A method for active safety obstacle avoidance control of scenic area roaming vehicles

By fusing data from multiple types of sensing sensors and implementing hierarchical obstacle avoidance decision-making, the obstacle avoidance problem of scenic area roaming vehicles in complex scenarios has been solved, achieving accurate detection and safety control, and improving the safety and operational efficiency of scenic area roaming vehicles.

CN123113876APending Publication Date: 2026-11-17SHENZHEN JIUYOU ZHIXING TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611408970.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-11
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing obstacle avoidance control methods for scenic area roaming vehicles suffer from insufficient detection accuracy, high misjudgment rate, and delayed response in scenarios with dense crowds, diverse and dynamically changing obstacles. Furthermore, they cannot distinguish between obstacle types and movement states, leading to frequent sudden braking that affects ride comfort or untimely response. In addition, they do not take into account the operational boundaries of the scenic area's electronic fence, posing a risk of exceeding the designated area, making it difficult to balance safety and operational efficiency.

Method used

By employing data fusion processing from multiple types of sensing sensors, a unified benchmark sensing dataset is obtained through time synchronization and spatial registration. This dataset is then used to extract and classify obstacle targets, predict motion trajectories, and classify collision risk levels. A hierarchical obstacle avoidance decision-making mechanism is constructed, and path constraints are applied in conjunction with the electronic fence operation boundary to achieve precise obstacle avoidance control.

Benefits of technology

It improves the reliability of obstacle detection and the accuracy of obstacle avoidance perception, takes into account both ride comfort and driving safety, ensures that the obstacle avoidance process is carried out within the legal operating area, reduces operational risks, and improves the safety and operational efficiency of the scenic area roaming vehicle.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113876A_ABST
    Figure CN123113876A_ABST
Patent Text Reader

Abstract

This invention discloses an active safety obstacle avoidance control method for scenic area tour vehicles, comprising the following steps: multi-source sensor data acquisition and spatiotemporal registration processing; obstacle target detection, extraction, and attribute classification; obstacle trajectory prediction and collision risk level classification; construction of a hierarchical obstacle avoidance decision-making mechanism and strategy matching; target deceleration calculation and control output in active braking mode; path planning and steering tracking control in active obstacle avoidance mode; and obstacle avoidance execution status feedback and operational boundary constraint verification. This invention achieves accurate detection and identification of obstacles in complex scenic areas through multi-source sensor data fusion processing, improving the reliability of obstacle avoidance perception. The hierarchical obstacle avoidance decision-making approach matches corresponding control strategies according to collision risk levels, balancing passenger comfort and driving safety.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of active safety control technology for intelligent sightseeing vehicles in scenic areas, specifically to an active safety obstacle avoidance control method for sightseeing vehicles. Background Technology

[0002] As a core vehicle in cultural tourism scenarios, scenic area shuttle buses serve functions such as tourist shuttle and attraction guidance. Their operational safety directly impacts the tourist experience and the operational efficiency of the scenic area. Current obstacle avoidance control for scenic area shuttle buses mostly employs single ultrasonic radar or ordinary camera solutions. In scenarios with dense crowds and diverse, dynamically changing obstacles, these solutions suffer from insufficient detection accuracy, high false alarm rates, and delayed response times. Some solutions use simple fixed threshold braking methods, which cannot distinguish between obstacle types and movement states, easily leading to frequent sudden braking that affects passenger comfort or delayed response to unexpected pedestrians. Furthermore, existing obstacle avoidance methods do not incorporate constraints based on the scenic area's electronic fence operating boundaries, posing a risk of vehicles exceeding designated operating areas during obstacle avoidance. They also lack a layered and graded obstacle avoidance control system adapted to the low-speed conditions of scenic areas, making it difficult to balance safety and operational efficiency. Summary of the Invention

[0003] The purpose of this invention is to provide an active safety obstacle avoidance control method for scenic area roaming vehicles, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for active safety obstacle avoidance control of a scenic area roaming vehicle, comprising the following steps: Step 1: Collect operational data from various sensing sensors mounted on the vehicle, perform time synchronization and spatial coordinate registration on the multi-source data, and obtain a sensing dataset under a unified benchmark. Step 2: Extract obstacle targets from the registered perception dataset, and complete the attribute classification and feature parameter extraction of obstacles by matching the multi-source detection results of the same obstacle through data association; Step 3: Predict the trajectory of the obstacle based on its current characteristic parameters, combine it with the vehicle's own trajectory to conduct a collision risk analysis, calculate the corrected collision time, and classify the risk level. Step 4: Construct a hierarchical obstacle avoidance decision-making mechanism, and match the corresponding obstacle avoidance control strategy according to the collision risk level. The obstacle avoidance control strategy includes four types: early warning prompts, gradual deceleration, emergency braking, and active obstacle avoidance. Step 5: When matching a gradual deceleration or emergency braking strategy, calculate the target deceleration for active braking based on the obstacle distance and relative speed, and output braking control commands to the brake-by-wire system; Step 6: When matching the active obstacle avoidance strategy, plan the obstacle avoidance driving path based on the Ackerman steering geometry, calculate the target steering angle, and output steering control commands to the steering system; Step 7: Collect vehicle status data during the obstacle avoidance process in real time and feed it back to the decision-making end. Combine the electronic fence operation boundary to constrain and verify the obstacle avoidance path, and dynamically adjust the control commands.

