Turn-round control method and device of automatic driving vehicle and automatic driving vehicle

Abstract

The invention discloses a U-turn control method and device of an automatic driving vehicle and the automatic driving vehicle. The method comprises the steps that a first prediction track of a traffic participation object is acquired, and meanwhile, a second prediction track of a target automatic driving vehicle is acquired; when it is determined that a coincident trajectory exists between the first prediction trajectory and the second prediction trajectory based on a comparison result of the first prediction trajectory and the second prediction trajectory, obtaining first arrival time when the traffic participant arrives at each prediction coincident point on the coincident trajectory; evaluating the first arrival time to obtain the influence degree of the traffic participant on the turning-around operation; adjusting the original turning-around strategy according to the influence degree to obtain a target turning-around strategy; and controlling the target autonomous vehicle to execute a turning-around operation according to the target turning-around strategy. According to the invention, the technical problem that the traffic capacity of the autonomous vehicle is limited due to the obvious insufficiency of the autonomous vehicle turn-round path planning in the aspects of dynamic traffic flow influence processing and adaptive adjustment in the related technology is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle control, in particular to a U-turn control method and device of an autonomous vehicle and the autonomous vehicle. BACKGROUND

[0002] With the continuous evolution of intelligent transportation systems and autonomous driving technologies, the autonomous driving capability of autonomous vehicles in complex urban traffic environments has become one of the key points of technological development. U-turn operation, as a common driving behavior, becomes exceptionally complex due to its high dependence on the dynamic characteristics of the surrounding traffic environment. Especially in the urban environment with dense traffic flow and complex road layout, autonomous vehicles face multiple challenges, including but not limited to space limitations, safety considerations, traffic rule compliance, and minimization of impact on other traffic participants.

[0003] Currently, the research on U-turn path planning for autonomous vehicles mainly focuses on the following aspects: 1) static obstacle avoidance: existing technologies usually focus on avoiding fixed obstacles and following road geometric features for path planning. For example, the method disclosed in Comparative Document D1 determines the U-turn starting point and turning-out point by analyzing static obstacles in the U-turn scenario, such as kerbs, pedestrian crossings, and fixed traffic facilities, to generate a safe U-turn path. Although this method can ensure safe U-turn of the vehicle to some extent, it is not sufficient when facing dynamic traffic flow, which can lead to path planning failure or unsatisfactory results. 2) Road geometry feature driven decision: Comparative Document D2 proposes a U-turn path planning method based on road geometry features, which triggers open space path planning by evaluating the curvature of the trajectory points on the front road reference line to find a suitable U-turn opportunity. However, this method ignores the influence of real-time traffic flow, making it difficult to find a safe and efficient U-turn path during busy traffic periods, especially at intersections and busy road sections. Vehicles may have to wait for a long time, not only reducing traffic efficiency, but also causing traffic congestion and even increasing the risk of accidents. 3) Lack of adaptive adjustment mechanism: existing U-turn path planning algorithms mostly use preset constraints and search strategies. Once encountering complex traffic scenarios, especially when traffic flow patterns change rapidly, fixed planning strategies are difficult to adapt to different driving environments, resulting in either overly conservative planned paths that affect traffic efficiency or overly aggressive planned paths that increase safety risks.

[0004] In summary, the existing U-turn path planning technology for autonomous vehicles has obvious shortcomings in handling dynamic traffic flow and adaptive adjustment, which limits the traffic capacity and safety of autonomous vehicles in complex urban traffic environments.

[0005] To address the above problems, no effective solutions have been proposed so far. SUMMARY

[0006] The embodiments of the present application provide a U-turn control method and device of an autonomous vehicle, and the autonomous vehicle, to at least solve the technical problem of obvious deficiency in U-turn path planning of the autonomous vehicle in handling dynamic traffic flow influence and adaptive adjustment, which limits the traffic capacity of the autonomous vehicle.

[0007] According to an aspect of the embodiments of the present application, a U-turn control method of an autonomous vehicle is provided, including: obtaining a first predicted trajectory of a traffic participant within a predetermined range centered on a target autonomous vehicle, and obtaining a second predicted trajectory of the target autonomous vehicle performing a U-turn operation according to an original U-turn strategy; when it is determined that the first predicted trajectory and the second predicted trajectory exist in a coincident trajectory based on a comparison result of the first predicted trajectory and the second predicted trajectory, obtaining a first arrival time of the traffic participant to each predicted coincident point on the coincident trajectory; evaluating the first arrival time to obtain an influence degree of the traffic participant on the U-turn operation; adjusting the original U-turn strategy according to the influence degree to obtain a target U-turn strategy; and controlling the target autonomous vehicle to perform the U-turn operation according to the target U-turn strategy.

[0008] Optionally, the obtaining of the first predicted trajectory of the traffic participant within the predetermined range centered on the target autonomous vehicle, and the obtaining of the second predicted trajectory of the target autonomous vehicle performing the U-turn operation according to the original U-turn strategy, include: collecting, by the target autonomous vehicle, first motion state information of the traffic participant within the predetermined range, and processing the motion state information by using a trajectory prediction model to obtain the first predicted trajectory, wherein the trajectory prediction model is a model trained by machine learning using a plurality of sets of training data; and obtaining second motion state information of the target autonomous vehicle, and generating the second predicted trajectory according to the second motion state information and the original U-turn strategy.

[0009] Optionally, the determination of the first predicted trajectory and the second predicted trajectory existing in the coincident trajectory based on the comparison result of the first predicted trajectory and the second predicted trajectory, includes: converting the first predicted trajectory and the second predicted trajectory into the same time-space coordinate system to obtain a third predicted trajectory and a fourth predicted trajectory; determining a shortest distance between each point on the third predicted trajectory and the fourth predicted trajectory to obtain a shortest distance set; and determining that the first predicted trajectory and the second predicted trajectory exist in the coincident trajectory when it is determined that there is a trajectory point less than a preset safety threshold in the shortest distance set.

