Autonomous driving methods, devices, and autonomous vehicles based on obstacle intent

Abstract

This invention provides an autonomous driving method, device, and vehicle based on obstacle intent. The method includes: acquiring local map data corresponding to the vehicle's current position; determining a target obstacle based on the current position, the vehicle's target driving area, and the positional relationship between the obstacle and the target obstacle; performing lane filtering on the local map data based on the current position and the target obstacle's position to determine the target lane for the obstacle; establishing a lane driving path sequence for the target obstacle based on the connectivity between the target lane and other lanes; predicting the target obstacle's intended trajectory based on the lane driving path sequence and collected feature data of the target obstacle; and updating the vehicle's path planning based on the correlation between the intended trajectory and the target driving area. This method makes the predicted trajectory more consistent with real-world complex scenarios, solving the problem of practical application of autonomous driving in complex scenarios and improving the adaptability of autonomous driving to the environment.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and more specifically to autonomous driving methods, devices, and autonomous vehicles based on obstacle intent. Background Technology

[0002] Currently, when mining areas, factory areas, or industrial parks require engineering vehicles (such as dump trucks, mixer trucks, wide-body trucks, semi-trailers, etc.) for transportation operations, they often need to complete the corresponding tasks and pass through designated areas (such as weighing areas, loading and unloading areas, washing areas, monitoring areas, etc.). These designated areas are fixed for a certain period of time and have corresponding rule attributes, such as fixed location, area division by blocks, two-way traffic, intersection of roads, and the presence of traffic lights, warning signs, etc.

[0003] With the rise and development of autonomous driving technology, there are increasing attempts to apply driverless / unmanned operations in the aforementioned closed / semi-closed scenarios. However, the current level of technological maturity and the complexity of the scenarios mean that human intervention cannot be completely eliminated. Often, other traffic participants, such as human-driven vehicles, pedestrians, or electric bicycles, are active in the driverless / unmanned operation areas. Currently, there are many autonomous driving methods based on obstacle trajectory prediction (i.e., obstacle intent), but these are all designed for single or low-complexity scenarios, such as autonomous driving on urban motor vehicle lanes. Existing trajectory prediction methods do not specifically address mixed scenarios (complex scenarios of driverless / unmanned operations involving mixed traffic activities of autonomous vehicles, human vehicles, pedestrians, electric bicycles, etc.). Their implementation is highly coupled and bound to the scenario, resulting in a lack of replicability. Because mixed scenarios differ greatly from completely single scenarios, such autonomous driving methods, detached from the scenario, are often difficult to adapt to and cannot be applied in practice. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide an autonomous driving method, device and autonomous vehicle based on obstacle intent, to overcome the problem that existing vehicle autonomous driving methods based on trajectory prediction are difficult to apply in complex mixed scenarios.

[0005] This invention provides an autonomous driving method based on obstacle intent, comprising:

[0006] Obtain local map data corresponding to the current location based on the vehicle's current location;

[0007] Based on the current location and the positional relationship between the vehicle's target driving area and each obstacle, the target obstacle is determined;

[0008] Based on the current location and the location of the target obstacle, lane filtering is performed on the local map data to determine the target lane of the obstacle;

[0009] Based on the connectivity between the target lane and other lanes, a lane travel path sequence for the target obstacle is established;

[0010] Based on the lane driving path sequence and the collected feature data of the target obstacle, predict the intended trajectory of the target obstacle;

[0011] Based on the correlation between the intended trajectory and the target driving area, the vehicle's path planning is updated.

[0012] Optionally, the step of filtering lanes on the local map data based on the current location and the location of the target obstacle to determine the target lane of the obstacle includes:

[0013] Based on the current location and the location of the target obstacle, all lanes within a set range are filtered on the local map data;

[0014] The lanes are sorted based on the distance between the selected lanes and the target obstacle and / or the angle between the selected lanes and the target obstacle.

[0015] The lane that is closest to the target obstacle and / or has the smallest angle with the target obstacle is determined as the target lane.

[0016] Optionally, predicting the intended trajectory of the target obstacle based on the lane driving path sequence and the collected feature data of the target obstacle includes:

[0017] The lane driving path sequence and the feature data of the target obstacle are input into the intention trajectory prediction model to obtain the predicted trajectory points and movement direction of the target obstacle;

[0018] The intended trajectory of the target obstacle is generated based on the predicted trajectory points and direction of movement of the target obstacle.

[0019] Optionally, the method further includes:

[0020] Obtain the actual trajectory of the obstacle;

[0021] Error analysis is performed on the intended trajectory using the feature data of the actual trajectory, the lane driving path sequence, and the target obstacle.

[0022] The intention trajectory prediction model is optimized based on the analysis results.