[0005] Furthermore, in step one, the various perception sensors include surround-view cameras, lidar, ultrasonic radar, and millimeter-wave radar; the time synchronization processing uses the unified clock of the vehicle chassis as a reference, and uses linear interpolation to align the timestamps of sensor data with different sampling frequencies; the spatial coordinate registration transforms the obstacle coordinates in each sensor coordinate system to the vehicle coordinate system through the rotation and translation parameters obtained by calibration, with the vehicle coordinate system having the rear axle center as the origin and the vehicle's forward direction as the positive direction of the horizontal axis.

[0006] Furthermore, in step two, the obstacle attribute classification includes three categories: static obstacles, dynamic pedestrians, and dynamic vehicles; the extracted obstacle feature parameters include longitudinal distance, lateral distance, relative speed, external dimensions, and direction of movement; and the data association is matched and determined based on three dimensions: position overlap, size similarity, and consistency of motion state.

[0007] Furthermore, in step three, the longitudinal distance of the obstacles is calculated using a confidence-weighted fusion method, and the calculation equation is as follows:

[0008] In the formula, The longitudinal distance of the merged obstacles; For the first Longitudinal distance to obstacles measured by sensors; For the first Confidence weights for distance measurements corresponding to different types of sensors; The number of sensor types participating in the fusion.

[0009] Furthermore, in step three, the corrected equation for calculating the collision time is:

[0010] In the formula, The corrected collision time; The relative longitudinal velocity between the vehicle and the obstacle; This is the scene risk correction coefficient, and its value corresponds to the obstacle category. The correction coefficient value for dynamic pedestrians is less than the correction coefficient value for static obstacles.

[0011] Furthermore, in step three, the risk level is divided into four levels: low risk, medium risk, high risk, and extremely high risk. Low risk corresponds to a collision time greater than the safety threshold, medium risk corresponds to a collision time between the safety threshold and the warning threshold, high risk corresponds to a collision time between the warning threshold and the emergency threshold, and extremely high risk corresponds to a collision time less than the emergency threshold.

[0012] Furthermore, in step five, the calculation equation for the target deceleration during active braking is:

[0013] In the formula, The target deceleration for active braking; It provides a preset safe stopping distance margin; in the gradual deceleration mode, the deceleration is output in a gradient increasing manner, while in the emergency braking mode, the target deceleration is output directly.

[0014] Furthermore, in step six, the triggering conditions for active obstacle avoidance include three conditions: the collision risk reaches an extremely high level, there is sufficient avoidance space on the side of the vehicle, and the obstacle avoidance path is within the operating range of the electronic fence; the obstacle avoidance path adopts a double-arc S-shaped path, which is divided into three stages: the entry segment, the holding segment, and the straightening segment.

[0015] Furthermore, in step six, the equation for calculating the obstacle-avoidance target turning angle is:

[0016] In the formula, The target turning angle for the steering wheels; This refers to the vehicle's wheelbase. This is the lateral offset distance required to bypass the obstacle.

[0017] Furthermore, in step seven, the electronic fence boundary constraint verification calculates the distance between the vehicle position and the fence boundary in real time. When the distance reaches the deceleration threshold, the vehicle speed is automatically reduced. When the distance reaches the parking threshold, the obstacle avoidance action is terminated and braking control is switched. The obstacle avoidance execution status feedback adopts a closed-loop control method, which dynamically adjusts the control commands according to the deviation between the actual vehicle state and the target state.

[0018] Compared with the prior art, the beneficial effects of the present invention are: This invention achieves accurate obstacle detection and identification in complex scenic environments through multi-source sensor data fusion processing, improving the reliability of obstacle avoidance perception. Employing a hierarchical obstacle avoidance decision-making approach, it matches corresponding control strategies based on collision risk levels, balancing passenger comfort and driving safety. By constructing an obstacle avoidance control calculation model adapted to low-speed conditions in scenic areas, it enables smooth execution of braking and obstacle avoidance maneuvers, preventing abrupt control actions from impacting the visitor experience. Combining obstacle avoidance path constraints with electronic fence operating boundaries ensures that the obstacle avoidance process always remains within the legal operating area, reducing operational risks in the scenic area. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 This invention provides an active safety obstacle avoidance control method for scenic area roaming vehicles, comprising the following steps: Step 1: Data Acquisition and Spatiotemporal Registration Processing from Multi-Source Sensors Operational data from various sensing sensors mounted on the vehicle are collected, and the multi-source data undergoes time synchronization and spatial coordinate registration to obtain a sensing dataset under a unified benchmark. The core function of this step is to eliminate the differences between different sensors in the temporal and spatial dimensions, providing standardized data input for subsequent fusion processing.