[0010] Optionally, the original U-turn strategy is adjusted according to the influence degree to obtain a target U-turn strategy, including: when it is determined that the first predicted trajectory and the second predicted trajectory have a coincident trajectory based on a comparison result of the first predicted trajectory and the second predicted trajectory, acquiring a first arrival time of the traffic participant to each predicted coincident point on the coincident trajectory, and simultaneously analyzing reachability of a road network corresponding to a road where the target autonomous vehicle is located and a bottleneck area of the road network by using graph theory connectivity detection; determining a dynamic feature and an obstacle point of a traffic flow corresponding to the road according to the reachability and the bottleneck area; obtaining a traffic flow mode of the traffic flow according to the dynamic feature and the obstacle point; generating an adaptive soft constraint according to the traffic flow mode, and adjusting the original U-turn strategy according to the adaptive soft constraint and the influence degree to obtain the target U-turn strategy.

[0011] Optionally, the original U-turn strategy is adjusted according to the influence degree to obtain a target U-turn strategy, including: in a process of driving of the target autonomous vehicle, acquiring environmental change information of the target autonomous vehicle; updating the traffic flow mode and a weight of the adaptive soft constraint according to the environmental change information to obtain an updated traffic flow mode and an updated adaptive soft constraint; adjusting the original U-turn strategy according to the updated traffic flow mode and the updated adaptive soft constraint to obtain the target U-turn strategy.

[0012] Optionally, the first arrival time is evaluated to obtain the influence degree of the traffic participant on the U-turn operation, including: acquiring a starting time point at which the target autonomous vehicle starts to perform the U-turn operation, and comparing the starting time point with the first arrival time to calculate a time difference value between the starting time point and the first arrival time; acquiring a type and a behavior of the traffic participant; determining the influence degree according to the type, the behavior and the time difference value.

[0013] Optionally, the original U-turn strategy is adjusted according to the influence degree to obtain a target U-turn strategy, including: when it is determined that the influence degree is greater than an influence degree threshold, postponing the U-turn operation, or adjusting the second predicted trajectory.

[0014] Optionally, the original U-turn strategy is adjusted according to the influence degree to obtain a target U-turn strategy, including: adopting a multi-level progressive degradation strategy to dynamically adjust a search constraint condition to adjust the original U-turn strategy to obtain the target U-turn strategy, wherein the search constraint condition is used to constrain a U-turn path of the target autonomous vehicle, and the multi-level progressive degradation strategy is used to adjust a margin of the search constraint condition.

[0015] Optionally, a multi-stage progressive degradation strategy is adopted to dynamically adjust the search constraint condition, so as to adjust the original U-turn strategy to obtain the target U-turn strategy, including: defining the multi-stage progressive degradation strategy; starting from the highest strategy of the multi-stage progressive degradation strategy, using a search algorithm to determine a U-turn path from the current position to the destination until the U-turn path meets the safety constraint condition.

[0016] Optionally, the original U-turn strategy is adjusted according to the influence degree to obtain a target U-turn strategy, including: predicting traffic flow in a predetermined time window in a future time period to obtain a traffic flow prediction result; dynamically adjusting a U-turn starting point of the target autonomous vehicle according to the traffic flow prediction result and the adaptive software constraint to obtain an adjusted U-turn starting point; and adjusting the original U-turn strategy according to the adjusted U-turn starting point to obtain the target U-turn strategy.

[0017] Optionally, the U-turn control method further includes: evaluating a U-turn path corresponding to the target U-turn strategy to obtain a U-turn path evaluation result during the process of controlling the target autonomous vehicle to perform the U-turn operation according to the target U-turn strategy; when the U-turn path evaluation result indicates that the U-turn path is in a failure state, re-planning a U-turn path according to path planning information corresponding to the U-turn path and current traffic flow to obtain a new U-turn path; and controlling the target autonomous vehicle to perform the U-turn operation according to a new target U-turn strategy.

[0018] According to another aspect of the embodiment of the present application, an U-turn control device of an autonomous vehicle is also provided, including: a first acquisition unit configured to acquire a first predicted trajectory of a traffic participant within a predetermined range centered on a target autonomous vehicle, and acquire a second predicted trajectory of the target autonomous vehicle performing a U-turn operation according to an original U-turn strategy; a second acquisition unit configured to, when it is determined that the first predicted trajectory and the second predicted trajectory have a coincident trajectory based on a comparison result of the first predicted trajectory and the second predicted trajectory, acquire a first arrival time of the traffic participant to each predicted coincident point on the coincident trajectory; an evaluation unit configured to evaluate the first arrival time to obtain an influence degree of the traffic participant on the U-turn operation; an adjustment unit configured to adjust the original U-turn strategy according to the influence degree to obtain a target U-turn strategy; and a control unit configured to control the target autonomous vehicle to perform the U-turn operation according to the target U-turn strategy.

[0019] Optionally, the second acquisition unit comprises: a first processing module, configured to collect first motion state information of the traffic participant within the predetermined range by the target automatic driving vehicle, and process the motion state information by using a trajectory prediction model to obtain the first predicted trajectory, wherein the trajectory prediction model is a model trained by machine learning using a plurality of sets of training data; and a first acquisition module, configured to acquire second motion state information of the target automatic driving vehicle, and generate the second predicted trajectory according to the second motion state information and the original turning strategy.

[0020] Optionally, the second acquisition unit comprises: a second acquisition module, configured to convert the first predicted trajectory and the second predicted trajectory into the same time-space coordinate system to obtain a third predicted trajectory and a fourth predicted trajectory; a first determination module, configured to determine a shortest distance between each point on the third predicted trajectory and the fourth predicted trajectory to obtain a shortest distance set; and a second determination module, configured to determine that the first predicted trajectory and the second predicted trajectory have a coincident trajectory when it is determined that there is a trajectory point less than a preset safety threshold in the shortest distance set.

[0021] Optionally, the adjustment unit comprises: a third acquisition module, configured to acquire a first arrival time of the traffic participant to each predicted coincidence point on the coincident trajectory while determining that the first predicted trajectory and the second predicted trajectory have a coincident trajectory based on a comparison result of the first predicted trajectory and the second predicted trajectory, and analyze reachability of a road network corresponding to a road where the target automatic driving vehicle is located and a bottleneck area of the road network by using graph theory connectivity detection; a third determination module, configured to determine a dynamic feature and an obstacle point of a traffic flow corresponding to the road according to the reachability and the bottleneck area; a fourth acquisition module, configured to obtain a traffic flow pattern of the traffic flow according to the dynamic feature and the obstacle point; and a fifth acquisition module, configured to generate an adaptive soft constraint according to the traffic flow pattern, and adjust the original turning strategy according to the adaptive soft constraint and the influence degree to obtain the target turning strategy.

