A path planning system based on driverless new energy vehicles

Abstract

The present invention discloses a path planning system based on driverless new energy vehicles, which relates to the field of driverless technology and includes: an environment construction unit that integrates the real-time perception data of the vehicle surrounding environment by the sensor module to construct a dynamic three-dimensional road condition model; an obstacle detection module that detects pedestrians, vehicles, and road obstacles in real time and continuously updates the position and speed information; an adaptive data acquisition model that adjusts the data acquisition method based on the road condition complexity in the dynamic three-dimensional road condition model; a local path generation unit that generates multiple candidate trajectories that meet the vehicle movement restrictions according to the current local map, evaluates the safety and efficiency of each trajectory within the time window, and optimally selects the best path. Based on the collaborative perception strategy of vehicle networking communication and multi-sensor data fusion, it improves the vehicle's cognitive and decision-making abilities for the surrounding environment in high traffic flow and complex road conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of driverless technology, and specifically to a path planning system for new energy vehicle driverless driving. Background Art

[0002] The path planning technology of driverless vehicles usually adopts a hierarchical architecture, splitting the entire driving process into multiple stages, each stage having a dedicated algorithm and processing module. This design not only ensures the optimization of global goals but also takes into account the requirements of real-time dynamic regulation. Most driverless vehicles adopt a hierarchical system architecture, which includes not only path planning but also modules such as perception, prediction, and control, with close cooperation between layers.

[0003] Publication No. CN112068548B discloses a path planning method for driverless vehicles in special scenarios under 5G environment, relying on a driverless vehicle path planning system based on 5G communication. The path planning system includes driverless vehicles, roadside units, cloud platforms, and 5G networks. The characteristics of fast transmission speed, large amount of transmitted information, and high signal quality can overcome communication problems in special scenarios, realizing efficient, stable, and fast information interaction between driverless vehicles, roadside units, and cloud platforms; through the organic combination of historical information and real-time perception, it effectively reduces the computing resources occupied by the cloud platform during normal operation, ensuring the safety and reachability of driverless driving; the interaction among vehicle, road, and cloud can achieve full-time, all-weather, full-coverage, and high-precision collaborative perception, jointly completing the decision-making of path planning and adjustment, thereby improving vehicle operation efficiency.

[0004] The planning of driving paths includes global path planning and local path planning. The global path generates an overall optimal or approximately optimal route from the starting point to the end point. Global planning relies on a pre-constructed high-precision digital map, road network model, and historical traffic data;

[0005] Local path planning, on the basis of global planning, generates specific trajectories covering a short period according to real-time sensor data (such as lidar, cameras, millimeter-wave radars, etc.). Local planning needs to quickly respond to changes in road conditions to ensure safe obstacle avoidance and smooth driving. Complex road conditions are often accompanied by a large number of dynamic obstacles, such as pedestrians, non-motor vehicles, and other vehicles driving densely. Local path planning needs to collect data from sensors in real time in an extremely short time and make quick judgments and responses to these dynamic targets. In a traffic environment with multiple vehicles and pedestrians participating, the behavior of each traffic participant has a large degree of uncertainty. Local planning not only needs to consider the current environmental state but also predict the interactions that may occur in the next few seconds. Summary of the Invention

[0006] One of the objectives of the present invention is to provide a path planning system based on the driverless of new energy vehicles, which uses the data of multiple transfer stations and real-time perception data for fusion to reduce the data processing volume.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: A path planning system based on the driverless of new energy vehicles, comprising:

[0008] An environment construction unit, which integrates the real-time perception data of the sensor module for the vehicle surrounding environment, constructs a dynamic three-dimensional road condition model, and marks static and dynamic obstacles, lane lines, and traffic sign information;

[0009] An obstacle detection module, which detects pedestrians, vehicles, and road obstacles in real time, and continuously updates the position and speed information;

[0010] An adaptive data acquisition model, which adjusts the data acquisition method based on the complexity of the road conditions in the dynamic three-dimensional road condition model:

[0011] Multi-source sensor acquisition. When the traffic flow is low, rely on the sensor module to perceive the surrounding environment and construct a local map;

[0012] Transfer station data acquisition. In the vehicle high-density scenario, establish a data transfer station based on the information terminal, realize the sharing of road condition data through the information terminal, collect the shared road condition data and verify the shared road condition data through the real-time perception data, and then perform data fusion to generate the road condition fusion data, and construct a local map based on the road condition fusion data;

[0013] A local path generation unit, which generates multiple candidate trajectories that meet the vehicle movement restrictions according to the current local map, evaluates the safety and efficiency of each trajectory within the time window, and selects the optimal path.