[0023] Optionally, based on the correlation between the intended trajectory and the target driving area, the vehicle's path planning is updated, including:

[0024] Based on the intended trajectory, it is determined whether the target obstacle will enter the target driving area within a first set time period;

[0025] When the target obstacle enters the target driving area within a set time, the intended trajectory is sampled;

[0026] Based on the positional relationship between each sampling point and the target driving area and the current driving parameters of the vehicle, the driving route and speed of the vehicle in the target driving area are planned;

[0027] Determine whether the intended trajectory and the driving route intersect at the same time;

[0028] When the intended trajectory and the driving route intersect at the same time, the vehicle's speed is adjusted based on the time corresponding to the intersection, and the process returns to the step of whether the intended trajectory and the driving route intersect at the same time, until the intended trajectory and the driving route do not intersect at any time.

[0029] Optionally, when the intended trajectory of the target obstacle remains on the driving route for a second set time, a warning message is sent to the remote control terminal;

[0030] The remote control terminal receives the early warning information and plans a new driving route accordingly.

[0031] The vehicle is controlled to travel along the new route to avoid the target obstacle.

[0032] Optionally, the current driving parameters of the vehicle include: the vehicle's attitude, speed, acceleration, and the traffic light detection results of the target driving area.

[0033] This invention also provides an autonomous driving device based on obstacle intent, comprising:

[0034] The acquisition module is used to acquire local map data corresponding to the current location of the vehicle.

[0035] The first processing module is used to determine the target obstacle based on the current position and the positional relationship between the vehicle's target driving area and each obstacle;

[0036] The second processing module is used to perform lane filtering on the local map data based on the current location and the location of the target obstacle, and to determine the target lane of the obstacle;

[0037] The third processing module is used to establish a lane travel path sequence for the target obstacle based on the connectivity between the target lane and other lanes.

[0038] The fourth processing module is used to predict the intended trajectory of the target obstacle based on the lane driving path sequence and the feature data of the target obstacle collected.

[0039] The fifth processing module is used to update the vehicle's path planning based on the correlation between the intended trajectory and the target driving area.

[0040] This invention also provides an autonomous driving vehicle, including a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the method provided in this invention.

[0041] This invention also provides a computer-readable storage medium storing computer instructions for causing a computer to execute the method provided in this invention.

[0042] The technical solution of this invention has the following advantages:

[0043] This invention provides an autonomous driving method based on obstacle intent. The method involves acquiring local map data corresponding to the vehicle's current location; determining target obstacles based on the current location, the vehicle's target driving area, and the positional relationship between obstacles; filtering lanes on the local map data based on the current location and the target obstacle's position to determine the target lane for the obstacle; establishing a lane driving path sequence for the target obstacle based on the connectivity between the target lane and other lanes; predicting the target obstacle's intent trajectory based on the lane driving path sequence and collected feature data; and updating the vehicle's path planning based on the correlation between the intent trajectory and the target driving area. By utilizing local map data of the vehicle's location, potential target obstacles that may affect vehicle driving are identified. Then, by combining the actual vehicle position and obstacle positional relationship in a mixed scenario, the target lane for the obstacle is determined. All possible lane driving paths for the target obstacle are established based on the actual connectivity between lanes. The intent trajectory is predicted based on the target obstacle's features, and the driving route and adaptive speed are adjusted in real-time based on the predicted trajectory to avoid collision risks. This improves transportation efficiency while ensuring safety, solves the problem of practical application of autonomous driving in complex scenarios, and enhances the adaptability of autonomous driving to the environment. Attached Figure Description

[0044] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0045] Figure 1 This is a flowchart of an autonomous driving method based on obstacle intent in an embodiment of the present invention;

[0046] Figure 2 This is an architecture diagram of an autonomous driving system based on obstacle intent in an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of the lane driving path sequence in an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of the structure of an autonomous driving device based on obstacle intent in an embodiment of the present invention;

[0049] Figure 5 This is a schematic diagram of the structure of an autonomous vehicle in an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0052] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can also refer to the internal connection of two components; and they can refer to a wireless connection or a wired connection. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0053] The technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0054] To address the aforementioned issues, this invention provides an obstacle-intention-based autonomous driving method, applied to an autonomous driving system for engineering vehicles in a specific area within a mixed scenario. First, the terms used in this invention are explained: Mixed scenario: A complex scenario involving unmanned driving / unmanned operation where autonomous vehicles, manual vehicles, pedestrians, and electric bicycles engage in mixed traffic activities. Specific area: A designated area (such as a weighing area, loading / unloading area, washing area, monitoring area, etc.) through which unmanned vehicles complete their tasks. This area is fixed within a certain timeframe and possesses corresponding rule attributes, such as fixed location, area divided into blocks, two-way traffic, intersections, and attributes including at least one such as traffic lights or warning signs. Obstacle intention: The trajectory of an obstacle in the next few seconds.

[0055] like Figure 1 As shown, this obstacle-intention-based autonomous driving method specifically includes the following steps:

[0056] Step S101: Obtain local map data corresponding to the current location based on the vehicle's current location.