[0022] Optionally, the adjustment unit comprises: a sixth acquisition module, configured to acquire environmental change information of the target automatic driving vehicle during driving of the target automatic driving vehicle; an update module, configured to update a traffic flow pattern and a weight of an adaptive soft constraint according to the environmental change information to obtain an updated traffic flow pattern and an updated adaptive soft constraint; and a first adjustment module, configured to adjust the original turning strategy according to the updated traffic flow pattern and the updated adaptive soft constraint to obtain the target turning strategy.

[0023] Optionally, the evaluation unit comprises: a seventh acquisition module, configured to acquire a starting time point at which the target automatic driving vehicle starts to perform the turning operation, and compare the starting time point with the first arrival time to calculate a time difference between the starting time point and the first arrival time; an eighth acquisition module, configured to acquire a type and a behavior of the traffic participant; and a fourth determination module, configured to determine the influence degree according to the type, the behavior and the time difference.

[0024] Optionally, the adjustment unit comprises: a second processing module, configured to delay the turning operation or adjust the second predicted trajectory when it is determined that the influence degree is greater than an influence degree threshold.

[0025] Optionally, the adjustment unit comprises: a second adjustment module, configured to dynamically adjust a search constraint condition by using a multi-level progressive degradation strategy to adjust the original turning strategy to obtain the target turning strategy, wherein the search constraint condition is used to constrain a turning path of the target automatic driving vehicle, and the multi-level progressive degradation strategy is used to adjust a margin of the search constraint condition.

[0026] Optionally, the second adjustment module comprises: a definition submodule, configured to define the multi-level progressive degradation strategy; and a determination submodule, configured to determine a turning path from a current position to a destination by using a search algorithm from a highest strategy of the multi-level progressive degradation strategy until the turning path meets a safety constraint condition.

[0027] Optionally, the adjustment unit comprises: a prediction module, configured to predict a traffic flow in a predetermined time window in a future time period to obtain a traffic flow prediction result; a third adjustment module, configured to dynamically adjust a turning starting point of the target automatic driving vehicle according to the traffic flow prediction result and an adaptive software constraint to obtain an adjusted turning starting point; and a fourth adjustment module, configured to adjust the original turning strategy according to the adjusted turning starting point to obtain the target turning strategy.

[0028] Optionally, the turning control apparatus further comprises: an evaluation unit, configured to evaluate a turning path corresponding to the target turning strategy in a process in which the target automatic driving vehicle performs the turning operation according to the target turning strategy to obtain a turning path evaluation result; a planning unit, configured to perform turning path planning again according to path planning information corresponding to the turning path and a current traffic flow when the turning path evaluation result indicates that the turning path is in a failure state to obtain a new turning path; and the control unit is further configured to control the target automatic driving vehicle to perform the turning operation according to a new target turning strategy.

[0029] According to another aspect of the embodiments of the present application, there is further provided an autonomous vehicle, which uses the U-turn control method of the autonomous vehicle according to any one of the preceding embodiments.

[0030] According to another aspect of the embodiments of the present application, there is further provided a computer readable storage medium comprising a stored program, wherein the program performs the U-turn control method of the autonomous vehicle according to any one of the preceding embodiments.

[0031] According to another aspect of the embodiments of the present application, there is further provided a processor configured to execute a program, wherein the program performs the U-turn control method of the autonomous vehicle according to any one of the preceding embodiments.

[0032] According to another aspect of the embodiments of the present application, there is further provided a computer program product comprising computer instructions, which, when executed by a processor, perform the U-turn control method of the autonomous vehicle according to any one of the preceding embodiments.

[0033] In the embodiments of the present application, the first predicted trajectory of the traffic participant within a predetermined range centered on the target autonomous vehicle is acquired, and the second predicted trajectory of the target autonomous vehicle performing the U-turn operation according to the original U-turn strategy is acquired; when it is determined that the first predicted trajectory and the second predicted trajectory have a coincident trajectory based on the comparison result of the first predicted trajectory and the second predicted trajectory, the first arrival time of the traffic participant to each predicted coincident point on the coincident trajectory is acquired; the first arrival time is evaluated to obtain the influence degree of the traffic participant on the U-turn operation; the original U-turn strategy is adjusted according to the influence degree to obtain a target U-turn strategy; and the target autonomous vehicle is controlled to perform the U-turn operation according to the target U-turn strategy. Through the above embodiments of the present application, the potential conflict point of the target autonomous vehicle and the traffic participant is predicted in advance, the influence of the traffic participant on the U-turn operation is evaluated, and the original U-turn strategy is adjusted in combination with the dynamic environment evaluation to realize the purpose of real-time path planning and optimization, so as to improve the safe driving capability of the autonomous vehicle in the complex traffic environment, reduce unnecessary waiting time, and improve the traffic efficiency of the overall traffic flow, thereby solving the technical problems that the U-turn path planning of the autonomous vehicle in the related art has obvious deficiencies in processing the influence of the dynamic traffic flow and adaptive adjustment, and limits the traffic capacity of the autonomous vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0034] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and serve to explain the principles of the application. In the drawings:

[0035] Figure 1A hardware structure block diagram of a mobile terminal of a U-turn control method of an autonomous vehicle according to an embodiment of the present application is shown in FIG. 1.

[0036] Figure 2 A flow chart of a U-turn control method of an autonomous vehicle according to an embodiment of the present application is shown in FIG. 2.

[0037] Figure 3 A schematic diagram of a U-turn control device of an autonomous vehicle according to an embodiment of the present application is shown in FIG. 3.