[0022] In the data acquisition phase, surround-view cameras are responsible for collecting image data around the vehicle, covering a large visual scene around the vehicle body, and are used for obstacle category recognition and contour extraction; LiDAR is responsible for collecting point cloud data around the vehicle, and obtains high-precision obstacle position and size information through laser ranging; ultrasonic radar is responsible for collecting distance data of obstacles close to the vehicle body, focusing on covering low obstacles and blind spots around the vehicle body; millimeter-wave radar is responsible for collecting speed information of dynamic obstacles, and can stably obtain relative speed data in complex weather conditions. The four types of sensors cover different detection distances, different environmental adaptability, and different information dimensions, and work together to form a comprehensive perception system.

[0023] In the time synchronization process, the unified clock of the vehicle chassis control system is used as the global time reference, and all sensor data frames are marked with a global timestamp corresponding to the given moment. Due to differences in sampling frequencies among different sensors, the time nodes of the data output cannot be perfectly aligned. To address this issue, a linear interpolation time registration method is adopted. Using a set unified sampling period as a reference, the equivalent measurement value at the reference time is calculated by linearly interpolating the measurement data of each sensor at non-reference moments between two consecutive frames. Ultimately, the data from all sensors are aligned to the same time node, ensuring consistency in the time dimension. Time synchronization processing avoids obstacle position prediction errors caused by time deviations, improving the accuracy of subsequent fusion processing.

[0024] In the spatial coordinate registration process, a unified vehicle coordinate system is first established, with the rear axle center as the origin, the vehicle's forward direction as the positive X-axis, the horizontal direction to the left of the vehicle as the positive Y-axis, and the vertical upward direction as the positive Z-axis. Each type of sensor has its own independent installation coordinate system, with the origin located at the sensor's installation position, and the coordinate axes aligned with the sensor's measurement reference direction. The rotation and translation parameters of each type of sensor relative to the vehicle coordinate system are pre-calibrated. The rotation parameters include rotation angles around the three coordinate axes, and the translation parameters include offset distances along the three coordinate axes. Using rotation matrices and translation vectors, obstacle coordinates in the sensor coordinate system are transformed to the vehicle coordinate system. The transformation process sequentially performs rotation and translation transformations, ultimately unifying the position data of all obstacles into the vehicle coordinate system.

[0025] During spatial registration, data from overlapping detection areas are prioritized for fusion, taking into account the differences in field of view of different sensors, while data from non-overlapping areas are processed separately to ensure comprehensive perception coverage around the entire vehicle body. For multiple similar sensors installed at different locations on the vehicle body, such as ultrasonic radars distributed around the vehicle, coordinate transformations are performed separately and then unified into the vehicle coordinate system to form a complete peripheral distance detection array.

[0026] Step 2: Obstacle Target Detection, Extraction, and Attribute Classification Obstacle targets are extracted from the registered perception dataset. By matching multi-source detection results of the same obstacle through data association, obstacle attribute classification and feature parameter extraction are completed. The core function of this step is to filter out effective obstacle targets from complex environmental data and clarify their type and motion characteristics, providing basic parameters for risk assessment.

[0027] In the obstacle target extraction stage, corresponding processing methods are adopted for different types of sensor data. For LiDAR point cloud data, ground point filtering is first performed. By fitting the ground plane equation, point cloud data belonging to the ground are removed to avoid interference from ground points in obstacle detection. Then, point cloud denoising is performed to remove discrete isolated points and noise points. Finally, Euclidean clustering is used to divide point clouds with close spatial distances into the same cluster. Each cluster corresponds to an obstacle target. Information such as the center coordinates, maximum and minimum coordinate ranges, and number of point clouds of each cluster are extracted, and then the dimensional parameters of the obstacle, such as longitudinal distance, lateral distance, length, width, and height, are calculated.

[0028] For camera image data, obstacle detection is performed using image processing methods. This involves edge extraction and feature matching to identify obstacle targets in the image, obtaining the obstacle's bounding box position and category information. By combining the camera's intrinsic and extrinsic parameter matrices, the pixel coordinates in the image can be converted into three-dimensional coordinates in the vehicle's coordinate system, thus obtaining the obstacle's position information.

[0029] For ultrasonic radar data, based on the installation location and measurement distance of each radar, it is directly converted into obstacle distance points in the vehicle coordinate system to supplement obstacle detection in close-range areas, especially in the area near the bottom of the vehicle where lidar and cameras have blind spots.

[0030] For millimeter-wave radar data, the relative distance and relative velocity of each target are extracted and converted into position and velocity information in the vehicle coordinate system, which is mainly used for the speed detection of dynamic obstacles.

[0031] In the data association stage, obstacle targets detected by different sensors are matched to determine whether they belong to the same physical target. Association determination is based on three dimensions: first, positional overlap, comparing the deviation of the center positions of two targets in the vehicle coordinate system; the smaller the deviation, the higher the probability that they belong to the same target; second, size similarity, comparing the differences in the external dimensions of two targets; the closer the dimensions, the higher the probability that they belong to the same target; and third, consistency of motion state, for dynamic targets, comparing the magnitude and direction of their velocities; the more consistent the motion states, the higher the probability that they belong to the same target. When the overall matching degree of the three dimensions reaches a set threshold, the two targets are determined to be the same obstacle, and the multi-source data is fused; targets with a matching degree below the threshold are treated as independent targets.

