, and the dynamic obstacle tracking is used for the fusion of environmental data, trajectory prediction, and obstacle analysis;

[0009] A vehicle group communication layer, which is used for real-time communication of position data between vehicles;

[0010] The collaborative decision-making layer, which includes a dynamic road right game and a path planning module. The dynamic road right game is used to analyze the passing priority of vehicles, and the path planning module provides route planning for vehicles according to the passing priority.

[0011] The execution control layer, which includes motion control. The motion control includes steer-by-wire and power distribution. The steer-by-wire uses fuzzy PID control for lateral deviation and active rear-wheel steering to compensate for the vehicle body attitude.

[0012] The feedback optimization layer is used to calibrate and enhance the parameters of the sightseeing vehicle obstacle avoidance database.

[0013] Among them, the situation of the passing road surface is analyzed according to the collected data. When the vehicle is in an uphill and downhill road environment, the sightseeing vehicle approaching the bend waiting area and going uphill has the right of way; when the vehicle is in a horizontal road environment, the sightseeing vehicle approaching the bend waiting area has the right of way.

[0014] Preferably, the multi-sensor synchronous acquisition includes a lidar array, a dual-mode camera, and a millimeter-wave radar; the lidar array constructs a 360° point cloud map in real time; the dual-mode camera identifies the obstacle attributes within 20m; the millimeter-wave radar monitors the relative speed of moving targets within 50m.

[0015] Preferably, the dynamic obstacle tracking fuses multi-source data through Kalman filtering on the one hand; after fusion, an obstacle trajectory prediction model with spatio-temporal correlation is established; special obstacles are marked according to the collected obstacle information.

[0016] Preferably, the vehicle group communication layer includes V2X information interaction and network topology maintenance. The V2X information interaction is used for the interaction of vehicle-to-vehicle data; the network topology maintenance dynamically allocates communication time slots based on the TDMA protocol on the one hand; on the other hand, LDPC coding is used to ensure communication reliability under a 90% packet loss rate.

[0017] Preferably, the calculation of the passing priority score:

[0018] P riority = 0.4Q passenger + 0.3E battery + 0.3U urgency .

[0019] Preferably, the dynamic road right game predicts potential conflicts within 3 seconds through a Bayesian game model on the one hand, and records the road right transfer agreement through blockchain on the other hand.

[0020] Preferably, the path planning module generates candidate paths based on the improved artificial potential field method:

[0021]

[0022] Preferably, the path planning module applies the geese flock following model to optimize the overall energy consumption of the vehicle fleet.

[0023] Preferably, the feedback optimization layer includes real-time parameter calibration, which dynamically adjusts the repulsive force coefficient of the virtual force field and the following distance based on the actual obstacle avoidance effect; the real-time parameter calibration also updates online through a deep learning model (using a federated learning framework to protect privacy data; synchronously updating the new obstacle classifier every 24 hours).

[0024] Preferably, the feedback optimization layer further includes enhanced simulation of the scenario library. The feedback optimization layer performs stress testing through a digital twin platform; and records special avoidance cases in the actual application scenario.

[0025] Compared with the related technologies, the obstacle avoidance system for an autonomous electric sightseeing vehicle provided by the present invention has the following

[0026] Beneficial effects:

[0027] By identifying and differentiating the road surface environment for the sightseeing vehicle to distinguish the driving environment where the sightseeing vehicle is located; when the sightseeing vehicle identifies that the vehicle is in an uphill or downhill road environment, when meeting at a turning point, the uphill vehicle is given priority; when the sightseeing vehicle identifies that the vehicle is in a horizontal road environment, when meeting at a turning point, the vehicle with a higher score according to the vehicle passing priority score is selected to be given priority; it is convenient to provide different modes of vehicle passing priority selection for the sightseeing vehicle according to different driving environments, providing support for the safe driving and avoidance of the autonomous sightseeing vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] 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 for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on the structures shown in these drawings without creative efforts.