[0038] In the above drawings, the following reference signs are used:

[0039] 102, processor; 104, memory; 106, transmission device; 108, input / output device. DETAILED DESCRIPTION

[0040] In order to make the personnel in the technical field better understand the present application scheme, the technical scheme in the embodiment of the present application will be described clearly and completely below in combination with the drawings in the embodiment of the present application. Obviously, the described embodiment is only a part of the embodiment of the present application, not all. Based on the embodiment in the present application, all other embodiments obtained by the person skilled in the art without creative labor should belong to the protection scope of the present application.

[0041] It should be noted that the terms "first", "second" and the like in the specification and claims of the present application and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0042] As introduced in the background, the U-turn path planning of the autonomous vehicle in the related art has obvious deficiencies in processing the influence of dynamic traffic flow and adaptive adjustment, which limits the passing capacity of the autonomous vehicle. There is an urgent need for a more intelligent and flexible U-turn path planning method, which not only can accurately identify the dynamic traffic flow pattern, but also has an adaptive adjustment mechanism to cope with various complex traffic conditions and realize efficient and safe U-turn operation. Therefore, in the embodiments of the present application, a U-turn control method and device of an autonomous vehicle, an autonomous vehicle, a computer readable storage medium, a processor and a computer program product are provided.

[0043] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0044] The methods and embodiments provided in this invention can be executed on a mobile terminal, a computer terminal, or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal for a U-turn control method for an autonomous vehicle according to an embodiment of the present invention. Figure 1 As shown, a mobile terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. The mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal described above. For example, the mobile terminal may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0045] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the U-turn control method for autonomous vehicles in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the aforementioned networks may include wireless networks provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0046] Example 1

[0047] According to an embodiment of the present application, a method embodiment of a U-turn control method of an autonomous vehicle is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown.

[0048] Figure 2 is a flowchart of a U-turn control method of an autonomous vehicle according to an embodiment of the present application, as shown in Figure 2 The U-turn control method of the autonomous vehicle includes the following steps:

[0049] Step S202, acquiring a first predicted trajectory of a traffic participant within a predetermined range centered on a target autonomous vehicle, and acquiring a second predicted trajectory of the target autonomous vehicle performing a U-turn operation according to an original U-turn strategy.

[0050] Optionally, the target autonomous vehicle refers to a vehicle capable of running in an autonomous driving mode in a specific scenario (such as urban roads, highways), including but not limited to Robotaxi, automatic freight vehicle, etc. The target autonomous vehicle has the ability to perceive the surrounding environment, plan the path, and execute the driving operation.

[0051] Optionally, the traffic participant within the predetermined range refers to all dynamic objects within a certain range around the target autonomous vehicle (usually determined by the effective detection distance of the vehicle sensor), including all possible dynamic objects that may affect or be affected by the target vehicle, such as other vehicles, pedestrians, bicycles, etc. The motion state of these objects has a direct impact on the driving strategy of the target vehicle.

[0052] Optionally, the first predicted trajectory refers to the possible motion trajectory of the traffic participant within a future period of time generated by using the motion state information of the traffic participant collected by the vehicle sensor and combining the prediction algorithm. The first predicted trajectory is used to evaluate the potential impact of the traffic flow on the U-turn operation of the target autonomous vehicle. The second predicted trajectory refers to the predicted motion trajectory of the vehicle when performing the U-turn action, which is calculated based on the current motion state information (such as position, speed, acceleration, etc.) of the target autonomous vehicle and the preset U-turn strategy.

[0053] In this embodiment, the first motion state information of surrounding traffic participants (including vehicles, pedestrians, and other moving objects) such as position, speed, direction, etc. can be collected by using on-board sensors (such as cameras, radars, etc.). Then, through a pre-trained trajectory prediction model, the motion trajectories of these traffic participants in the future period of time are predicted to form the first predicted trajectory, which helps to predict the pattern of traffic flow and possible conflict points. At the same time, based on the current state information (such as position, speed) of the target autonomous vehicle and the predetermined original turning strategy, the trajectory of the vehicle itself when performing the turning operation is predicted to form the second predicted trajectory, which provides a basis for subsequent path evaluation and adjustment. This step relies on advanced machine learning and prediction algorithms, combined with real-time perception data, to lay the foundation for subsequent path optimization.

[0054] In step S204, when it is determined based on the comparison result of the first predicted trajectory and the second predicted trajectory that the first predicted trajectory and the second predicted trajectory exist overlapping trajectories, the first arrival time of the traffic participant to each predicted overlapping point on the overlapping trajectory is obtained.

[0055] Optionally, the overlapping trajectory refers to when the first predicted trajectory and the second predicted trajectory exist intersection in time and space, that is, at least one point of the two trajectories is at the same or similar spatial position at the same time point, and this intersection is defined as the overlapping trajectory. The existence of the overlapping trajectory indicates a potential collision risk.

[0056] Optionally, the first arrival time refers to the time point at which the surrounding traffic participant is expected to arrive at a certain predicted overlapping point on the turning operation overlapping trajectory. The first arrival time is a key parameter for evaluating the influence degree of the traffic participant on the turning operation.

[0057] In this embodiment, the predicted trajectories of the traffic participants and the autonomous vehicle are unified into the same time-space coordinate system, which facilitates the comparison between the trajectories. If the shortest distance between the trajectory points is less than a preset safety threshold, it is determined that there is an overlapping trajectory, indicating a potential collision risk.

[0058] In step S206, the first arrival time is evaluated to obtain the influence degree of the traffic participant on the turning operation.

[0059] Optionally, the influence degree refers to the possible influence degree of the traffic participant on the turning operation of the target autonomous vehicle, which can be comprehensively evaluated by factors such as time difference, relative speed, type of traffic participant, etc. The influence degree is used to determine whether the turning strategy needs to be adjusted and the specific way of adjustment.

[0060] Optionally, the time difference between the traffic participant and the predicted overlapping point of the target vehicle, as well as the type and behavior of the traffic participant, are evaluated to comprehensively judge their influence degree on the turning operation. The smaller the time difference, the greater the influence degree may be.

[0061] By the above steps, the first predicted trajectory and the second predicted trajectory are spatio-temporally matched to identify whether there is a trajectory overlap. When it is found that the two trajectories may intersect at a certain time point, the time point and its corresponding first arrival time, i.e., the time of the traffic participant reaching the predicted intersection point, are recorded. Subsequently, the influence degree of the traffic participant on the U-turn operation within this time window is evaluated, mainly considering factors such as traffic density and speed difference.