[0032] In the attribute classification stage, obstacles are divided into three categories based on their size, movement, and location characteristics: static obstacles, dynamic pedestrians, and dynamic vehicles. Static obstacles include fixed objects such as trees, railings, signs, and landscape facilities within the scenic area, which move at zero speed and vary widely in size. Dynamic pedestrians include tourists walking within the scenic area, whose size is relatively small, their movement speed is low, and their movement direction is highly random. Dynamic vehicles include other roaming vehicles and service vehicles within the scenic area, whose size is relatively large, their movement speed is high, and their movement direction is relatively stable.

[0033] After classification, complete feature parameters of each obstacle are extracted, including longitudinal distance, lateral distance, relative longitudinal velocity, relative lateral velocity, length, width, height, and motion direction angle. All parameters are unified under the vehicle coordinate system to form a standardized obstacle feature set.

[0034] Step 3: Obstacle trajectory prediction and collision risk level classification The obstacle's trajectory is predicted based on its current characteristic parameters. Collision risk analysis is then performed by combining this with the vehicle's own trajectory, calculating the corrected collision time, and classifying the risk level. The core function of this step is to quantitatively assess the collision hazard level of the obstacle, providing a quantitative basis for subsequent obstacle avoidance decisions.

[0035] In the trajectory prediction stage, different prediction methods are used for different types of obstacles. For static obstacles, their positions remain fixed, and their future positions are consistent with their current positions, resulting in a fixed trajectory. For dynamic obstacles, a uniform motion model is used for prediction. Based on the current position coordinates and velocity components, the position coordinates at multiple future moments are calculated to form a continuous predicted trajectory. The prediction duration is set according to the vehicle's speed, covering the time range required for obstacle avoidance decisions. For dynamic pedestrians, the prediction duration is appropriately shortened, and the trajectory update frequency is increased to accommodate the randomness of pedestrian movement; for dynamic vehicles, the prediction duration can be appropriately extended, and the update frequency reduced.

[0036] Simultaneously, based on the vehicle's current speed and steering angle, and combined with the vehicle's kinematic model, the vehicle's future trajectory is calculated. The trajectory prediction takes into account Ackermann steering characteristics to ensure that the trajectory conforms to the vehicle's actual motion.

[0037] In the collision risk analysis phase, the predicted trajectory of the obstacle is spatially compared with the predicted trajectory of the vehicle itself to determine whether the two trajectories intersect. If there is no intersection, it is determined that there is no collision risk; if there is an intersection, the time required for the vehicle to reach the collision point is calculated to obtain the original collision time.

[0038] To improve the accuracy of risk assessment, a confidence-weighted fusion method is used to calculate the fused longitudinal distance of obstacles. The calculation equation is as follows:

[0039] In the formula, The longitudinal distance of the merged obstacles; For the first Longitudinal distance to obstacles measured by sensors; For the first Confidence weights for distance measurements corresponding to different types of sensors; This represents the number of sensor types participating in the fusion. The confidence weights are set based on the ranging accuracy and applicable scenarios of different sensors. LiDAR has the highest confidence for mid-to-long-range ranging, ultrasonic radar has relatively high confidence for short-range ranging, millimeter-wave radar has relatively high confidence for dynamic target ranging, and camera visual ranging has relatively low confidence. Through weighted fusion, the advantages of different sensors can be combined to improve the overall accuracy of distance measurement.

[0040] In the collision time correction stage, a scenario risk correction coefficient is introduced to correct the original collision time. The formula for calculating the corrected collision time is as follows:

[0041] In the formula, The corrected collision time; The relative longitudinal velocity between the vehicle and the obstacle; This is a scene risk correction coefficient, and its value corresponds to the obstacle category. The correction coefficient for dynamic pedestrians is less than 1, which is equivalent to shortening the collision time judgment value, increasing the risk level, and reflecting the safety principle of pedestrian priority; the correction coefficient for static obstacles is equal to 1, and the judgment is based on the original collision time; the correction coefficient for dynamic vehicles is greater than 1, which is equivalent to extending the collision time judgment value, reducing the probability of misjudgment, because vehicle obstacles usually have autonomous avoidance capabilities.

[0042] In the risk level classification stage, based on the corrected collision time, the collision risk is divided into four levels: low risk level, corresponding to a corrected collision time greater than the safety threshold, where the obstacle is far away and there is no immediate risk of collision; medium risk level, corresponding to a corrected collision time between the safety threshold and the warning threshold, where there is a potential collision risk and deceleration measures are required; high risk level, corresponding to a corrected collision time between the warning threshold and the emergency threshold, where the collision risk is high and emergency braking measures are required; and very high risk level, corresponding to a corrected collision time less than the emergency threshold, where the collision risk is extremely high, the braking distance is insufficient, and active obstacle avoidance needs to be considered.

[0043] Step 4: Construction of a hierarchical obstacle avoidance decision-making mechanism and strategy matching A hierarchical obstacle avoidance decision-making mechanism is established, matching corresponding obstacle avoidance control strategies based on the collision risk level. These strategies include four categories: early warning prompts, gradual deceleration, emergency braking, and active obstacle avoidance. The core function of this step is to employ differentiated control methods based on the degree of risk, minimizing interference with normal driving while ensuring safety, thereby improving passenger comfort and operational efficiency.