[0029] Figure 1 It is a system block diagram of the obstacle avoidance system for an autonomous electric sightseeing vehicle provided by the present invention;

[0030] Figure 2 It is a system block diagram of the environment perception layer provided by the present invention;

[0031] Figure 3 It is a system block diagram of the vehicle group communication layer provided by the present invention;

[0032] Figure 4System block diagram of the collaborative decision-making layer provided by the present invention;

[0033] Figure 5 System block diagram of the execution control layer provided by the present invention;

[0034] Figure 6 System flowchart of the obstacle avoidance system for an autonomous driving electric sightseeing vehicle provided by the present invention;

[0035] Figure 7 Selection flowchart of the priority passage mode provided by the present invention;

[0036] Figure 8 Formula diagram of the calculation method of F_attract provided by the present invention;

[0037] Figure 9 Formula diagram of the calculation method of F_repel^obstacle provided by the present invention;

[0038] Figure 10 Formula diagram of the calculation method of F_repel^vehicle provided by the present invention.

[0039] The realization, functional features and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed implementation manners

[0040] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0041] The present invention provides an obstacle avoidance system for an autonomous driving electric sightseeing vehicle.

[0042] Please refer to Figures 1 to 2 , in the first embodiment of the present invention, the obstacle avoidance system for an autonomous driving electric sightseeing vehicle includes:

[0043] An environment perception layer, where the environment perception layer includes multi-sensor synchronous acquisition and dynamic obstacle tracking. The multi-sensor synchronous acquisition is used to collect environmental data of the sightseeing vehicle, and the dynamic obstacle tracking is used for the fusion of environmental data, trajectory prediction and obstacle analysis;

[0044] A vehicle group communication layer, where the vehicle group communication layer is used for real-time communication of position data between vehicles;

[0045] The collaborative decision-making layer, which includes a dynamic right-of-way game and a path planning module. The dynamic right-of-way game is used to analyze the passing priority of vehicles, and the path planning module provides route planning for vehicles according to the passing priority.

[0046] The execution control layer, which includes motion control. The motion control includes steer-by-wire and power distribution. The steer-by-wire uses fuzzy PID control for lateral deviation and active rear-wheel steering to compensate for the vehicle body attitude.

[0047] The feedback optimization layer is used to calibrate and enhance the parameters of the sightseeing vehicle obstacle avoidance database.

[0048] Among them, according to the collected data analysis of the passing road surface conditions, when the vehicle is in an uphill and downhill road environment, the sightseeing vehicle going uphill and approaching the turning area of the curve has the right of way; when the vehicle is in a horizontal road environment, the sightseeing vehicle with a higher passing priority score has the right of way.

[0049] Since when the sightseeing vehicle passes on the uphill and downhill road surfaces, the frequent start and stop of the uphill vehicle requires higher energy consumption and braking requirements. Therefore, how to provide obstacle avoidance convenience for the uphill vehicle during the uphill and downhill process of the sightseeing vehicle to reduce energy consumption and braking demand remains to be studied.

[0050] In this embodiment, the environment perception layer analyzes and identifies the corresponding driving environment (road surface slope) according to the collected environmental data, and distinguishes the driving environment according to the road surface slope, including two types:

[0051] The road surface slope angle > 5°; it is an uphill and downhill road environment, and the passing vehicle or vehicle group is in the uphill and downhill slope driving environment.

[0052] The road surface slope angle is between 0 and 5°; it is a horizontal road environment, and the passing vehicle or vehicle group is in the flat driving environment of the flat road surface.

[0053] The obstacle avoidance system of the sightseeing vehicle includes two passing avoidance modes:

[0054] The uphill and downhill avoidance mode. When the sightseeing vehicle is in this mode, when the uphill sightseeing vehicle meets or meets at a curve with the downhill sightseeing vehicle, the uphill sightseeing vehicle has direct passing priority; when the sightseeing vehicle is in the uphill and downhill road environment, the system automatically identifies and automatically switches the passing avoidance mode to the uphill and downhill avoidance mode.

[0055] The flat surface avoidance mode. When the sightseeing vehicle is in this mode, calculate the passing priority score of each sightseeing vehicle, and the sightseeing vehicle with a higher score has the passing priority; the system automatically identifies and automatically switches the passing avoidance mode to the flat surface avoidance mode.