[0062] In step S208, the original U-turn strategy is adjusted according to the influence degree to obtain a target U-turn strategy.

[0063] Optionally, the original U-turn strategy refers to the preset U-turn operation execution scheme of the autonomous vehicle without considering the dynamic changes of the surrounding environment. The strategy may include U-turn path, speed control, acceleration and deceleration scheme, etc. The target U-turn strategy refers to the adjustment of the original U-turn strategy after evaluating the environmental influence (such as the motion state of the traffic participant, the analysis of the overlapping trajectory) to meet the new strategy with higher safety and efficiency requirements. The target U-turn strategy better adapts to the dynamic traffic environment and reduces the potential collision risk.

[0064] In this embodiment, adaptive soft constraints can be generated for traffic flow dense areas based on traffic flow patterns and dynamic characteristics. Soft constraints allow the path to be flexibly adjusted within a safe range to reduce interference with traffic flow while ensuring vehicle safety.

[0065] In step S210, the target autonomous vehicle is controlled to perform the U-turn operation according to the target U-turn strategy.

[0066] In this embodiment, during the vehicle driving process, environmental changes (such as the appearance of new obstacles or changes in traffic flow patterns) are monitored in real time, and the traffic flow patterns and the weights of soft constraints are updated accordingly to dynamically adjust the path planning strategy.

[0067] According to the influence degree of the traffic participant, it is determined whether and how to adjust the original U-turn strategy. If the influence degree is large, the U-turn operation time may need to be postponed, or an alternative U-turn path needs to be found to ensure the safety of the operation. The adjusted strategy is called the target U-turn strategy, and the target autonomous vehicle will perform the U-turn operation according to this strategy to realize functions such as avoidance, waiting, or selection of a more suitable path.

[0068] Notably, by the above-mentioned manner, useful information is extracted from dynamic traffic flow, and the U-turn strategy of the autonomous vehicle is adjusted in real time by predicting future trajectories and evaluating the impact of the dynamic environment. With the help of machine learning models to predict the motion trend of traffic participants, combined with graph analysis and space-time gridding technology, reachable paths and bottleneck areas in the traffic flow can be effectively identified, soft constraints can be generated and different weights can be set, so as to build a safer and more efficient U-turn path planning scheme.

[0069] Suppose a Robotaxi is driving on a busy urban road and is about to reach a U-turn location. The system first collects data on surrounding vehicles using on-board sensors and predicts their motion trajectories through a prediction model. At the same time, it also predicts the path of the Robotaxi if it performs a U-turn according to the original strategy. If analysis finds that the trajectory of another vehicle that is about to go straight coincides with the U-turn trajectory of the Robotaxi at time point t, and the straight vehicle is faster, the system considers that the straight vehicle has a greater impact on the U-turn operation by evaluating the time difference, vehicle type (the straight vehicle may be a bus or an emergency vehicle), and other information. At this time, the system can choose to postpone the U-turn and wait for the straight vehicle to pass, or adjust the U-turn path and choose a more open lane to reduce the risk of conflict. Finally, according to the adjusted target strategy, the Robotaxi successfully and safely completes the U-turn operation, and this process is almost completed without the passenger's awareness, reflecting a high level of automation and intelligence.

[0070] As can be seen from the above, in the embodiment of the application, the first predicted trajectory of the traffic participants within a predetermined range centered on the target autonomous vehicle can be obtained, and the second predicted trajectory of the target autonomous vehicle performing the U-turn operation according to the original U-turn strategy can be obtained; when it is determined that the first predicted trajectory and the second predicted trajectory exist a coincident trajectory based on the comparison result of the first predicted trajectory and the second predicted trajectory, the first arrival time of the traffic participants to each predicted coincident point on the coincident trajectory is obtained; the first arrival time is evaluated to obtain the influence degree of the traffic participants on the U-turn operation; the original U-turn strategy is adjusted according to the influence degree to obtain a target U-turn strategy; and the target autonomous vehicle is controlled to perform the U-turn operation according to the target U-turn strategy, which realizes the purpose of predicting potential conflict points between the target autonomous vehicle and the traffic participants in advance, evaluating the influence of the traffic participants on the U-turn operation, and adjusting the original U-turn strategy for real-time path planning and optimization, thereby achieving the technical effects of improving the safe driving capability of the autonomous vehicle in a complex traffic environment, reducing unnecessary waiting time, and improving the traffic efficiency of the overall traffic flow.

[0071] Therefore, through the above technical solutions provided by the embodiments of the present application, the technical problem that the U-turn path planning of the autonomous vehicle has obvious deficiencies in processing the influence of dynamic traffic flow and adaptive adjustment, and limits the passing capacity of the autonomous vehicle in the related art is solved.

[0072] According to the above embodiments of the present application, the first prediction trajectory of the traffic participants within the predetermined range centered on the target autonomous vehicle is obtained, and the second prediction trajectory when the target autonomous vehicle performs the U-turn operation according to the original U-turn strategy is obtained, including: collecting the first motion state information of the traffic participants within the predetermined range by the target autonomous vehicle, and processing the motion state information by using a trajectory prediction model to obtain the first prediction trajectory, wherein the trajectory prediction model is a model trained by machine learning using a plurality of sets of training data; obtaining the second motion state information of the target autonomous vehicle, and generating the second prediction trajectory according to the second motion state information and the original U-turn strategy.

[0073] In this embodiment, the motion state information of the surrounding traffic participants, including but not limited to position, speed, acceleration, etc., can be collected in real time by sensors on the autonomous vehicle, such as lidar and camera. The trajectory prediction model trained in advance is used to process the motion state of the collected traffic participants to generate their prediction trajectories in the future period of time. The model may be based on deep learning technology and trained by a large amount of historical data, and can predict the possible behaviors of different traffic participants. At the same time, based on the current state of the target autonomous vehicle (such as position, speed, acceleration, etc.) and the original U-turn strategy, the prediction trajectory of the vehicle when performing the U-turn is calculated.