[0044] The obstacle avoidance decision-making mechanism follows the principle of safety first. High-priority strategies can override low-priority strategies. When the risk level increases, it automatically switches to a higher-priority control strategy; when the risk level decreases, it gradually downgrades to a lower-priority control strategy to avoid frequent switching of control actions. At the same time, the braking strategy has a higher priority than the obstacle avoidance strategy. When braking can achieve a safe stop, the braking method is used first. The obstacle avoidance method is only used when braking cannot meet the safety requirements and obstacle avoidance conditions are available.

[0045] The specific details of the four obstacle avoidance control strategies are as follows: The first category is the early warning and alert strategy, corresponding to the low-risk level. Under this strategy, the vehicle maintains its current speed and route, without issuing braking or steering control commands. It only displays the location and type of surrounding obstacles on the in-vehicle display interface and provides mild voice prompts to alert the driver to their surroundings. The purpose of this strategy is to enhance the driver's perception while avoiding unnecessary control actions that could affect the smoothness of the ride.

[0046] The second type is the gradual deceleration strategy, corresponding to the medium-risk level. Under this strategy, the vehicle gradually reduces its speed, increasing the safe distance through smooth deceleration and mitigating potential collision risks. The deceleration process uses gradual control, with the deceleration gradually increasing to avoid abrupt braking that could cause passengers to lurch forward, ensuring passenger comfort. Simultaneously, a voice prompt will indicate an obstacle ahead and that the vehicle is decelerating, requesting passengers to hold on tight. Once the obstacle is removed or the risk is reduced, the vehicle gradually resumes normal speed.

[0047] The third category is emergency braking strategy, corresponding to a high-risk level. Under this strategy, the vehicle rapidly increases braking force to achieve sufficient deceleration and ensure a safe stop before the obstacle. The primary goal of braking is safe stopping, while also ensuring braking smoothness to avoid wheel lock-up and braking shock. When emergency braking is triggered, in-vehicle warning prompts are simultaneously activated to remind passengers to prepare for braking. Once the obstacle is removed and a safe distance is restored, the brakes are gradually released, and normal driving resumes.

[0048] The fourth category is active obstacle avoidance strategy, corresponding to an extremely high risk level. Under this strategy, the vehicle maneuvers around obstacles using steering maneuvers to avoid collisions caused by insufficient braking distance. Active obstacle avoidance requires strict triggering conditions; it is only permitted when braking cannot guarantee a safe stop, there is sufficient side clearance, and the obstacle avoidance path is within the operational range of an electronic fence. During obstacle avoidance, the vehicle maintains a stable speed and smooth steering movements, avoiding skidding and significant swaying. After completing the obstacle avoidance maneuver, the vehicle gradually returns to its original driving route.

[0049] The decision-making mechanism also includes obstacle failure judgment rules. If an obstacle is not detected for several consecutive cycles, it is determined that the obstacle has been removed or disappeared, and the corresponding obstacle avoidance control strategy is automatically deactivated, restoring normal driving status. A certain delay is set for failure judgment to avoid frequent switching of control strategies due to brief frame drops from sensors.

[0050] Step 5: Target deceleration calculation and control output in active braking mode When matching progressive deceleration or emergency braking strategies, the target deceleration for active braking is calculated based on the obstacle distance and relative speed, and braking control commands are output to the brake-by-wire system. The core function of this step is to translate obstacle avoidance decisions into specific braking control quantities, achieving precise braking control.

[0051] First, based on the merged longitudinal distance to the obstacle, the relative longitudinal velocity between the vehicle and the obstacle, and the safe stopping distance margin, calculate the target deceleration required for braking. The formula for calculating the target deceleration of active braking is:

[0052] In the formula, The target deceleration for active braking; This is the preset safe stopping distance margin, which is the safe distance that needs to be maintained between the front of the vehicle and the obstacle after the vehicle has come to a complete stop, in order to avoid collisions caused by braking control errors.

[0053] After the target deceleration is calculated, different output methods are adopted according to the braking strategy type. For the progressive deceleration mode, a gradient-increasing output method is used, dividing the target deceleration into multiple gradient levels. The deceleration level is increased by one level at set time intervals until the target deceleration value is reached. The granularity of the gradient division is set according to comfort performance requirements. Too small a granularity will lead to excessively frequent adjustments, while too large a granularity will cause braking shocks. A smooth deceleration process can be achieved by setting the gradient appropriately. During progressive deceleration, changes in obstacle distance and relative velocity are monitored in real time, and the target deceleration is dynamically updated. When the risk decreases, the deceleration is synchronously reduced using the gradient to ensure the smoothness of the control process.

[0054] In emergency braking mode, the calculated target deceleration is directly output, and the brake-by-wire system responds quickly to achieve the target braking intensity within a short time. During emergency braking, the remaining braking distance and the stopping distance at the current speed are calculated in real time, and the braking force is dynamically adjusted to ensure that the vehicle stops smoothly within a safe distance, avoiding over-braking or under-braking. Once the vehicle speed drops to zero and remains stationary, the braking state is maintained until the obstacle risk is eliminated, at which point the braking is released.