[0057] Specifically, such as Figure 2As shown, in the aforementioned autonomous driving system for engineering vehicles based on obstacle intent in a specific area within a mixed scenario, the perception sensor and data acquisition unit acquire sensor data, including radar data or image data. The corresponding sensors include radar modules (LiDAR, millimeter-wave radar, ultrasonic radar, etc.) and camera modules (ordinary cameras, infrared cameras, depth cameras, etc.). The data preprocessing unit stores and corrects the perception sensor data in a unified format (e.g., converting image data into CV::Mat or Image format, performing width and height alignment, and correcting the image data), and records the data frame number and corresponding data status information. Vehicle and environmental information includes the vehicle's positioning information, vehicle information (speed, acceleration, attitude, etc.), and environmental information such as routes and lanes composed of high-precision maps, providing decision-making basis for the entire autonomous driving system.

[0058] First, the system initializes and checks the integrity of its parameters, including but not limited to: whether the high-precision map data is complete (including road information, scene information, and fixed traffic sign information); whether the vehicle's own status parameters are normal; whether the perception sensors have data and their data status; whether predetermined points in specific areas are set, and whether obstacle intent parameters are set, etc. Based on the inheritance and connection relationships, the road is divided into segments, and a road data warehouse is established based on the global high-precision map. If it has already been established and the global map or operating routes and facilities have not changed, no update is needed, and this step is skipped. The system determines whether the vehicle is on the preset route. If not, the data is uploaded to the remote dispatch unit's dispatch platform for correction; if so, proceed to the next step. The system obtains real-time information from the local high-precision map. This information is map data for a local segment based on the vehicle's location. Based on this information, the data in the road data warehouse is converted into data in the vehicle's coordinate system and stored and labeled. For example, the local map data in the world coordinate system of the road data warehouse is converted to the vehicle coordinate system. The specific coordinate conversion method is existing technology and will not be elaborated here. The aforementioned local map data can be obtained by performing coordinate conversion on the local map data.

[0059] Step S102: Determine the target obstacle based on the current location and the positional relationship between the vehicle's target driving area and each obstacle.

[0060] Specifically, the system acquires scene information (including predetermined points, curves, straight sections, intersections, etc.) and corresponding obstacle intent parameters from map data, and determines whether to enter the set scene and updates the corresponding parameters in the scene. Obstacle intent parameters include: search range for lanes and obstacles, angle difference and distance difference thresholds, and dynamic / static state judgment thresholds, etc., which are preset. Then, the system judges the state of obstacles, classifying them as dynamic or static, and sets their priority levels based on conditions, categorizing them as attention-worthy obstacles, ignore obstacles, and general obstacles. These conditions can be based on whether the obstacle is on the lane, near a predetermined point in a specific area, or within 100 meters in front of the vehicle. For example, if the obstacle is near a predetermined point in a specific area (the vehicle's target driving area) and relatively close to the vehicle (e.g., within 50 meters), or relatively close to the front or rear of the vehicle's target lane, then the obstacle is set as a attention-worthy obstacle (the threshold here is further refined according to the actual scene and driving route), i.e., the aforementioned target obstacle.

[0061] Step S103: Based on the current location and the location of the target obstacle, perform lane filtering on the local map data to determine the target lane of the obstacle.

[0062] Step S104: Based on the connectivity between the target lane and other lanes, establish the lane driving path sequence for the target obstacle.

[0063] Specifically, possible path sequences are established and stored by utilizing the connectivity between the selected target lane and other lanes, such as lane1, lane2, lane3, etc. The maximum value of the sum of the lengths of all roads in the road sequence is set in advance, such as 100m, etc. This invention is not limited to this.

[0064] For example, a lane travel path sequence refers to all possible lane sequences at the location of the obstacle, such as... Figure 3 As shown, the obstacle is located in lane 0, and its possible routes are: lane 0->lane 4->lane 5, lane 0->lane 2->lane 6, lane 1->lane 3->lane 7. Then, a recursive comparison is made between the maximum predicted distance of the obstacle within the prediction time and all lanes of the obstacle. That is, for example, if the maximum predicted distance of the obstacle is only a small segment (e.g., it cannot reach lane 7), then the final possible road sequence is lane 0->lane 4->lane 5, lane 0->lane 2, lane 1->lane 3. In this way, the lane driving path sequence is established.

[0065] Step S105: Based on the lane driving path sequence and the collected feature data of the target obstacle, predict the intended trajectory of the target obstacle.

[0066] The obstacle's feature data includes, but is not limited to, perceived data about the obstacle, such as its length, width, height, speed, orientation, and type. The target obstacle's intended trajectory is the predicted movement trajectory of the obstacle. For example, this intended trajectory is the aforementioned road sequence lane0->lane2, meaning it is predicted that the target obstacle will move along road lane0 to lane1.

[0067] Step S106: Update the vehicle's path planning based on the correlation between the intended trajectory and the target driving area.