[0074] The trajectory prediction model based on machine learning can learn the dynamic behavior patterns of traffic participants from historical data to predict future motion trajectories. At the same time, the real-time perception data is combined with the pre-trained model output to generate the first prediction trajectory, and the vehicle state information is used to generate the second prediction trajectory.

[0075] The above prediction of dynamic traffic environment by using advanced machine learning model can improve the accuracy and adaptability of path planning. By predicting the future behavior of surrounding traffic participants, potential conflict points can be identified earlier, providing timely information for path adjustment.

[0076] According to the above embodiment of the present application, the determination of the existence of the coincident trajectory between the first predicted trajectory and the second predicted trajectory based on the comparison result of the first predicted trajectory and the second predicted trajectory comprises: converting the first predicted trajectory and the second predicted trajectory into the same time-space coordinate system to obtain a third predicted trajectory and a fourth predicted trajectory; determining the shortest distance between each point on the third predicted trajectory and the fourth predicted trajectory to obtain a shortest distance set; and determining the existence of the coincident trajectory between the first predicted trajectory and the second predicted trajectory when there is a trajectory point in the shortest distance set that is less than a preset safety threshold.

[0077] In this embodiment, the first predicted trajectory and the second predicted trajectory are unified into the same time-space coordinate system for comparison. The shortest distance between the converted trajectories is calculated point by point to obtain a series of distance values for evaluating the closeness of the two trajectories. If the shortest distance of one or more trajectory points is less than a preset safety threshold, it is determined that the first predicted trajectory and the second predicted trajectory have a coincident trajectory.

[0078] By converting different trajectories into the same reference system, the rationality of the comparison is ensured, and by setting a safety distance threshold, the intersection degree of the two trajectories is quantitatively evaluated to determine whether there is a potential collision risk.

[0079] The above steps can accurately identify the points that may collide with surrounding traffic participants during the U-turn operation, providing a basis for subsequent strategy adjustment. By identifying the coincident trajectory early, invalid attempts in the U-turn operation can be avoided, the waiting time is reduced, and the traffic efficiency is improved.

[0080] According to the above embodiment of the present application, the original U-turn strategy is adjusted according to the influence degree to obtain a target U-turn strategy, comprising: when it is determined that the first predicted trajectory and the second predicted trajectory have a coincident trajectory based on the comparison result of the first predicted trajectory and the second predicted trajectory, the first arrival time of the traffic participant object to each predicted coincident point on the coincident trajectory is obtained, and the reachability of the road network corresponding to the road where the target autonomous vehicle is located and the bottleneck area of the road network are analyzed using graph theory connectivity detection; the dynamic characteristics and obstacle points of the traffic flow corresponding to the road are determined according to the reachability and the bottleneck area; the traffic flow mode of the traffic flow is obtained according to the dynamic characteristics and the obstacle points; the adaptive soft constraint is generated according to the traffic flow mode, and the original U-turn strategy is adjusted according to the adaptive soft constraint and the influence degree to obtain the target U-turn strategy.

[0081] In this embodiment, the first arrival time of the traffic participant to the predicted intersection point can be obtained and compared with the turning start time point of the target autonomous vehicle to calculate the time difference. The type (such as vehicle, pedestrian, bicycle, etc.) and behavior (such as speed, direction, whether it is an emergency vehicle, etc.) of the traffic participant are considered to quantify the degree of their influence on the turning operation. And according to the evaluation of the influence degree, the original turning strategy is adjusted to reduce or avoid potential collision risks.

[0082] By combining the time difference and the type and behavior of the traffic participant, the influence of the turning operation is comprehensively evaluated. Based on the evaluation results, an adaptive turning strategy suitable for the current traffic conditions is generated to ensure safety and efficiency.

[0083] That is, in the embodiment of the present application, traffic flow pattern recognition is achieved through trajectory intersection analysis and graph connectivity detection. Among them, trajectory intersection analysis: predicting the trajectories of surrounding traffic participants, analyzing the potential intersection with the turning trajectory of the autonomous vehicle; graph connectivity detection: abstracting the road network as a graph structure, identifying the accessibility and bottleneck area of traffic flow by analyzing the connectivity of nodes and edges.

[0084] According to the specific situation, the turning strategy is dynamically adjusted, so that the path planning is more personalized and reasonable. By avoiding collision with high-risk traffic participants, both safety and unnecessary waiting are improved, and traffic efficiency is improved.

[0085] According to the above embodiment of the present application, the original turning strategy is adjusted according to the influence degree to obtain the target turning strategy, including: obtaining the environmental change information of the target autonomous vehicle during the driving process of the target autonomous vehicle; updating the traffic flow pattern and the weight of the adaptive soft constraint according to the environmental change information, obtaining the updated traffic flow pattern and the updated adaptive soft constraint; adjusting the original turning strategy according to the updated traffic flow pattern and the updated adaptive soft constraint to obtain the target turning strategy.

[0086] During driving, the surrounding environment change information of the target autonomous vehicle is continuously collected, including newly appearing traffic participants, traffic signal changes, etc. According to the environmental change information, the weight of the traffic flow pattern and the adaptive soft constraint is updated in real time to adapt to the new traffic conditions. Combined with the updated information, the original turning strategy is adjusted to obtain a new target turning strategy.

[0087] Using sensors to continuously monitor environmental changes ensures the real-time and accuracy of path planning information. According to the real-time monitoring of environmental changes, the weight of the soft constraint is dynamically adjusted to more accurately reflect the influence of the current traffic conditions on path planning.

[0088] Based on the identified traffic flow pattern, adaptive soft constraint generation is realized through the SoftConstraintGrid class, a soft constraint segment is added to the traffic flow dense area and the corresponding weight is set. Soft constraint allows flexible adjustment of the path according to the traffic condition under the premise of ensuring safety, and a penalty term for violating the soft constraint is introduced in the optimization objective to balance safety and flexibility.

[0089] Specifically, first, the predicted trajectories of surrounding traffic participants are traversed, and it is judged whether there is potential intersection with the target trajectory of the autonomous vehicle through trajectory intersection analysis, and the reachability and bottleneck area of the road network are analyzed by using graph connectivity detection. When it is detected that there is a trajectory intersection and the road network is connected, the system adds a corresponding soft constraint segment to the traffic flow dense area, and sets appropriate weights according to the density of the traffic flow, realizes accurate identification of the traffic flow pattern and adaptive constraint generation.