[0055] During braking control, a closed-loop control method is adopted, which collects the actual deceleration of the vehicle and the pressure of the brake lines in real time, compares them with the target values, and adjusts the control commands according to the deviation to ensure that the actual braking effect is consistent with the target. At the same time, a deceleration upper limit is set to avoid excessive deceleration that may cause passenger discomfort or loss of vehicle control. The deceleration upper limit in low-speed scenarios in scenic areas is set according to human comfort thresholds, taking into account both safety and comfort.

[0056] When the distance to the obstacle increases, the relative speed decreases, or the obstacle is removed, the braking system gradually reduces the braking force, smoothly releasing the brakes to prevent the vehicle from lurching forward due to a sudden loss of braking force. The brake release process also uses a gradual reduction method to gradually restore normal driving conditions.

[0057] Step Six: Path Planning and Steering Tracking Control in Active Obstacle Avoidance Mode When matching an active obstacle avoidance strategy, the obstacle avoidance path is planned based on Ackermann steering geometry, the target steering angle is calculated, and steering control commands are output to the steering system. The core function of this step is to plan a reasonable obstacle avoidance path while satisfying the vehicle's kinematic characteristics, thereby achieving smooth obstacle avoidance maneuvers.

[0058] First, obstacle avoidance conditions are verified, which includes three aspects: First, collision risk verification confirms that the collision risk has reached an extremely high level and the braking distance is insufficient to guarantee a safe stop. Second, lateral space verification uses sensors to detect the distance to obstacles in the avoidance direction, confirming that there is sufficient lateral space to complete the obstacle avoidance; this lateral space must be greater than the sum of the vehicle width, the obstacle avoidance offset, and the safety margin. Third, operational boundary verification confirms that the planned obstacle avoidance path remains within the operational area of ​​the electronic fence and will not exceed the designated operational zone. Only when all three conditions are met can the active obstacle avoidance strategy be executed.

[0059] The obstacle avoidance path employs a double-circular-arc S-shaped design, divided into three phases: the first phase is the entry phase, where the vehicle gradually turns from its original route, with the lateral offset gradually increasing until it reaches the maximum offset distance; the second phase is the holding phase, where the vehicle maintains the maximum lateral offset distance while traveling in a straight line, completely avoiding the obstacle; the third phase is the straightening phase, where the vehicle gradually straightens its direction, with the lateral offset gradually decreasing, eventually returning to its original route. All three phases utilize circular transitions to ensure continuous changes in path curvature and avoid shocks caused by sudden changes in steering angle.

[0060] Based on the Ackermann steering geometry, and according to the vehicle wheelbase and the lateral offset distance around the obstacle, the turning radius of the obstacle avoidance path and the corresponding target steering angle of the steering wheels are calculated. The equation for calculating the target steering angle around the obstacle is:

[0061] In the formula, The target turning angle for the steering wheels; This refers to the vehicle's wheelbase. This represents the lateral offset distance required to maneuver around the obstacle. This formula is derived based on the geometry of a double-circular-arc path, conforming to the kinematic characteristics of Ackermann steering. It ensures that the wheels roll purely during obstacle maneuvering, reducing tire wear and the risk of sideslip.

[0062] After path planning is completed, the path is converted into steering control commands, and the steering system is controlled to execute them using a preview and tracking method. The preview distance is set according to the current vehicle speed; the higher the vehicle speed, the farther the preview distance, ensuring smooth steering control. During steering execution, the actual steering angle and vehicle yaw rate are collected in real time, compared with the target values, and the steering commands are dynamically adjusted to ensure that the vehicle accurately tracks the planned path.

[0063] During obstacle avoidance, the movement status of the obstacle and changes in the side space are continuously monitored. If the obstacle moves in the avoidance direction, resulting in insufficient side space, or if the obstacle avoidance path is about to exceed the boundary of the electronic fence, the obstacle avoidance action is immediately terminated and the emergency braking strategy is switched to prioritize safety.

[0064] After successfully navigating the obstacle, confirm the original driving route is safe, and gradually straighten the steering wheels to smoothly return the vehicle to its original path and resume normal driving. The straightening process should also be gradual to avoid excessively rapid steering correction that could cause the vehicle to wobble.

[0065] Step 7: Obstacle Avoidance Execution Status Feedback and Operational Boundary Constraint Verification The system collects vehicle status data in real time during obstacle avoidance and feeds it back to the decision-making end. This data, combined with the electronic fence operating boundary, constrains and verifies the obstacle avoidance path, dynamically adjusting control commands. The core function of this step is to form a complete closed-loop control chain, while ensuring that the obstacle avoidance process complies with the scenic area's operating boundary requirements, avoiding the risk of boundary violations.

[0066] In the status feedback phase, the vehicle chassis system collects real-time status data such as vehicle speed, deceleration, steering angle, yaw rate, and position coordinates, and feeds this data back to the obstacle avoidance decision controller via the onboard bus. The controller compares the actual status with the target status, calculates the deviation between the two, and adjusts the output of control commands based on the deviation, forming a closed-loop control. Closed-loop control can eliminate the influence of actuator errors and external interference, improving the accuracy and stability of obstacle avoidance control.