[0068] Specifically, based on the intended trajectory of the obstacle, it is determined whether the obstacles around a specific area will be within the target driving area in the current and the next few seconds (e.g., 3s, 5s), and the vehicle's path planning is updated accordingly to avoid the risk of collision with the obstacle.

[0069] By performing the above steps, the obstacle-intention-based autonomous driving method provided in this embodiment of the invention obtains target obstacles that may affect vehicle driving by utilizing local map data of the vehicle's location. Then, by combining the actual vehicle position and obstacle position in the mixed scene, the target lane of the obstacle is determined, and all possible lane driving paths of the target obstacle are established based on the actual connectivity between lanes. The intention trajectory is predicted by combining the characteristics of the target obstacle, and the driving route and adaptive speed are adjusted in real time based on the predicted trajectory to avoid collision risks. This improves the transportation efficiency of the vehicle under the premise of safety, solves the problem of the practical application of autonomous driving in complex scenarios, and improves the adaptability of autonomous driving to the environment.

[0070] Specifically, in one embodiment, step S103 above includes the following steps:

[0071] Step S201: Based on the current location and the location of the target obstacle, filter all lanes within a set range on the local map data.

[0072] Specifically, local map data based on the vehicle's location is converted into road information data under the vehicle's coordinates and stored. The data is stored in a queue and associated containers. For example, the queue size is set to 5, which means that when the size is greater than 5, the oldest data will be updated.

[0073] Step S202: Sort each lane based on the distance between the selected lane and the target obstacle and / or the angle of the lane relative to the target obstacle.

[0074] Step S203: Determine the lane with the smallest distance to the target obstacle and / or the smallest angle with the target obstacle as the target lane.

[0075] Specifically, the nearest road data location points for the target obstacle and the vehicle's location points are found separately. Specifically, all lanes are found based on the obstacle (including the vehicle) location points and the search range (e.g., within 10m). Then, the angle difference between the direction of the lane containing the nearest road data location point of the obstacle and the direction of the obstacle is calculated. The obstacle location point is mapped to the lane coordinates and the distance to the boundary of all found lanes is calculated to find the nearest lane. The lane is obtained by combining the minimum angle difference and the nearest distance, and the lane direction and the obstacle direction are recorded.

[0076] The minimum angle difference is calculated by taking the angle difference between the direction of the lane containing the nearest lane data point (one angle) and the orientation of the obstacle (another angle). This is iterated to find the lane with the minimum angle difference from the obstacle's orientation. The nearest distance is calculated by mapping the obstacle's location to lane coordinates and finding the distance to the boundaries of all found lanes. These two factors are then combined and processed according to different scenarios (e.g., Scenario 1: angle difference less than threshold 1 or distance less than threshold 2; Scenario 2: angle difference less than threshold 3 and distance less than threshold 4, with thresholds determined based on the scenario). For different scenarios, Scenario 1 is sufficient for straight roads, while Scenario 2 is chosen for intersections or intersections of multiple directions to avoid selecting the wrong lane.

[0077] Specifically, in one embodiment, step S105 specifically includes the following steps:

[0078] Step S301: Input the lane driving path sequence and the feature data of the target obstacle into the intention trajectory prediction model to obtain the predicted trajectory points and movement direction of the target obstacle.

[0079] For example, the intention trajectory prediction model can employ existing technologies, such as LSTM. This model outputs the confidence level (probability) of obstacles and trajectory points, and the direction of movement can be determined by judging the predicted trajectory points.

[0080] Step S302: Generate the intended trajectory of the target obstacle based on the predicted trajectory points and movement direction of the target obstacle.

[0081] Specifically, by inputting the feature data of the aforementioned obstacles and the lane driving path sequence of the obstacles into the intention trajectory prediction model, the intention trajectory is predicted, and the probability value and the trajectory point and direction with the highest probability are output (in the case of outputting the trajectory of the reverse direction when the obstacle is traveling in the wrong direction), the intention trajectory of the obstacle is obtained.

[0082] Specifically, in practical applications, after obtaining the intended trajectory of the target obstacle, the above method further acquires the actual trajectory of the obstacle; performs error analysis on the intended trajectory using the actual trajectory, lane driving path sequence, and feature data of the target obstacle; and optimizes the intended trajectory prediction model based on the analysis results. For example, by recording the measured trajectory within a few seconds, the actual trajectory, and the features of the obstacle and path, and sending them to the system's error analysis module, a feedback mechanism is used to adjust model parameters and optimize the model. The resulting obstacle's intended trajectory is associated with a specific area in the mixed scene, without being restricted by traffic roads or rules, making it highly modifiable and thus increasing its adaptability. By adding an error analysis module, the feedback mechanism adjusts the model parameters. Specifically, it first analyzes the error between the model-predicted trajectory and the actual trajectory (e.g., the obstacle's orientation is opposite, the trajectory deviation is too large, or there are issues with speed and direction), and adjusts the parameters accordingly. Here, deviations are not necessarily model problems; they could be due to issues with high-precision map data, or problems with the perception module's conversion of obstacle orientation, etc. Therefore, this error analysis module categorizes and corrects deviations. Secondly, for the model, this is equivalent to feeding in real training data, facilitating parameter adjustment and model optimization. The road information collected is based on data from a local segment of the vehicle, and older data is processed using an update method. Furthermore, even when a vehicle is traveling in the wrong direction, its normal trajectory direction will be recorded.