[0014] In one or more embodiments of the present invention, the vehicle and the transfer station are realized through a low-latency communication protocol, and a short-distance communication protocol is used to transmit between the vehicle and the data transfer station in a short time. Based on the vehicle, a position coordinate is established, and the coordinate of the data transfer station is mapped within the position coordinate to identify the vehicle position. Based on the current position of the vehicle and the position of the data transfer station, the shared road condition data is split, and the useless data based on the position relationship between the vehicle and the transfer station is removed.

[0015] In one or more embodiments of the present invention, a position coordinate is established and the useless data based on the position relationship between the vehicle and the transfer station is removed according to the position coordinate:

[0016] Based on the current vehicle, a position coordinate is established. The real-time positions of the current vehicle and the transfer station are obtained through GPS. Taking the current vehicle coordinate as the coordinate origin, the coordinate of the transfer station is mapped within the position coordinate;

[0017] The transfer station attaches the acquisition location data to the road condition sharing data. When the current vehicle receives the road condition sharing data from the transfer station, it splits the road condition sharing data;

[0018] Using the positional relationship between the current position of the vehicle and the coordinates of the transfer station, distance and area thresholds are set to retain the road condition sharing data related to the driving route and driving area of the current traffic flow, filter the noise data based on the positional relationship, and reduce the transmission load and computational complexity.

[0019] In one or more embodiments of the present invention, the steps for constructing a dynamic three-dimensional road condition model are as follows:

[0020] Use lidar to obtain continuous point clouds, and through filtering, segmentation, and registration, a three-dimensional road condition model around the vehicle is generated in real time;

[0021] Combined with the vehicle movement information, use simultaneous localization and mapping to update the three-dimensional road condition model;

[0022] For targets that do not change with time, they are labeled and identified through point cloud segmentation and clustering algorithms;

[0023] For vehicles, pedestrians, and temporary obstacles, use object detection and tracking technology to judge the motion state, and update the position and speed information in real time;

[0024] Use the image segmentation technology based on deep learning of camera images to detect lane lines, and combine point cloud data to perform three-dimensional completion and correction of the lane boundaries.

[0025] In one or more embodiments of the present invention, based on the road condition sharing data of the transfer station, the lane line position is determined to achieve dynamic judgment and correction of the lane position:

[0026] The current vehicle relies on its own sensor module to detect the distance from the road boundary or guardrail as the initial reference information to determine the relative position of the vehicle and the surrounding fixed obstacles;

[0027] Obtain the position coordinates of the nearby transfer station. When the current vehicle moves to the transfer station area, a new polling is triggered, and the distance relationship between the current vehicle and the road boundary or guardrail at this time is recorded;

[0028] The detection data of the current vehicle at the transfer station position is compared with the initial perception data;

[0029] Combining the two measurements and the information shared by the transfer station, the relative position of the current vehicle and the planned lane is determined in the coordinates, and the lane where the vehicle is located is determined by comparing the actual distance between the vehicle and the road boundary with the preset lane width model.

[0030] In one or more embodiments of the present invention, the steps for verifying road condition sharing data based on real-time perception data are as follows:

[0031] Synchronize the road condition sharing data and the real-time perception data to the same time window to ensure the consistency of the data source timeliness, and map the time-synchronized road condition sharing data and real-time perception data within the position coordinates based on the current vehicle;

[0032] Extract the key geometric features in the road condition sharing data, determine the corresponding features in the real-time perception data based on the geometric features, and detect the consistency of the geometric features and the corresponding features through matching;

[0033] Set a feature error threshold. If the geometric features and the corresponding features are within the allowable error range, the road condition sharing data is considered valid. If it exceeds the threshold, the road condition sharing data is corrected based on the real-time perception data.