[0056] Specifically, such as Figure 5As shown in the figure, the execution control layer includes motion control instructions and a fault tolerance mechanism. The motion control instructions are used for wire-controlled steering and power distribution (fuzzy PID control is used for lateral deviation during linear steering; active rear-wheel steering is used to compensate for the vehicle body attitude; dynamic coupling of regenerative braking and mechanical braking is used during power distribution; and hub motor torque vector control). The fault tolerance mechanism is used for controlled units and emergency modes (when the main control fails, the backup control unit is automatically switched; when communication is interrupted, the degradation mode is triggered: switching to pure vision positioning; enabling the mechanical limit device of the emergency steering wheel).

[0057] By identifying and differentiating the road surface environment of the sightseeing vehicle, the driving environment of the sightseeing vehicle can be distinguished. When the sightseeing vehicle identifies that it is in an uphill or downhill road environment and meets other vehicles at a turning point, the uphill vehicle is given priority. When the sightseeing vehicle identifies that it is in a horizontal road environment and meets other vehicles at a turning point, the vehicle with a higher score is selected according to the vehicle passing priority score to be given priority. This facilitates providing different modes of vehicle passing priority selection for the sightseeing vehicle according to different driving environments, and provides support for the safe driving and avoidance of the autonomous driving sightseeing vehicle.

[0058] Further, as Figure 2 shown, the multi-sensor synchronous acquisition includes a lidar array, a dual-mode camera, and a millimeter-wave radar. The lidar array constructs a 360° point cloud map in real time. The dual-mode camera identifies the attributes of obstacles within 20m. The millimeter-wave radar monitors the relative speed of moving targets within 50m.

[0059] The lidar array uses 8-line lidar * 4, which is distributed in a diamond shape to achieve 360° coverage without dead angles. The lidar array uses an adaptive point cloud density algorithm to maintain an accuracy of 200 points / ㎡ within 20 meters.

[0060] The dual-mode camera uses a visible light + thermal imaging dual-mode camera, and the dual-mode camera is based on a dynamic obstacle classification module for deep learning (pedestrian recognition rate > 99.8%).

[0061] The millimeter-wave radar is used for road surface state perception. On the one hand, it analyzes the ground echo of the millimeter-wave radar, and on the other hand, it also monitors the grip force in real time through a tire mechanics sensor.

[0062] This facilitates providing support for data acquisition for the perception fusion of three-dimensional data of the sightseeing vehicle obstacle avoidance system.

[0063] Further, as Figure 2 shown, the dynamic obstacle tracking, on the one hand, fuses multi-source data through Kalman filtering; after fusion, an obstacle trajectory prediction model with spatio-temporal correlation is established; special obstacles are marked according to the collected obstacle information.

[0064] Specifically, the dynamic obstacle tracking includes data fusion, trajectory prediction, and obstacle recognition. The data fusion fuses multi-source data through Kalman filtering;

[0065] The trajectory prediction establishes a spatio-temporal associated obstacle trajectory prediction model after data fusion;

[0066] The obstacle recognition labels special obstacles according to the collected obstacle information, and makes special marks on children, animals, or other high-risk targets for emergency obstacle avoidance / braking.

[0067] When the autonomous sightseeing vehicle faces a suddenly emerging special obstacle, the autonomous sightseeing vehicle enters the emergency obstacle avoidance / braking mode, and the autonomous sightseeing vehicle realizes automatic avoidance or emergency braking of the sightseeing vehicle while avoiding collision accidents.

[0068] It is convenient to perform adaptive recognition and analysis of obstacles in the driving environment, so as to provide support for the acquisition of obstacle data and the construction of models for the autonomous driving recognition of the sightseeing vehicle.

[0069] Further, as Figure 3 shown, the vehicle group communication layer includes V2X information interaction and network topology maintenance. The V2X information interaction is used for the interaction of vehicle-to-vehicle data; the network topology maintenance dynamically allocates communication time slots based on the TDMA protocol on the one hand; on the other hand, LDPC coding is used to ensure communication reliability under a 90% packet loss rate.

[0070] In this embodiment, the system adopts a distributed TDMA communication protocol, and has a self-organizing network time slot allocation algorithm, a 50ms-level dynamic topology update, and packet fusion of tightly coupled positioning information of GPS + IMU + wheel speed.

[0071] Specifically, the perception (communication loop) shares the laser point cloud and visual recognition results through V2X; the vehicle-to-vehicle positioning error compensation adopts the cooperative SLAM algorithm.