[0067] The frequency of status feedback is matched with the control cycle to ensure real-time control. For braking control, the focus is on feeding back the actual deceleration and braking pressure; for steering control, the focus is on feeding back the actual steering angle and yaw rate. The controller uses a proportional-integral (PI) regulation method, dynamically adjusting the control output based on the deviation to achieve zero steady-state error control.

[0068] In the operational boundary constraint verification stage, the electronic fence boundary data of the scenic area's operational area is pre-stored in the system. The boundary data consists of a series of coordinate points connected to form a closed polygon. Vehicles obtain their own position coordinates in real time through the positioning system, calculate the shortest distance between their current position and the electronic fence boundary, and simultaneously perform a full path verification of the planned obstacle avoidance path to confirm that all points on the path are inside the fence boundary.

[0069] When the distance between the vehicle and the fence boundary reaches the deceleration threshold, the system automatically limits the vehicle's maximum speed, reduces the speed in the boundary area, and improves safety. When the distance reaches the parking threshold, the system prohibits the vehicle from continuing to move towards the boundary. If the vehicle is performing obstacle avoidance maneuvers at this time, the obstacle avoidance will be terminated immediately, and braking control will be switched to ensure that the vehicle stops within the operating area.

[0070] Electronic fence constraints have the highest priority. No obstacle avoidance strategy may exceed the electronic fence boundary. When obstacle avoidance needs conflict with boundary constraints, the boundary constraints shall be satisfied first, and safety shall be ensured through braking to prevent vehicles from going out of the operating area and causing safety accidents and management risks.

[0071] In addition, the system records process data for all obstacle avoidance events, including trigger time, obstacle type, risk level, executed control strategy, and vehicle status changes, for subsequent data analysis and strategy optimization to continuously improve the performance of obstacle avoidance control.

[0072] Example: This embodiment uses a low-speed intelligent roaming vehicle in a scenic area as the application object, and adopts the active safety obstacle avoidance control method proposed in this invention to achieve active safety obstacle avoidance in complex crowd flow scenarios in scenic areas. The specific implementation process is as follows: First, the installation and calibration of multiple sensors were completed. Four types of sensors—surround-view cameras, LiDAR, ultrasonic radar, and millimeter-wave radar—were deployed on the vehicle body. The rotation and translation parameters of each sensor relative to the vehicle's coordinate system were calibrated to ensure spatial registration accuracy. Electronic fence boundary data was then configured and the operating area of ​​the scenic area was entered.

[0073] During normal vehicle operation, Step 1 continuously collects data from four types of sensors, performs time synchronization and spatial registration processing, and outputs a unified benchmark perception dataset. Step 2 performs obstacle detection on the perception data, extracting obstacle targets through point cloud clustering, image processing, distance detection, etc., classifying them after data association and matching, and outputting the feature parameters of each obstacle.

[0074] When a pedestrian appears in front of the vehicle, step three predicts the pedestrian's trajectory and the vehicle's own trajectory, calculates the fused longitudinal distance, and uses a weighted fusion method to combine the distance data from lidar and ultrasonic radar to obtain an accurate obstacle distance. The relative longitudinal velocity is calculated, and a risk correction coefficient corresponding to the pedestrian is introduced to obtain the corrected collision time. The risk level is determined based on the collision time; if it falls within the medium-risk range, a gradual deceleration strategy is applied.

[0075] Step five calculates the target deceleration based on distance, relative speed, and safety margin. Braking commands are output using a gradient-increasing method, causing the vehicle to decelerate smoothly and gradually increase the safe distance from the pedestrian. Simultaneously, a voice prompt informs the tourist that there is a pedestrian ahead and the vehicle is decelerating. Once the pedestrian leaves the vehicle's path, the risk level decreases to low risk, the braking system gradually reduces braking force, and the vehicle gradually returns to normal speed.

[0076] If a pedestrian suddenly runs in front of the vehicle, and the collision time drops drastically to a high-risk range, the emergency braking strategy is immediately switched on, rapidly increasing braking force to achieve the calculated target deceleration and ensure a safe stop in front of the pedestrian. Simultaneously, in-vehicle warnings are triggered to alert passengers.

[0077] If the braking distance is insufficient and there is enough space on the side, and the vehicle is within the electronic fence area, the active obstacle avoidance strategy is triggered. Step six involves planning an S-shaped obstacle avoidance path, calculating the target steering angle based on the wheelbase and offset, controlling the steering system to smoothly steer, and gradually straightening the vehicle after passing the pedestrian, returning to the original driving route.

[0078] Throughout the obstacle avoidance process, step seven provides real-time feedback on the vehicle's status, dynamically adjusts the control parameters, and continuously verifies the electronic fence boundaries to ensure that all actions are completed within the operational area.