[0034] In one or more embodiments of the present invention, use particle filtering to perform weighted correction on the road condition sharing data based on the actual perception data, calculate the feature variables of the same geometric feature, and evaluate the difference between the road condition sharing data and the real-time perception data through the calculated feature variables.

[0035] In one or more embodiments of the present invention, integrate the corrected road condition sharing data and the real-time perception data through a data fusion algorithm:

[0036] Use their respective confidence levels to perform dynamic weighting during the fusion process, adjust the fusion ratio, and if the error exceeds the tolerance range, correct the road condition sharing data.

[0037] In one or more embodiments of the present invention, the transfer station obtains the control data of the vehicle from the control system, and the current vehicle obtains the vehicle control data of the transfer station to determine the motion state of the transfer station within the time window:

[0038] Set a fixed time window, and the current vehicle performs statistics and analysis on all received transfer station control data;

[0039] For consecutive time instants t1, t2, t3,......, tn, calculate the position change of the transfer station;

[0040] Based on the estimation results within the entire time window, determine the speed and direction of the transfer station within the time window, and obtain the obstacle path generated by the motion of the transfer station within the time window.

[0041] In one or more embodiments of the present invention, combine the obstacle path with the local path planning to generate multiple candidate trajectories that meet the vehicle motion constraints, establish a short-term prediction model, use the obstacle path and the future path of the obstacle as constraint conditions, and generate the optimal path through online optimization.

[0042] Through the above technical solutions, the present invention has the following beneficial effects:

[0043] 1. Based on the collaborative perception strategy of vehicle networking communication and multi-sensor data fusion, it improves the vehicle's cognitive and decision-making abilities regarding the surrounding environment in high traffic volume and complex road conditions, complements the deficiencies in environmental cognition from multiple perspectives. At the same time, it tightly couples the braking system with the path planning system, and by real-time transmitting the braking operation status, it improves the immediate response and accurate prediction of vehicle problem handling.

[0044] 2. Utilizing vehicle networking technology, multiple vehicles and infrastructure form a dynamic and real-time data sharing network, enhancing the integrity and accuracy of environmental perception.

[0045] 3. Taking advantage of the low-latency and short-distance communication protocol in vehicle networking, it ensures that data transmission can be completed between the vehicle and the data transfer station in an extremely short time. Then, with the help of a unified coordinate system, the geographical positions of the vehicle and the transfer station are precisely matched, thereby effectively screening and filtering out irrelevant data, which is particularly effective in high traffic volume and complex road conditions. It can not only expand the perception boundary of a single vehicle but also reduce the interference of redundant information on the basis of ensuring data timeliness.

[0046] 4. Integrating multiple sensor modules and using real-time data collection to construct a dynamic three-dimensional road condition model covering all-round information. This model not only marks static obstacles, lane lines, and traffic signs on the road but also can track dynamic targets, providing accurate and real-time environmental cognition for the vehicle.

[0047] Other features and advantages of the present invention will be described in the subsequent specification, and some of them will become obvious from the specification or be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the written specification and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the path planning system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] The following will disclose multiple embodiments of the present invention. For the sake of clarity, many practical details will be described together in the following narrative. However, it should be understood that these practical details are not used to limit the present invention. That is to say, in some embodiments of the present invention, these practical details are not necessary. And if possible in implementation, the features of different embodiments can be interactively applied.

[0050] Unless otherwise defined, all terms (including technical and scientific terms) used herein have their ordinary meanings, which can be understood by those skilled in the art. Further, the definitions of the above terms in commonly used dictionaries shall be interpreted as being consistent with the meanings in the relevant fields of the present invention. Unless specifically defined otherwise, these terms will not be construed in an idealized or overly formal sense.

[0051] Referring to Figure 1 As shown, the present invention provides a path planning system for new energy vehicle driverless, which can adjust the acquisition method of path planning data according to different road conditions, reduce the algorithm response frequency while not reducing the response accuracy of path planning, and is especially suitable for road conditions with large traffic flow and complex vehicle changes.