[0072] In this embodiment, the V2X information interaction on a sightseeing vehicle broadcasts the status of the vehicle every 100ms; at the same time, it receives the real-time status data of adjacent vehicles (within a radius of 50m).

[0073] The status of the vehicle includes: tightly coupled positioning coordinates (GNSS + IMU + wheel speed odometer), current driving intention (going straight / turning / parking), and remaining battery life (affecting the right-of-way priority).

[0074] The network topology maintenance is used to maintain the reliability of vehicle-to-vehicle communication.

[0075] In this embodiment, a fixed communication device is provided at the turning of the passing road. A network communication device and an online monitoring device are provided on the fixed communication device. The network communication device is used for network data interaction between the vehicle and the fixed communication device, and the online monitoring device is used for image acquisition and safety monitoring at the turning of the passing road.

[0076] Further, the fixed communication devices can be laid successively at a certain interval according to the usage requirements, providing safety guarantee for the image acquisition of the sightseeing vehicle's autonomous driving and the communication data.

[0077] It facilitates data communication between vehicles, providing stable communication support for vehicle group collaborative autonomous driving and avoidance.

[0078] Further, the calculation of the passing priority score is as follows:

[0079] P riority = 0.4Q passenger + 0.3E battery + 0.3U urgency .

[0080] The system adopts a dynamic road right allocation mechanism; multi-objective optimization based on passenger capacity, battery remaining capacity, and task urgency; uses a Bayesian game model to predict conflict resolution solutions; supports a blockchain recording system for temporary road right transfer.

[0081] Specifically, the meanings of Q, E, U, and P are as follows:

[0082] Q: Passenger capacity coefficient; the current passenger number (or passenger occupancy rate) of the vehicle; for example: empty vehicle: 0.0; full vehicle: 1.0;

[0083] E: Battery remaining capacity coefficient; the percentage of the remaining battery power in the total battery capacity; for example: 80% battery power → 0.8;

[0084] U: Task urgency coefficient; the priority weight assigned according to the vehicle task; for example: regular patrol: 0.2; emergency transfer: 1.0;

[0085] P: Comprehensive priority score; the finally calculated road right priority value (the larger the value, the higher the priority).

[0086] 0.4Q represents the priority of operation efficiency: the passenger capacity directly affects the passenger experience, and full vehicles need to pass first to reduce the waiting time of the group;

[0087] 0.3E represents the priority of battery life safety: low-battery vehicles need to pass first to avoid stalling midway and at the same time balance the overall battery life of the vehicle fleet;

[0088] 0.3U represents the priority of task flexibility: urgent tasks (such as rapid rescue, VIP pick-up) need to dynamically adjust the road right allocation.

[0089] Calculation of Q value: actual number of passengers / maximum passenger capacity (for example, a vehicle with a capacity of 8 passengers carrying 6 passengers → Q = 0.75);

[0090] Calculation of E value: remaining battery power / full battery capacity (directly mapped by SOC status);

[0091] Definition of U value: preset task level table (for example: 0.0 - 0.3 routine / 0.4 - 0.6 important / 0.7 - 1.0 urgent).

[0092] In the flat road mode:

[0093] Assume that two vehicles conflict at an intersection:

[0094] Vehicle A, passenger load factor 90% (Q = 0.9), battery power 30% (E = 0.3), performing a routine task (U = 0.2);

[0095] P A = 0.4×0.9 + 0.3×0.3 + 0.3×0.2 = 0.36 + 0.09 + 0.06 = 0.51;

[0096] Vehicle B: passenger load factor 40% (Q = 0.4), battery power 80% (E = 0.8), performing an emergency repair task (U = 0.9);

[0097] P B = 0.4×0.4 + 0.3×0.8 + 0.3×0.9 = 0.16 + 0.24 + 0.27 = 0.67;

[0098] It can be seen that P B > P A ; The decision result is: Vehicle B (P = 0.67) obtains the right of way.

[0099] Finally, by quantifying the three core elements of operation efficiency, energy management and task flexibility, the fairness and efficiency balance of multi - vehicle collaborative obstacle avoidance in a closed scenario is achieved; the optimal passing and obstacle avoidance strategies are provided.