[0090] In order to reduce the computational complexity, the system adopts a three-level optimization strategy: first, the continuous trajectory is discretized into time-space grid points through spatiotemporal rasterization, reducing the dimension of the trajectory data; second, spatiotemporal correlation pruning is applied to filter out the trajectories of traffic participants whose time difference with the turning trajectory of the autonomous vehicle is more than 3 seconds or the spatial distance is greater than 50 meters; finally, an incremental update mechanism is adopted, only the trajectories of traffic participants whose position changes exceed the threshold are recalculated, avoiding repeated processing of full data. Through these optimization measures, the average calculation time is reduced by 40%, meeting the real-time requirements of the autonomous driving system.

[0091] In the above manner, the traffic environment can be continuously adapted, improving the flexibility and real-time response capability of the path planning. By dynamically adjusting the weight of the soft constraint, conflicts with other traffic participants can be more effectively avoided to reduce conflicts, improving the overall traffic efficiency and safety.

[0092] According to the above embodiment of the application, the first arrival time is evaluated to obtain the influence degree of the traffic participant object on the turning operation, including: acquiring a starting time point at which the target autonomous vehicle starts to perform the turning operation, and comparing the starting time point with the first arrival time to calculate a time difference value between the starting time point and the first arrival time; acquiring the type and behavior of the traffic participant object; determining the influence degree according to the type, the behavior and the time difference value.

[0093] In this embodiment, based on the predicted trajectories of the vehicle and surrounding traffic participants, the time difference of both reaching the same location is calculated. By analyzing the time difference and the characteristics of the traffic participants such as type, speed, size, etc., the degree of influence on the vehicle's U-turn operation is quantified.

[0094] By establishing a model or algorithm, the time difference and participant characteristics are taken as input, and the potential impact on the U-turn operation is output. According to the evaluation results of the influence degree, the U-turn strategy of the vehicle is dynamically adjusted to ensure safety and traffic efficiency.

[0095] The above-mentioned manner can automatically analyze the influence of surrounding traffic on the U-turn operation and intelligently adjust the strategy to cope with different traffic conditions. By quantifying the influence, direct conflict between the U-turn operation and high-risk traffic flow can be avoided, improving overall safety.

[0096] According to the above-mentioned embodiments of the present application, the original U-turn strategy is adjusted according to the influence degree to obtain the target U-turn strategy, including: delaying the U-turn operation or adjusting the second predicted trajectory when the influence degree is greater than the influence degree threshold.

[0097] In this embodiment, in the SoftConstraintGrid class, the soft constraint section of the traffic flow dense area is defined, and different weights are set. In the path planning optimization objective, a penalty term for soft constraint violation is introduced to guide the vehicle to avoid these areas as much as possible.

[0098] According to the density of traffic flow and the predicted dynamic adjustment of the weight of soft constraint, the flexibility and safety balance of path planning are realized. Through the penalty term in the optimization algorithm, the vehicle is guided to choose a path that does not violate the soft constraint as much as possible under the premise of meeting the safety conditions.

[0099] Through the above-mentioned manner, the soft constraint allows a certain degree of path deviation, but ensures not to violate the safety principle through the penalty mechanism, improving the flexibility and safety of path planning. The soft constraint mechanism helps the autonomous vehicle better adapt to complex road conditions with dense traffic flow and reduces the disturbance to normal traffic.

[0100] According to the above-mentioned embodiments of the present application, the original U-turn strategy is adjusted according to the influence degree to obtain the target U-turn strategy, including: adopting a multi-level progressive degradation strategy to dynamically adjust the search constraint condition to adjust the original U-turn strategy to obtain the target U-turn strategy, wherein the search constraint condition is used to constrain the U-turn path of the target autonomous vehicle, and the multi-level progressive degradation strategy is used to adjust the margin of the search constraint condition.

[0101] In this embodiment, each level corresponds to different search constraints to define the policy hierarchy, from the most stringent level 0 to the most lenient level 7. In addition, the destination_level parameter is dynamically adjusted according to the number of consecutive failures of path planning, gradually relaxing the constraints until a feasible path is found.

[0102] When no feasible path is found at the current level, the system automatically downgrades to try more lenient constraints, expanding the search space for policy downgrade. Starting from the highest level, the system gradually relaxes constraints such as lateral error, heading error, etc., until a successful turnaround path is planned.

[0103] It should be noted that in the ModifyProblem function in the embodiment of the present application, the eight-level progressive downgrade strategy is implemented through the destination_level parameter, and the search constraints are dynamically adjusted according to the success rate of path planning. The eight-level progressive downgrade strategy is shown in Table 1 below.

[0104] Table 1

[0105]

[0106] Table 1 above shows the processing strategy and application scenario at each level. In the ModifyProblem function of reversible_solver_kturn.txt, the eight-level progressive downgrade strategy is implemented through the destination_level parameter. The system dynamically adjusts the value of destination_level according to the number of consecutive failures of path planning, gradually relaxing the constraints from level 0 to level 7. The target curve corresponds to the rightmost lane by default, and the standby curves correspond to the left lanes in turn. After completing the U-turn, the vehicle prefers to enter the rightmost lane. When the traffic flow does not allow safe entry into the rightmost lane, the system automatically downgrades to select the inner lane to avoid crossing the traffic flow and improve traffic efficiency.