[0052] The path planning system includes:

[0053] An environment construction unit, which integrates the real-time perception data of the vehicle surrounding environment by the sensor module, constructs a dynamic three-dimensional road condition model, and marks static and dynamic obstacles, lane lines, and traffic sign information;

[0054] An obstacle detection module, which detects pedestrians, vehicles, and road obstacles in real time, and continuously updates the position and speed information;

[0055] An adaptive data acquisition model, which adjusts the data acquisition method based on the complexity of the road conditions in the dynamic three-dimensional road condition model:

[0056] Multi-source sensor acquisition. When the traffic flow is low, rely on the sensor module to perceive the surrounding environment and construct a local map;

[0057] Transfer station data acquisition. In the high vehicle density scenario, establish a data transfer station based on the information terminal, realize road condition data sharing through the information terminal, collect road condition shared data and verify the road condition shared data through real-time perception data, and then perform data fusion to generate road condition fusion data, and construct a local map based on the road condition fusion data;

[0058] A local path generation unit, which generates multiple candidate trajectories that meet the vehicle movement restrictions according to the current local map, evaluates the safety and efficiency of each trajectory within the time window, and selects the optimal path.

[0059] In an implementable manner, through multiple data acquisition methods, different data acquisition methods can be adopted in different environments to adapt to complex and changeable road conditions. In ordinary or low traffic flow situations, the vehicle mainly relies on its own sensors for data acquisition. However, when the traffic flow increases and the road conditions become complex, the vehicle sensors are affected by factors such as vision occlusion and environmental interference, resulting in incomplete single-vehicle data and prone to misjudgment.

[0060] Among them, the establishment of the data transfer station can effectively make up for the blind spots of bicycle perception. Through the data sharing and fusion of multiple information terminals, a more comprehensive road condition information can be formed. Compared with the real-time perception data of bicycles, the data transfer station can significantly improve the data coverage and accuracy. It is necessary to verify the authenticity and timeliness of the road condition sharing data through real-time perception data to ensure the reliability and safety of path planning.

[0061] In this application, each vehicle is regarded as a transfer station. Through the information interaction between vehicles, a dynamic data network is constructed to achieve multi-vehicle collaborative perception, further improving the comprehensiveness and real-time nature of road condition data, and ensuring the accuracy and efficiency of path planning in complex environments.

[0062] In another implementation, the transfer station can also be set as a mobile phone terminal device. This scenario is applicable to urban areas with dense pedestrians. The mobile phone terminal, as a mobile transfer station, is used to determine the movement state and position of pedestrians and transmit them to the vehicle system in real time through a wireless network, assisting the vehicle to accurately identify the dynamics of pedestrians, optimize the obstacle avoidance strategy, and improve driving safety and traffic efficiency.

[0063] In one embodiment, the vehicle and the transfer station are implemented through a low-latency communication protocol. The short-range communication protocol is used to transfer between the vehicle and the data transfer station in a short time, and based on the vehicle, a position coordinate is established. The data transfer station coordinate is mapped within the position coordinate to identify the vehicle position. Based on the current position of the vehicle and the position of the data transfer station, the road condition sharing data is split to eliminate the useless data based on the position relationship between the vehicle and the transfer station.

[0064] In an implementable manner, a position coordinate is established at the current vehicle position. When the lane is in different forms, it is affected by vehicles in different directions. Therefore, the data of the data transfer station is split to determine the magnitude of the association between the current vehicle position and different data, eliminate the useless data, reduce the volume of data processing, and thus improve the data processing efficiency to ensure the real-time nature and accuracy of path planning.

[0065] When the vehicle is driving in different lanes and moving in different directions, it is affected differently. Exemplarily, when the vehicle is driving in the middle of the road, for the data transfer station, the pedestrian data shared by the roadside transfer station is eliminated, which can greatly reduce the processing burden in terms of data volume and improve the response speed.