[0100] Furthermore, please combine Figure 1 and Figure 4 In the dynamic road - right game, on the one hand, the potential conflicts within 3 seconds are predicted through the Bayesian game model, and on the other hand, the road - right transfer agreement is recorded through the blockchain.

[0101] Furthermore, the path planning module generates candidate paths based on the improved artificial potential field method:

[0102]

[0103] Among them, F_total: Total Force represents the total potential field force, which is the vector sum of all virtual forces acting on the vehicle in the environment and determines the movement direction and acceleration of the vehicle;

[0104] F_attract: Attractive Force represents the attractive force that guides the vehicle towards the target point and is usually proportional to the distance from the target point;

[0105] F_repel^obstacle: Repulsive Force(Obstacle) represents the repulsive force from obstacles, which prevents the vehicle from colliding with static or dynamic obstacles and is inversely proportional to the distance from the obstacles;

[0106] F_repel^vehicle: Repulsive Force(Vehicle) represents the repulsive force between vehicles, which prevents the vehicle from colliding with other vehicles and is calculated in real time based on V2X communication.

[0107] The calculation method of F_attract is as Figure 8 shown;

[0108] Among them, k attract : attraction coefficient; P goal : target point coordinates; P current : current vehicle position; it can be seen that the attractive force decreases linearly as the vehicle approaches the target point.

[0109] The calculation method of F_repel^obstacle is as Figure 9 shown:

[0110] Among them, k_repel: repulsive force coefficient (adjustable parameter); d: distance between the vehicle and the obstacle; d_0: obstacle influence range (threshold); it can be seen that when the vehicle enters the obstacle influence range (d ≤ d_0), the repulsive force increases sharply.

[0111] The calculation method of F_repel^vehicle is as Figure 10 shown;

[0112] Among them, k_vehicle: repulsive force coefficient between vehicles (adjustable parameter); d_vehicle: distance between two vehicles; d_safe: safe following distance (dynamically adjusted and related to speed); it can be seen that it is calculated in real time based on V2X communication to prevent collisions between vehicles in the convoy.

[0113] Scenario application:

[0114] The vehicle is in the following environment:

[0115] Target point: 10 meters directly ahead;

[0116] Obstacle: There is a tree 3 meters to the right.

[0117] Other vehicle: There is an oncoming vehicle 5 meters to the left.

[0118] Calculation process:

[0119] Calculate the attractive force (F_attract): It points to the target point and has a magnitude of k_attract×10.

[0120] Calculate the repulsive force of the obstacle (F_repel^obstacle): It points to the left and has a magnitude of k_repel×(1 / 3 - 1 / d0).

[0121] Calculate the repulsive force between vehicles (F_repel^vehicle): It points to the right and has a magnitude of k_vehicle×(1 / 5 - 1 / d_safe).

[0122] Synthesize the total potential field force (F_total): Adjust the driving direction according to the vector superposition result.

[0123] According to the calculation results, the vehicle finally deviates to the left, avoiding obstacles and oncoming vehicles, and at the same time continues to move forward towards the target point.

[0124] By improving the artificial potential field method, the system can efficiently generate safe and smooth candidate paths, which are suitable for the autonomous driving requirements in complex low-speed scenarios such as scenic spots and industrial parks.

[0125] Furthermore, as Figure 4 shown, the path planning module applies the geese following model to optimize the overall energy consumption of the vehicle fleet.

[0126] In this embodiment, the obstacle avoidance algorithm adopts the geese cooperation model and the multi-modal avoidance strategy library;

[0127] The geese cooperation model integrates the leading vehicle dynamic election mechanism, the hydrodynamic following algorithm, and the electromagnetic field type virtual force field (the leading vehicle adopts the path with the optimal energy consumption, and the following vehicle utilizes the aerodynamic wake of the preceding vehicle);

[0128] The multi-modal avoidance strategy library integrates the progressive lane offset (for oncoming vehicles), the queue reorganization strategy (for overtaking in the same direction), and the safety island emergency parking protocol (for sudden obstacles).

[0129] Optimize the following vehicle scheme to reduce the overall energy consumption in the state of vehicle fleet following.