[0083] Specifically, in one embodiment, step S106 above includes the following steps:

[0084] Step S401: Based on the intended trajectory, determine whether the target obstacle will enter the target driving area within a first set time.

[0085] Specifically, a 3D mesh is created for a specific area, i.e., the target driving area, and a database is established. If the database has already been created and the specific area remains unchanged, there is no need to rebuild or update it. The height of the 3D mesh or a range is set based on the maximum height of vehicles passing through the specific area as a baseline. Database creation can be achieved by dividing the 3D mesh into smaller meshes in an equal manner, and storing the 3D center coordinates of each mesh using a KD-tree. Obstacle fusion detection results are obtained and their normality is determined. A limit is then placed on the height of obstacles in the detection results, or a template value is used to prevent abrupt or abnormal height changes in the detection results. For obstacles with normal detection results, a matching judgment is performed with the 3D mesh of the specific area, i.e., by matching the obstacle's center point, length, width, height, and vertices. The KD-tree search is performed on the 3D mesh points to find the nearest points for matching. Specifically, the mesh point closest to the center point of the searched obstacle location is used as the matching point for the obstacle's center point. Then, based on the four vertices with the smallest angular difference between the obstacle's orientation and the mesh's normal vector, points at the same height adjacent to the nearest mesh point of the searched obstacle's center point are found as matching points for the obstacle vertices. A distance loss function is calculated based on the matching points; this can be Euclidean distance, Manhattan distance, or cosine distance, etc. It is then determined whether the obstacle is within a specific region by comparing the calculated distance loss with a normal proximity threshold range (pre-calibrated based on the actual scene and vehicle proximity). If the distance is within the range, the obstacle is considered inside the boundary; otherwise, it is outside the boundary. The states of the obstacle and the boundary / outside the boundary are stored sequentially. The above process is repeated for the sampling points and poses of the obstacle's intended trajectory.

[0086] Step S402: When the target obstacle enters the target driving area within a set time, the intended trajectory is sampled.

[0087] Specifically, for the intended trajectory of an obstacle, it is first determined whether it exists. If it does not exist, it is likely a static obstacle or an unknown obstacle. In this case, the current pose of the obstacle is determined whether it is within the area and whether its position within the area poses a safety issue to the vehicle's predetermined route. If the intended trajectory of the obstacle exists, the trajectory points are sampled and the results are sent to the running logic judgment and processing submodule.

[0088] Step S403: Based on the positional relationship between each sampling point and the target driving area and the vehicle's current driving parameters, plan the vehicle's driving route and speed in the target driving area.

[0089] The vehicle's current driving parameters include: vehicle attitude, speed, acceleration, and traffic light detection results for the target driving area. For example, by acquiring traffic light detection results for a specific area (e.g., green for no operation, red for operation, white for abnormal), and for example, if the light is red, the vehicle waits at a predetermined location; once it turns green, the vehicle's driving route is planned based on the relationship between the obstacle sampling points and the specific area, as well as the vehicle's attitude, speed, and acceleration information. Specific planning of the vehicle's driving route and speed can be implemented with reference to existing methods for path and speed planning in autonomous driving, but this invention is not limited thereto.

[0090] Step S404: Determine whether there is an intersection between the intended trajectory and the driving route at the same time.

[0091] Step S405: When the intended trajectory and the driving route intersect at the same time, adjust the vehicle's speed based on the time corresponding to the intersection, and return to the step of whether the intended trajectory and the driving route intersect at the same time, until the intended trajectory and the driving route do not intersect at any time.

[0092] Specifically, by determining whether the intended trajectory of the obstacle next to the planned driving route intersects with the vehicle's driving route at the set speed, the speed and acceleration are adjusted within a safe range based on this intersection (such as speeding up or slowing down and then speeding up again) to avoid intersections within a certain period of time, so as to improve traffic efficiency. That is, if the intersection can be avoided by following the adaptive speed of the trajectory, then the vehicle will proceed according to the driving route; otherwise, it will enter the overall control scheduling and release module.

[0093] Step S406: When the intended trajectory of the target obstacle remains on the driving route for a second set time, send a warning message to the remote control terminal.

[0094] Step S407: Receive the remote control terminal and plan a new driving route based on the early warning information.

[0095] Step S408: Control the vehicle to travel along the new route to avoid the target obstacle.