[0066] In one embodiment, a position coordinate is established and the useless data based on the position relationship between the vehicle and the transfer station is eliminated according to the position coordinate:

[0067] Establish position coordinates based on the current vehicle, obtain the real-time positions of the current vehicle and the transfer station through GPS, take the coordinates of the current vehicle as the coordinate origin, and map the coordinates of the transfer station within the position coordinates;

[0068] The transfer station attaches the acquisition position data to the road condition sharing data. When the current vehicle receives the road condition sharing data from the transfer station, it splits the road condition sharing data;

[0069] Utilize the positional relationship between the current position of the vehicle and the coordinates of the transfer station, set distance and area thresholds, retain the road condition sharing data related to the current traffic flow driving route and driving area, filter the noise data based on the positional relationship, and reduce the transmission load and computational complexity.

[0070] In an implementable manner, when the transfer station collects or receives road condition data by itself, it splits the road condition data according to the position of the transfer station, marks the direction of the road condition data, and conducts area division. Exemplarily, with the transfer station as the center, the data within a radius of 100 meters is the current area information, and the data beyond this range is considered to have a lower correlation with the current vehicle driving state.

[0071] In another embodiment, the relevance with the vehicle is determined based on the data in the front, back, left, and right directions of the transfer station itself.

[0072] When vehicle A (the current vehicle) receives data from the transfer station, it compares the positional relationship between its own position and the position of transfer station B. Vehicle A only retains the data that falls near its predetermined driving route. For example, a data record from another vehicle, whose acquisition position deviates more than 30 meters from the front driving corridor of vehicle A and is located in the left-turn lane, while vehicle A is in the right-turn lane, is regarded as irrelevant information and is excluded. Finally, when performing data fusion, only the most relevant and valuable information is used to update the environmental model.

[0073] In one embodiment, the steps for constructing a dynamic three-dimensional road condition model are as follows:

[0074] Utilize lidar to obtain continuous point clouds, and through filtering, segmentation, and registration, generate a three-dimensional road condition model around the vehicle in real time;

[0075] Combine vehicle motion information and use simultaneous localization and mapping to update the three-dimensional road condition model;

[0076] For targets that do not change with time, annotate and identify them through point cloud segmentation and clustering algorithms;

[0077] For vehicles, pedestrians, and temporary obstacles, judge the motion state through target detection and tracking technologies, and update the position and speed information in real time;

[0078] Detect lane lines using deep learning-based image segmentation technology on camera images, and perform 3D completion and correction of lane boundaries in combination with point cloud data.

[0079] In an implementable manner, vehicle lines and traffic signs can also use object detection algorithms to identify traffic signs and traffic lights, and through geometric position mapping, mark their accurate positions in the 3D road condition model.

[0080] The constructed 3D road condition model provides a fine-grained environmental input for local path planning, helping the vehicle more accurately predict the road conditions ahead and formulate driving strategies. The environment model after data fusion not only helps to calibrate the current road conditions, but also provides real-time reference for decision-making in emergency situations (such as obstacle avoidance and braking).

[0081] In one embodiment, determine the lane line position based on the road condition sharing data of the transfer station to achieve dynamic judgment and correction of the lane position:

[0082] The current vehicle relies on its own sensor module to detect the distance between itself and the road boundary or guardrail as the initial reference information to determine the relative position of the vehicle and the surrounding fixed obstacles;

[0083] Obtain the position coordinates of the nearby transfer station. When the current vehicle moves to the transfer station area, trigger a new polling and record the distance relationship between the current vehicle and the road boundary or guardrail at this time;

[0084] Compare the detection data of the current vehicle at the transfer station position with the initial perception data;

[0085] Combine the two measurements and the information shared by the transfer station to determine the relative position of the current vehicle and the planned lane in the coordinate system. By comparing the actual distance between the vehicle and the road boundary with the preset lane width model, determine the lane where the vehicle is located.

[0086] In an implementable manner, make full use of the high-precision environmental data shared by the transfer station to supplement the deficiencies of the vehicle's self-perception: Vehicle A obtains the distance relationship with the road boundary or guardrail through its own sensors to provide a preliminary reference for the fuzzy lane line; use low-latency communication to obtain the accurate road and lane information of the transfer station, and map the sensing data to a unified coordinate system; when approaching the transfer station area, measure again and compare with the transfer station data to eliminate noise, so as to accurately determine the vehicle position; finally determine the lane where the vehicle is located to provide accurate information for subsequent path planning and vehicle control.