[0130] Furthermore, the feedback optimization layer includes real-time parameter calibration, which dynamically adjusts the repulsive force coefficient of the virtual force field and the following distance based on the actual obstacle avoidance effect; the real-time parameter calibration also updates online through a deep learning model (using a federated learning framework to protect privacy data; synchronously updating the new version of the obstacle classifier every 24 hours).

[0131] Among them, the following can be dynamically adjusted based on the actual obstacle avoidance effect:

[0132] The adjustable range of the repulsive force parameter of the virtual force field is between 0.8 and 1.2;

[0133] The adjustable range of the following distance is between 1.5 and 3 s.

[0134] The online update of the deep learning model uses a federated learning framework to protect privacy data; synchronize the new version of the obstacle classifier every 24 hours.

[0135] Furthermore, the feedback optimization layer also includes scenario library simulation enhancement. The feedback optimization layer conducts stress tests through a digital twin platform; and records special avoidance cases in actual application scenarios.

[0136] Adaptive optimization and addition of special scenario events in the scenario library. After a special scenario event occurs in one vehicle, all scenario emergency data of the sightseeing vehicles in the vehicle group can be synchronously updated.

[0137] The decision-making delay chain provided by this application:

[0138] Perception (10 ms) → Communication (15 ms) → Decision-making (20 ms) → Control (5 ms) = total delay ≤ 50 ms;

[0139] Obstacle avoidance success rate:

[0140] Static obstacle: 100% (speed < 15 km / h);

[0141] Dynamic obstacle: 99.3% (relative speed of 20 km / h).

[0142] Compared with the single-vehicle avoidance method, the relative energy consumption is reduced by at least 18% through the collaborative avoidance method.

[0143] As Figure 6 shown, the closed-loop control process from environmental perception to decision execution, and at the same time, the system continuously evolves through the feedback optimization layer to meet the real-time (end-to-end delay < 50 ms) and reliability (99.99% availability) requirements in complex scenic area scenarios.

[0144] Ultimately, through the closed-loop architecture of "perception-communication-decision-execution-optimization", intelligent collaborative obstacle avoidance of vehicle groups in complex scenarios is achieved, especially in typical scenarios such as meeting vehicles at intersections and following driving on continuous curves, showing significant advantages.

[0145] The working principle of the obstacle avoidance system for the autonomous driving electric sightseeing vehicle provided in this embodiment is as follows:

[0146] like Figure 7 As shown in the figure, the principle of switching the sightseeing car avoidance mode is:

[0147] Before the sightseeing car drives automatically, the sightseeing car collects the current road inclination in real time through the environmental perception layer;

[0148] Based on the collected road surface data, the system automatically identifies and determines whether the road surface is on a steep slope;

[0149] When the judgment result is a steep slope, the system automatically identifies and switches the traffic avoidance mode to the uphill and downhill avoidance mode, giving priority to uphill vehicles (when there is an emergency vehicle among the passing vehicles, the emergency vehicle enjoys a higher level of priority, and other vehicles give way to the vehicle with the higher level of priority).

[0150] When the judgment result is that the road is not a steep slope, the system automatically identifies and automatically switches the traffic avoidance mode to the flat surface avoidance mode, and selects vehicles with higher priority scores to pass first;

[0151] As the vehicle continues to drive, the inclination of the road surface is continuously collected, and the sightseeing car avoidance mode is adaptively switched according to the inclination of the road surface.

[0152] The principle of obstacle avoidance when the sightseeing car turns:

[0153] During the automatic driving of the sightseeing car, the time nodes in the turning area are predicted and planned in advance;

[0154] The system automatically matches sightseeing buses within the same time node;

[0155] When there are two sightseeing cars driving in opposite directions and both are in the turning area at the same time node, the traffic priority of the two sightseeing cars is determined, and the latter (the sightseeing car with low priority) is controlled to slow down and avoid according to the traffic priority, so that the latter and the former (the sightseeing car with high priority) are staggered in the time nodes of the turning area, while maintaining the normal driving speed of the front car, providing safety guarantee for the automatic driving of the front car;

[0156] Ultimately, the autonomous sightseeing car's curve obstacle avoidance can be predicted in advance and decelerated in advance, minimizing the simultaneous deceleration of two vehicles driving in opposite directions, ensuring traffic efficiency while improving the safety and reliability of system operation.