[0096] Specifically, the central control dispatching module, also known as the remote control terminal, issues warning information to the central control module based on static and dynamic conditions for obstacles that block the planned driving route for a certain period of time (e.g., more than 5 minutes). This allows the remote control of the vehicle to issue warnings and dispatch new routes for the autonomous vehicle. If the vehicle is unable to travel after multiple attempts, a new dispatch is initiated until it can travel on one of the preset routes or temporarily suspends travel until intervention resolves the issue.

[0097] The following will combine, for example, Figure 2The system shown provides a detailed description of the obstacle-intention-based autonomous driving method provided in this embodiment of the invention.

[0098] like Figure 2 As shown, in addition to the aforementioned perception sensors and data acquisition units, the system also includes: a central processing and computing unit comprising a perception module, a planning module, a control module, an obstacle intent logic processing unit, and a computing platform or central controller. The perception module processes the pre-processed images and radar data using algorithms to obtain fused target information, including target category, ID, target position and speed information, target length, width, height, and timestamp. The obstacle intent and logic processing unit combines the perception fusion information, the vehicle's planned path information from the previous frame provided by the planning module, control information, and information stored in a unified format by the vehicle and environment information units to predict the trajectory of the perceived obstacle and sends the predicted trajectory and the actual trajectory to the error analysis module. Simultaneously, the logic processing unit performs real-time logical operations to provide efficient autonomous driving operation; a remote scheduling unit is used for scheduling and remotely controlling the vehicle.

[0099] The sensing sensor module mainly acquires sensing sensor data (image data and / or radar data). The sensors include, but are not limited to, radar modules (LiDAR, millimeter-wave radar, ultrasonic radar, etc.) and camera modules (ordinary cameras, infrared cameras, depth cameras, etc.).

[0100] Central Processing and Computing Unit Module: The core of this system and the main control module for achieving safe and efficient operation of autonomous vehicles in mixed scenarios. It performs logical calculations on the relationship between the trajectory of perceived obstacles generated based on environmental and vehicle information and the trajectory to be driven by the vehicle, forming an efficient mechanism.

[0101] Remote dispatch unit: remotely controls and dispatches vehicles and operations, providing intervention guarantees for the efficient operation of autonomous vehicles.

[0102] For example, the specific working process of the above system is as follows:

[0103] S1. System parameter initialization settings and integrity checks, including but not limited to: whether the high-precision map data is complete (including road information, scene information, and fixed traffic sign information); whether the vehicle parameters are normal; whether the perception sensors have data and their data status; whether the predetermined points in specific areas are set and the obstacle intent parameters are set, etc.

[0104] S2. Divide the roads into segments according to the inheritance and connection relationships, and establish a road data warehouse based on the global high-precision map. If the data warehouse has already been established and the global map or the running routes and facilities have not changed, no update is required, and this step can be skipped.

[0105] S3. Determine if the vehicle is on the preset route. If not, upload the information to the dispatch platform for correction. If yes, proceed to the next step.

[0106] S4. Obtain real-time information from local high-precision maps. This information is based on local segment map data of the vehicle's location. Based on this information, the data from the road data warehouse is converted into data under the vehicle's coordinates and stored and labeled.

[0107] S5. Obtain the intended trajectory of the obstacle and store the intended trajectory and the actual trajectory.

[0108] S6. Determine whether the predetermined position of the specific area has been reached. If it has, update the scene information and obstacle intent parameters, including the search range of lanes and obstacles, vehicle speed, dynamic and static judgment thresholds, and other parameter information.

[0109] S7. Based on the obstacle's intended trajectory, determine whether the obstacles around a specific area will be within the area in the current and next few seconds (e.g., 3s, 5s). For the obstacle's intended trajectory, first determine if it exists. If it does not exist, it indicates that it is likely a static obstacle or an unknown obstacle. Then, only determine whether the obstacle's current pose is within the area and whether its position within the area poses a safety issue to the vehicle's predetermined route. If the obstacle's intended trajectory exists, sample the trajectory points and determine whether the current position point and the sampled point are within the area, and then send the data to the computing platform.

[0110] S8. Obtain the signal light detection results for a specific area, such as green for no work, red for work in progress, and white for abnormality, and send the results to the computing platform.

[0111] S9, the computing platform, receives and processes the results of S7 and S8. Specifically, if the result of S8 is red, it waits at the predetermined location point 2. When the result of S8 is green, it plans the driving route of the vehicle based on the relationship between the obstacle sampling points and the specific area in S7, as well as the vehicle's attitude and speed acceleration information. Then, it comprehensively judges whether it can reach the destination. Specifically, it first judges whether the intended trajectory of the relevant obstacles next to the planned driving route intersects with the vehicle's driving route at the set speed. Based on this intersection, it adjusts the speed and acceleration within a safe range (such as speeding up or slowing down and then speeding up again) to avoid intersections within a certain period of time in order to improve efficiency. That is, if the speed can avoid intersections according to the trajectory, it will proceed according to the driving route; otherwise, it will enter the remote scheduling unit.