[0087] The method of complementing lane line data through a transfer station not only enhances the environmental perception under the condition of blurred lane lines, but also improves the accuracy and safety of driving decisions through a comparison and correction mechanism. In another embodiment, the Kalman filter data fusion algorithm is combined to further improve the data correction effect, so as to achieve more accurate autonomous driving path planning.

[0088] In one embodiment, the steps for verifying road condition sharing data based on real-time perception data are as follows:

[0089] Synchronize the road condition sharing data and the real-time perception data to the same time window to ensure the consistency of the data source timeliness, and map the time-synchronized road condition sharing data and real-time perception data within the position coordinates based on the current vehicle.

[0090] Extract the key geometric features in the road condition sharing data, determine the corresponding features in the real-time perception data based on the geometric features, and detect the consistency of the geometric features and the corresponding features through matching.

[0091] Set a feature error threshold. If the geometric features and the corresponding features are within the allowable error range, the road condition sharing data is considered valid. If it exceeds the threshold, the road condition sharing data is corrected based on the real-time perception data.

[0092] In an implementable manner, the geometric features include information such as lane lines, obstacle boundaries, vehicle speed, and vehicle movement direction. Due to the different measurement accuracies of the transfer stations, there will be error variations in the geometric features and the corresponding features. By adjusting the error threshold, the adaptive optimization of the transfer station data can be achieved to ensure that the vehicle can obtain relatively accurate road condition data under various conditions.

[0093] Among them, the road condition sharing data is the data after filtering.

[0094] In one embodiment, the particle filter is used to perform weighted correction on the road condition sharing data based on the actual perception data, calculate the feature variables of the same geometric feature, and evaluate the difference between the road condition sharing data and the real-time perception data through the calculated feature variables.

[0095] In an implementable manner, for the same geometric target (such as a lane line or an obstacle edge), its geometric features are extracted respectively, such as position coordinates, angles, curvatures, etc., to form a feature variable vector.

[0096] That is, the feature obtained through preprocessing in the shared data is x s , and the feature extracted from the real-time perception data is z (usually z has higher accuracy).

[0097] Establish a state space:

[0098] Define the state vector x to represent the geometric feature to be calibrated (x = [a, b, c]), where the variables describe the position and orientation of the geometric feature in space;

[0099] When the target is considered static or has low dynamics, the state transition model can be simply set as the identity mapping, plus a certain amount of process noise (used to capture small vibrations or measurement errors).

[0100] Particle initialization: Generate N particles in the state space , and set a rough distribution range according to the prior information of the shared data x s (uniformly distributed within x s ± Δ);

[0101] When initializing, the weight of each particle is set to .

[0102] For static features or slowly changing features, the state prediction model is

[0103] , where v1 is the process noise, which follows a zero-mean Gaussian distribution.

[0104] Define the measurement model to associate the particle state x i (k) with the real-time perception data , and the measurement model is

[0105] , where h(x) is a function (e.g., direct mapping if both are in the same coordinate system), and v2 is the measurement noise.

[0106] Use the Gaussian function to calculate the measurement likelihood for each particle, that is, the difference between the particle state and the real-time observation data. Update the weights of each particle according to the likelihood and normalize all particles.

[0107] To prevent particle degeneracy (most weights are concentrated on a few particles), when the number of effective particles is lower than the preset threshold, resample the particles, retain the high-weight particles, and generate a new particle set.

[0108] After multiple iterations of the particle filter, the fusion result is obtained. Compare the result with the features extracted from the real-time data, calculate the error metric. If the error is within the preset tolerance range, it means that the shared data and the real-time data are basically consistent in this geometric feature; otherwise, a large error indicates that there is a deviation in the shared data, and its weight is reduced in the decision-making.

[0109] In one embodiment, the corrected road condition shared data and the real-time perception data are integrated through a data fusion algorithm:

[0110] During the fusion process, dynamic weighting is performed using their respective confidence levels to adjust the fusion ratio. If the error exceeds the tolerance range, the road condition sharing data is corrected.