[0079] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for active safety obstacle avoidance control of a scenic area tour vehicle, characterized in that: Includes the following steps: Step 1: Collect operational data from various sensing sensors mounted on the vehicle, perform time synchronization and spatial coordinate registration on the multi-source data, and obtain a sensing dataset under a unified benchmark. Step 2: Extract obstacle targets from the registered perception dataset, and complete the attribute classification and feature parameter extraction of obstacles by matching the multi-source detection results of the same obstacle through data association; Step 3: Predict the trajectory of the obstacle based on its current characteristic parameters, combine it with the vehicle's own trajectory to conduct a collision risk analysis, calculate the corrected collision time, and classify the risk level. Step 4: Construct a hierarchical obstacle avoidance decision-making mechanism, and match the corresponding obstacle avoidance control strategy according to the collision risk level. The obstacle avoidance control strategy includes four types: early warning prompts, gradual deceleration, emergency braking, and active obstacle avoidance. Step 5: When matching a gradual deceleration or emergency braking strategy, calculate the target deceleration for active braking based on the obstacle distance and relative speed, and output braking control commands to the brake-by-wire system; Step 6: When matching the active obstacle avoidance strategy, plan the obstacle avoidance driving path based on the Ackerman steering geometry, calculate the target steering angle, and output steering control commands to the steering system; Step 7: Collect vehicle status data during the obstacle avoidance process in real time and feed it back to the decision-making end. Combine the electronic fence operation boundary to constrain and verify the obstacle avoidance path, and dynamically adjust the control commands.

2. The active safety obstacle avoidance control method for a scenic area tour vehicle according to claim 1, characterized in that: In step one, the various perception sensors include surround-view cameras, lidar, ultrasonic radar, and millimeter-wave radar; the time synchronization processing uses the unified clock of the vehicle chassis as a reference, and uses linear interpolation to align the timestamps of sensor data with different sampling frequencies; the spatial coordinate registration transforms the obstacle coordinates in each sensor coordinate system to the vehicle coordinate system through the rotation and translation parameters obtained by calibration. The vehicle coordinate system takes the rear axle center of the vehicle as the origin and the forward direction of the vehicle as the positive direction of the horizontal axis.

3. The active safety obstacle avoidance control method for a scenic area tour vehicle according to claim 1, characterized in that: In step two, the obstacle attribute classification includes three categories: static obstacles, dynamic pedestrians, and dynamic vehicles; the extracted obstacle feature parameters include longitudinal distance, lateral distance, relative speed, external dimensions, and direction of movement. Data association is determined based on three dimensions: location overlap, size similarity, and consistency of motion state.

4. The active safety obstacle avoidance control method for a scenic area tour vehicle according to claim 1, characterized in that: In step three, the longitudinal distance of the obstacles is calculated using a confidence-weighted fusion method, and the calculation equation is as follows: ; In the formula, The longitudinal distance of the merged obstacles; For the first Longitudinal distance to obstacles measured by sensors; For the first Confidence weights for distance measurements corresponding to different types of sensors; The number of sensor types participating in the fusion.

5. The active safety obstacle avoidance control method for a scenic area tour vehicle according to claim 4, characterized in that: In step three, the corrected equation for calculating the collision time is: ; In the formula, The corrected collision time; The relative longitudinal velocity between the vehicle and the obstacle; This is the scene risk correction coefficient, and its value corresponds to the obstacle category. The correction coefficient value for dynamic pedestrians is less than the correction coefficient value for static obstacles.

6. The active safety obstacle avoidance control method for a scenic area tour vehicle according to claim 1, characterized in that: In step three, the risk levels are divided into four levels: low risk, medium risk, high risk, and extremely high risk. Low risk corresponds to a collision time greater than the safety threshold, medium risk corresponds to a collision time between the safety threshold and the warning threshold, high risk corresponds to a collision time between the warning threshold and the emergency threshold, and extremely high risk corresponds to a collision time less than the emergency threshold.

7. The active safety obstacle avoidance control method for a scenic area tour vehicle according to claim 1, characterized in that: In step five, the equation for calculating the target deceleration during active braking is: ; In the formula, The target deceleration for active braking; It provides a preset safe stopping distance margin; in the gradual deceleration mode, the deceleration is output in a gradient increasing manner, while in the emergency braking mode, the target deceleration is output directly.

8. The active safety obstacle avoidance control method for a scenic area roaming vehicle according to claim 1, characterized in that: In step six, the triggering conditions for active obstacle avoidance include three conditions: the collision risk reaches an extremely high level, there is sufficient avoidance space on the side of the vehicle, and the obstacle avoidance path is within the operating range of the electronic fence; the obstacle avoidance path adopts a double-arc S-shaped path, which is divided into three stages: the entry segment, the holding segment, and the straightening segment.

9. The active safety obstacle avoidance control method for a scenic area tour vehicle according to claim 8, characterized in that: In step six, the equation for calculating the turning angle of the obstacle avoidance target is: ; In the formula, The target turning angle for the steering wheels; This refers to the vehicle's wheelbase. This is the lateral offset distance required to bypass the obstacle.

10. The active safety obstacle avoidance control method for a scenic area tour vehicle according to claim 1, characterized in that: In step seven, the electronic fence boundary constraint verification calculates the distance between the vehicle position and the fence boundary in real time. When the distance reaches the deceleration threshold, the vehicle speed is automatically reduced. When the distance reaches the parking threshold, the obstacle avoidance action is terminated and braking control is switched. The obstacle avoidance execution status feedback adopts a closed-loop control method, which dynamically adjusts the control commands according to the deviation between the actual vehicle state and the target state.