[0112] S10, the remote dispatch unit, issues warning information to the dispatch platform for obstacles that block the planned driving route for a certain period of time (e.g., more than 5 minutes), based on static and dynamic conditions. It remotely controls the vehicle to issue warnings and dispatches the autonomous vehicle to a new route, then re-enters S4 to S9. If the vehicle is unable to drive after multiple attempts, it will be dispatched again until it can travel on one of the preset routes or temporarily suspend driving until intervention resolves the issue.

[0113] By using the intention prediction of obstacles to obtain their trajectory in the next few seconds, the relationship between the obstacle and the trajectory to be driven by the autonomous vehicle can be determined in advance. Based on the safety mechanism, an efficient operation logic scheme is added to form a safe and efficient mechanism and improve the efficiency of unmanned transportation. This truly solves the problem of the coexistence of autonomous driving and human driving in the actual application of unmanned vehicles.

[0114] By performing the above steps, the obstacle-intention-based autonomous driving method provided in this embodiment of the invention obtains target obstacles that may affect vehicle driving by utilizing local map data of the vehicle's location. Then, by combining the actual vehicle position and obstacle position in the mixed scene, the target lane of the obstacle is determined, and all possible lane driving paths of the target obstacle are established based on the actual connectivity between lanes. The intention trajectory is predicted by combining the characteristics of the target obstacle, and the driving route and adaptive speed are adjusted in real time based on the predicted trajectory to avoid collision risks. This improves the transportation efficiency of the vehicle under the premise of safety, solves the problem of the practical application of autonomous driving in complex scenarios, and improves the adaptability of autonomous driving to the environment.

[0115] This invention also provides an autonomous driving device based on obstacle intent, such as... Figure 4 As shown, the obstacle-indicating autonomous driving device specifically includes:

[0116] The acquisition module 101 is used to acquire local map data corresponding to the current location of the vehicle. For details, please refer to the relevant description of step S101 in the above method embodiment, which will not be repeated here.

[0117] The first processing module 102 is used to determine the target obstacle based on the current location and the positional relationship between the vehicle's target driving area and each obstacle. For details, please refer to the relevant description of step S102 in the above method embodiment, which will not be repeated here.

[0118] The second processing module 103 is used to perform lane filtering on local map data based on the current location and the location of the target obstacle to determine the target lane of the obstacle. For details, please refer to the relevant description of step S103 in the above method embodiment, which will not be repeated here.

[0119] The third processing module 104 is used to establish a lane travel path sequence for the target obstacle based on the connectivity between the target lane and other lanes. For details, please refer to the relevant description of step S104 in the above method embodiment, which will not be repeated here.

[0120] The fourth processing module 105 is used to predict the intended trajectory of the target obstacle based on the lane driving path sequence and the collected feature data of the target obstacle. For details, please refer to the relevant description of step S105 in the above method embodiment, which will not be repeated here.

[0121] The fifth processing module 106 is used to update the vehicle's path planning based on the correlation between the intended trajectory and the target driving area. For details, please refer to the relevant description of step S106 in the above method embodiment, which will not be repeated here.

[0122] For a more detailed description of each of the above functional modules, please refer to the relevant descriptions in the above method embodiments, which will not be repeated here.

[0123] Through the collaborative efforts of the aforementioned components, the obstacle-intention-based autonomous driving device provided in this embodiment of the invention obtains target obstacles that may affect vehicle driving by utilizing local map data of the vehicle's location. Then, by combining the actual vehicle position and obstacle position in the mixed scene, it determines the target lane of the obstacle and establishes all possible lane driving paths for the target obstacle based on the actual connectivity between lanes. It predicts the intended trajectory by combining the characteristics of the target obstacle and then adjusts the driving route and adaptive speed in real time based on the predicted trajectory to avoid collision risks. This improves the transportation efficiency of the vehicle under the premise of safety, solves the problem of practical application of autonomous driving in complex scenarios, and improves the adaptability of autonomous driving to the environment.

[0124] An autonomous driving vehicle is also provided according to embodiments of the present invention, such as Figure 5 As shown, the autonomous vehicle may include a processor 901 and a memory 902, wherein the processor 901 and the memory 902 can be connected via a bus or other means. Figure 5 Taking a bus connection as an example, the autonomous vehicle is, for instance, an engineering vehicle.

[0125] Processor 901 can be a Central Processing Unit (CPU). Processor 901 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0126] The memory 902, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the method embodiments of the present invention. The processor 901 executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions, and modules stored in the memory 902, thereby implementing the methods in the above method embodiments.

[0127] The memory 902 may include a program storage area and a data storage area. The program storage area may store application programs required for operating the device and at least one function; the data storage area may store data created by the processor 901, etc. Furthermore, the memory 902 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 902 may optionally include memory remotely located relative to the processor 901, and these remote memories may be connected to the processor 901 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0128] One or more modules are stored in memory 902, and when executed by processor 901, they perform the methods described in the above method embodiments.