[0111] In an implementable approach, real-time perception data usually has a high signal-to-noise ratio, low-latency updates, and adjustment of the fusion ratio to ensure the accuracy and real-time nature of the data fusion result. However, since some of the road condition sharing data is not time-sensitive, in another embodiment, when performing data fusion, a time decay factor can be used to adjust the weight of the road condition sharing data, causing the contribution of different data to change during the fusion process, and the non-time-sensitive data can be directly fused.

[0112] In one embodiment, a transfer station obtains control data of a vehicle from a control system, and the current vehicle obtains the vehicle control data of the transfer station to determine the motion state of the transfer station within a time window:

[0113] Set a fixed time window, and the current vehicle performs statistics and analysis on all received transfer station control data;

[0114] For consecutive moments t1, t2, t3,......, tn, calculate the position change of the transfer station;

[0115] Based on the estimation results within the entire time window, determine the speed and direction of the transfer station within the time window, and obtain the obstacle path generated by the motion of the transfer station within the time window.

[0116] In this application, the transfer stations all refer to vehicles moving on the road other than the current vehicle, and vehicle A also refers to the current vehicle. When the control data of the transfer station (other vehicle) is transmitted to the current vehicle through the transfer station, the motion trend of the transfer station can be predicted based on the control data. The prediction result is brought into the path to optimize the motion path of the current vehicle.

[0117] In one embodiment, the obstacle path is combined with local path planning to generate multiple candidate trajectories that meet the vehicle motion constraints, a short-term prediction model is established, and the obstacle path and the future path of the obstacle are used as constraint conditions to generate the optimal path through online optimization.

[0118] In an implementable approach, each candidate path not only needs to consider the safety interval from the obstacle model but also evaluate indicators such as the smoothness, energy consumption, and travel time of the path. A multi-objective cost function is used to comprehensively balance these indicators to select the optimal path.

[0119] Although the present invention is disclosed in connection with the above embodiments, it is not intended to limit the present invention. Any person skilled in the art can make various modifications and refinements without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be defined by the appended claims.

Claims

1. A path planning system based on the driverless of new energy vehicles, characterized in that, Including: An environment construction unit that integrates the real-time perception data of the sensor module for the vehicle's surrounding environment to construct a dynamic three-dimensional road condition model; An obstacle detection module that real-time detects pedestrians, vehicles, and road obstacles and continuously updates the position and speed information; An adaptive data acquisition model that adjusts the data acquisition method based on the complexity of the road conditions in the dynamic three-dimensional road condition model: Multi-source sensor acquisition. When the traffic flow is low, rely on the sensor module to perceive the surrounding environment and construct a local map; Relay station data acquisition. In a high-vehicle-density scenario, establish a data relay station based on the information terminal, realize the sharing of road condition data through the information terminal, collect the shared road condition data, and verify and fuse the shared road condition data with the real-time perception data to construct a local map; A local path generation unit that generates multiple candidate trajectories that meet the vehicle's movement restrictions based on the current local map and evaluates the optimal path.

2. The path planning system for new energy vehicle driverless according to claim 1, characterized in that, The communication between the vehicle and the relay station is realized through a low-latency communication protocol, and a short-distance communication protocol is used to transmit between the vehicle and the data relay station in a short time. Based on the vehicle's established position coordinates, the relay station coordinates are mapped within the position coordinates to identify the vehicle's position. Based on the vehicle's current position and the relay station's position, the shared road condition data is split, and the useless data based on the position relationship between the vehicle and the relay station is removed.

3. The path planning system for new energy vehicle driverless according to claim 2, characterized in that, Establish position coordinates and remove the useless data based on the position relationship between the vehicle and the relay station according to the position coordinates: Establish position coordinates based on the current vehicle, obtain the real-time positions of the current vehicle and the relay station through GPS, use the current vehicle coordinates as the coordinate origin, and map the relay station coordinates within the position coordinates; The relay station attaches the acquisition position data to the shared road condition data. When the current vehicle receives the shared road condition data from the relay station, the shared road condition data is split; Utilize the position relationship between the vehicle's current position and the relay station coordinates, set distance and area thresholds, retain the shared road condition data related to the current traffic flow driving route and driving area, filter the noise data based on the position relationship, and reduce the transmission load and computational complexity.