[0157] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structural transformation made under the concept of the present invention by using the content of the specification and drawings of the present invention, or directly / indirectly applied to other related technical fields, is included in the patent protection scope of the present invention.

Claims

1. An obstacle avoidance system for an autonomous electric sightseeing vehicle, characterized in that: include: Environmental perception layer, the environmental perception layer includes multi-sensor synchronous acquisition and dynamic obstacle tracking, the multi-sensor synchronous acquisition is used to collect environmental data of the sightseeing car, and the dynamic obstacle tracking is used for environmental data fusion, trajectory prediction and obstacle analysis; A vehicle group communication layer, which is used for real-time communication of position data between vehicles; A collaborative decision-making layer, which includes a dynamic road right game and a path planning module. The dynamic road right game is used to analyze the traffic priority of the vehicle, and the path planning module provides route planning for the vehicle according to the traffic priority; An execution control layer, the execution control layer includes motion control, the motion control includes wire-controlled steering and power distribution, the wire-controlled steering uses fuzzy PID to control lateral deviation and active rear wheel steering to compensate for vehicle body posture; A feedback optimization layer, which is used to calibrate and enhance the parameters of the sightseeing car obstacle avoidance database; Among them, based on the analysis of the road conditions according to the collected data, when the vehicle is in an uphill or downhill road environment, the sightseeing car approaching the curve waiting area and going uphill has the right of priority; when the vehicle is in a horizontal road environment, the sightseeing car approaching the curve waiting area has the right of priority.

2. The obstacle avoidance system for an autonomous driving electric sightseeing vehicle according to claim 1, characterized in that: The multi-sensor synchronous acquisition includes a laser radar array, a dual-mode camera and a millimeter-wave radar; the laser radar array constructs a 360° point cloud map in real time; the dual-mode camera identifies the attributes of obstacles within 20m; The millimeter wave radar monitors the relative speed of moving targets within 50m.

3. The obstacle avoidance system for an autonomous driving electric sightseeing vehicle according to claim 2, characterized in that: The dynamic obstacle tracking method fuses multi-source data through Kalman filtering; after fusion, a time-space-related obstacle trajectory prediction model is established; and special obstacles are marked according to the collected obstacle information.

4. The obstacle avoidance system for an autonomous driving electric sightseeing vehicle according to claim 3, characterized in that: The vehicle group communication layer includes V2X information interaction and network topology maintenance. The V2X information interaction is used for vehicle-to-vehicle data interaction. The network topology maintenance dynamically allocates communication time slots based on the TDMA protocol on the one hand, and adopts LDPC coding on the other hand to ensure communication reliability under a 90% packet loss rate.

5. The obstacle avoidance system for the automatic driving electric sightseeing vehicle according to claim 4, characterized in that: Calculation of the traffic priority score: P riority =0.4Q passenger +0.3E battery +0.3U urgency 。 6. The obstacle avoidance system for the automatic driving electric sightseeing vehicle according to claim 5, characterized in that: The dynamic road right game predicts potential conflicts within 3 seconds through the Bayesian game model on the one hand, and records the road right transfer agreement through the blockchain on the other hand.

7. The obstacle avoidance system for an autonomous driving electric sightseeing vehicle according to claim 6, characterized in that: The path planning module generates candidate paths based on the improved artificial potential field method:

8. The obstacle avoidance system for an autonomous driving electric sightseeing vehicle according to claim 7, characterized in that: The path planning module applies the goose following model to optimize the overall energy consumption of the fleet.

9. The obstacle avoidance system for an autonomous driving electric sightseeing vehicle according to claim 8, characterized in that: The feedback optimization layer includes real-time parameter calibration, which dynamically adjusts the virtual force field repulsion coefficient and the following distance based on the actual obstacle avoidance effect; the real-time parameter calibration is also updated online through a deep learning model (using a federated learning framework to protect privacy data; a new version of the obstacle classifier is synchronously updated every 24 hours).

10. The obstacle avoidance system for an autonomous driving electric sightseeing vehicle according to claim 9, characterized in that: The feedback optimization layer also includes scenario library simulation enhancement. The feedback optimization layer is stress-tested through a digital twin platform and special avoidance cases are recorded in actual application scenarios.

Images