[0129] The specific details of the aforementioned autonomous vehicles can be understood by referring to the relevant descriptions and effects in the above method embodiments, and will not be repeated here.

[0130] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The implemented program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

[0131] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. An autonomous driving method based on obstacle intent, characterized in that, include: Obtain local map data corresponding to the current location based on the vehicle's current location; Based on the current location and the positional relationship between the vehicle's target driving area and each obstacle, the target obstacle is determined; Based on the current location and the location of the target obstacle, lane filtering is performed on the local map data to determine the target lane of the obstacle; Based on the connectivity between the target lane and other lanes, a lane travel path sequence for the target obstacle is established; Based on the lane driving path sequence and the collected feature data of the target obstacle, predict the intended trajectory of the target obstacle; Based on the correlation between the intended trajectory and the target driving area, the vehicle's path planning is updated; Based on the correlation between the intent trajectory and the target driving area, the vehicle's path planning is updated, including: determining whether a target obstacle will enter the target driving area within a first set time period based on the intent trajectory; when the target obstacle will enter the target driving area within the set time period, sampling the intent trajectory; planning the vehicle's driving route and speed in the target driving area based on the positional relationship between each sampling point and the target driving area and the vehicle's current driving parameters; determining whether the intent trajectory and the driving route intersect at the same time; when the intent trajectory and the driving route intersect at the same time, adjusting the vehicle's speed based on the time corresponding to the intersection point, and returning to the step of determining whether the intent trajectory and the driving route intersect at the same time, until the intent trajectory and the driving route do not intersect at any time; When the intended trajectory of the target obstacle remains on the driving route for a second set time, a warning message is sent to the remote control terminal. The remote control terminal receives the early warning information and plans a new driving route accordingly. The vehicle is controlled to travel along the new route to avoid the target obstacle.

2. The method according to claim 1, characterized in that, The step of filtering lanes on the local map data based on the current location and the location of the target obstacle to determine the target lane for the obstacle includes: Based on the current location and the location of the target obstacle, all lanes within a set range are filtered on the local map data; The lanes are sorted based on the distance between the selected lanes and the target obstacle and / or the angle between the selected lanes and the target obstacle. The lane that is closest to the target obstacle and / or has the smallest angle with the target obstacle is determined as the target lane.

3. The method according to claim 1, characterized in that, The step of predicting the intended trajectory of the target obstacle based on the lane driving path sequence and the collected feature data of the target obstacle includes: The lane driving path sequence and the feature data of the target obstacle are input into the intention trajectory prediction model to obtain the predicted trajectory points and movement direction of the target obstacle; The intended trajectory of the target obstacle is generated based on the predicted trajectory points and direction of movement of the target obstacle.

4. The method according to claim 3, characterized in that, Also includes: Obtain the actual trajectory of the obstacle; Error analysis is performed on the intended trajectory using the feature data of the actual trajectory, the lane driving path sequence, and the target obstacle. The intention trajectory prediction model is optimized based on the analysis results.

5. The method according to claim 1, characterized in that, The vehicle's current driving parameters include: the vehicle's attitude, speed, acceleration, and the traffic light detection results for the target driving area.

6. An autonomous driving device based on obstacle intent, characterized in that, include: The acquisition module is used to acquire local map data corresponding to the current location of the vehicle. The first processing module is used to determine the target obstacle based on the current position and the positional relationship between the vehicle's target driving area and each obstacle; The second processing module is used to perform lane filtering on the local map data based on the current location and the location of the target obstacle, and to determine the target lane of the obstacle; The third processing module is used to establish a lane travel path sequence for the target obstacle based on the connectivity between the target lane and other lanes. The fourth processing module is used to predict the intended trajectory of the target obstacle based on the lane driving path sequence and the feature data of the target obstacle collected; The fifth processing module is used to update the vehicle's path planning based on the correlation between the intended trajectory and the target driving area; Based on the correlation between the intended trajectory and the target driving area, the vehicle's path planning is updated, including: determining whether the target obstacle will enter the target driving area within a first set time period based on the intended trajectory; sampling the intended trajectory when the target obstacle will enter the target driving area within the set time period; planning the vehicle's driving route and speed in the target driving area based on the positional relationship between each sampling point and the target driving area and the vehicle's current driving parameters; determining whether the intended trajectory and the driving route intersect at the same time; adjusting the vehicle's speed based on the time corresponding to the intersection point when the intended trajectory and the driving route intersect at the same time, and returning to the step of determining whether the intended trajectory and the driving route intersect at the same time, until the intended trajectory and the driving route do not intersect at any time; sending a warning message to a remote control terminal when the intended trajectory of the target obstacle remains on the driving route for a second set time period; receiving a new driving route planned by the remote control terminal based on the warning message; and controlling the vehicle to drive according to the new driving route to avoid the target obstacle.

7. An autonomous driving vehicle, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method of any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the method described in any one of claims 1-5.

Images