4. A path planning system based on driverless new energy vehicles according to claim 3, characterized in that, The steps to construct a dynamic three-dimensional road condition model are as follows: Use lidar to obtain continuous point clouds, and through filtering, segmentation, and registration, generate a three-dimensional road condition model around the vehicle in real time; Combine the vehicle's movement information and use simultaneous localization and mapping to update the three-dimensional road condition model; For targets that do not change with time, label and identify them through point cloud segmentation and clustering algorithms; For vehicles, pedestrians, and temporary obstacles, judge the motion state through target detection and tracking technology, and update the position and speed information in real time; Use the image segmentation technology based on deep learning of camera images to detect lane lines, and combine point cloud data to perform three-dimensional completion and correction of the lane boundaries.

5. The path planning system for new energy vehicle driverless according to claim 4, characterized in that Determine the lane line position based on the shared road condition data of the relay station to realize the dynamic judgment and correction of the lane position: The current vehicle relies on its own sensor module to detect the distance between itself and the road boundary or guardrail as the initial reference information to determine the relative position between the vehicle and the surrounding fixed obstacles; Obtain the position coordinates of nearby transfer stations. When the current vehicle moves to the transfer station area, trigger a new polling and record the distance relationship between the current vehicle and the road boundary or guardrail at this time; Compare the detection data of the current vehicle at the transfer station position with the initial perception data; Combining the two measurements and the information shared by the transfer station, determine the relative position of the current vehicle and the planned lane in the coordinates. By comparing the distance between the actual vehicle and the road boundary with the preset lane width model, determine the lane where the vehicle is located.

6. The path planning system based on driverless new energy vehicles according to claim 5, characterized in that, The steps for verifying the road condition sharing data based on real-time perception data are as follows: Synchronize the road condition sharing data and the real-time perception data to the same time window to ensure the consistency of the data source timeliness. Map the time-synchronized road condition sharing data and real-time perception data within the position coordinates based on the current vehicle; Extract the key geometric features in the road condition sharing data, determine the corresponding features in the real-time perception data based on the geometric features, and match the consistency of the detected geometric features and the corresponding features; Set a feature error threshold. If the geometric features and the corresponding features are within the allowable error range, the road condition sharing data is considered valid. If it exceeds the threshold, correct the road condition sharing data based on the real-time perception data.

7. A path planning system for new energy vehicle driverless according to claim 6, characterized in that, Use particle filtering to perform weighted correction on the road condition sharing data based on the actual perception data, calculate the feature variables of the same geometric feature, and evaluate the difference between the road condition sharing data and the real-time perception data through the calculated feature variables.

8. A path planning system for new energy vehicle driverless according to claim 7, characterized in that, Integrate the corrected road condition sharing data and the real-time perception data through a data fusion algorithm: Use their respective confidence levels to perform dynamic weighting during the fusion process, adjust the fusion ratio. If the error exceeds the tolerance range, correct the road condition sharing data.

9. The path planning system for new energy vehicle driverless according to claim 8, characterized in that, The transfer station obtains the control data of the vehicle from the control system. The current vehicle obtains the vehicle control data of the transfer station and determines the motion state of the transfer station within the time window; Set a fixed time window, and the current vehicle statistically analyzes all the received transfer station control data; For consecutive time moments t1, t2, t3,......, tn, calculate the position change of the transfer station; Based on the estimation results within the entire time window, determine the speed and direction of the transfer station within the time window, and obtain the obstacle path generated by the motion of the transfer station within the time window.

10. A path planning system for new energy vehicle driverless driving according to claim 9, characterized in that, Combine the obstacle path with the local path planning to generate multiple candidate trajectories that meet the vehicle motion restrictions, establish a short-term prediction model, use the obstacle path and the future path of the obstacle as constraint conditions, and generate the optimal path through online optimization.

Images