[0107] Level 0: Prefer to use the target curve (rightmost lane) and set strict lateral and heading error ranges. Path planning follows all preset safety criteria and rules, including minimum safety distance, maximum speed limit, specific lane and heading preference, etc. The constraints of the search algorithm are the tightest, ensuring the highest level of safety. The constraints of the initial state and target state are maximized, and the vehicle must strictly follow the initial position and speed, target position, speed, and heading. The violation of adaptive soft constraints is punished the most heavily, ensuring that the vehicle prioritizes avoiding traffic dense areas;

[0108] Level 1: Prioritize using target curve (rightmost lane), moderately relax lateral error range to ±1.5 meters, and heading error range to ±20 degrees. Slightly relax certain safety constraints, such as increasing the tolerance for lateral deviation to ±1.5 meters and heading deviation to ±20 degrees. Slightly relax the initial and target state constraints to allow the vehicle to adjust the initial and target positions within a certain range. Moderately reduce the penalty of adaptive soft constraints while still maintaining sensitivity to traffic flow patterns;

[0109] Level 2: Use all alternative curves (including left lanes), maintain error ranges; use all available alternative paths, including lanes other than the target lane, as long as they are considered feasible. Further relax the tolerance range for lateral and heading deviations. Further reduce the penalty of adaptive soft constraints to allow the vehicle to make more dynamic adjustments in traffic flow;

[0110] Level 3: Do not prioritize using target curves, dynamically select appropriate curves based on traffic flow; no longer prioritize a specific target lane, but dynamically select the safest and most efficient lane based on real-time traffic conditions. The tolerance range for lateral and heading deviations remains the same as level 2, but places more emphasis on the continuity and stability of the path. Adjust the weight of adaptive soft constraints to better reflect the actual impact of traffic flow;

[0111] Level 4: Do not prioritize using target curves, further relax error ranges. On the basis of level 3, further relax the tolerance range for lateral and heading deviations to increase the flexibility of path planning. Focus on the overall efficiency of path planning, moderately sacrificing some local safety requirements;

[0112] Level 5: Use all curves while relaxing error ranges. Use all available road resources and lanes, including but not limited to the target lane, allowing the vehicle to make more significant maneuvers. Further increase the tolerance range for lateral and heading deviations to improve the degree of freedom of path planning. Basically cancel the penalty of adaptive soft constraints, focusing on the feasibility and continuity of the path;

[0113] Level 6: Abandon using target curves, directly use target state for path planning. Abandon the preference for a specific target lane, and the vehicle can choose any lane to continue driving after the turn. Directly use the target state for path planning without strictly limiting the target position and heading. Reduce the impact of adaptive soft constraints on path planning to the minimum, focusing on the physical feasibility of the path. Completely relax all search constraints, and the vehicle can take any possible action as long as it can reach the target position from the current position. Adjust the target position to provide the vehicle with more open operating space, moving the target position forward by a certain distance in the direction of travel and appropriately offsetting it laterally. All soft and hard constraints are in the most relaxed state, and path planning relies entirely on physical feasibility and real-time traffic conditions;

[0114] Level 7: Adjust the target position by moving it 5.0 meters forward along the original heading and 3.0 meters to the left, while setting the most lenient constraints. Completely relax all search constraints, allowing the vehicle to take any possible action as long as it can reach the target position from the current location. Adjust the target position to provide more open operating space for the vehicle, moving it forward in the direction of travel and offsetting it laterally as appropriate. All soft and hard constraints are set to their most lenient settings, and path planning relies entirely on physical feasibility and real-time traffic conditions.

[0115] The transition between each level of strategy is based on the number of consecutive path planning failures or monitoring of environmental changes. When the upper-level strategy fails to find a safe and feasible path, it automatically downgrades to the next lower-level strategy, gradually expanding the search space until a viable path is found. This strategy ensures that the vehicle can still find and execute a reasonable path even in extremely complex traffic environments, thereby improving the success rate of path planning and the passability of the autonomous vehicle.

[0116] It is worth noting that when the high-level strategy fails to find a feasible path, the system automatically downgrades to a lower-level strategy, gradually relaxing the constraints to expand the search space and improve the success rate of path planning in complex traffic environments.

[0117] By gradually relaxing the constraints, the success rate of planning a U-turn path in complex traffic conditions is improved. The use of multi-level strategies increases the flexibility of path planning, enabling the identification of the most suitable U-turn solution in various traffic environments.

[0118] According to the above embodiments of the present application, a multi-level progressive degradation strategy is adopted to dynamically adjust the search constraints, thereby adjusting the original U-turn strategy to obtain a target U-turn strategy, which includes: defining a multi-level progressive degradation strategy; starting from the highest strategy of the multi-level progressive degradation strategy, using a search algorithm to determine a U-turn path from the current location to the destination until the U-turn path meets the safety constraints.

[0119] In this embodiment, based on the current vehicle position and travel direction, the ideal U-turn starting state is calculated. Considering the vehicle dynamics constraints, the starting state is dynamically adjusted to avoid dense traffic areas or potential obstacles. Combined with the vehicle dynamics characteristics, the appropriate starting state is evaluated and selected. By analyzing the surrounding traffic flow pattern, potential high-risk areas are predicted and avoided to safely start the U-turn operation.

[0120] By optimizing the starting state, potential collisions during the U-turn operation can be effectively avoided, improving safety. Selecting a more optimal starting state helps to plan a more efficient and safer U-turn path.

[0121] According to the above embodiment of the present application, the original turning strategy is adjusted according to the influence degree to obtain a target turning strategy, comprising: predicting the traffic flow of a predetermined time window in a future time period to obtain a traffic flow prediction result; dynamically adjusting the turning starting point of the target autonomous vehicle according to the traffic flow prediction result and the adaptive software constraint to obtain an adjusted turning starting point; and adjusting the original turning strategy according to the adjusted turning starting point to obtain the target turning strategy.

[0122] In this embodiment, the effectiveness of the current planning path is periodically evaluated to monitor changes in traffic conditions. When a change in traffic conditions or path failure is detected, path re-planning is triggered, and historical planning results are combined for optimization.

[0123] Here, the sensor is used to continuously monitor the changes in the environment around the vehicle, ensuring the real-time and effectiveness of the data. The previous path planning information is used to reduce unnecessary calculations and improve path refresh efficiency.

[0124] Specifically, the forward-looking starting state optimization can be implemented in the ModifyProblem function of reversible_solver_kturn.txt. The system calculates the ideal starting state in the future period of time according to the current position, speed and surrounding environment of the vehicle. First, the safe parking distance is calculated according to the current speed and comfortable deceleration of the vehicle, and then the ideal starting point without collision is searched within the distance range. The system preferentially selects a position in advance for a period of time as the starting state, rather than using the current position, to improve the forward-looking and success rate of path planning. If the ideal starting point has collision risk, the system will automatically adjust the starting position until a safe and feasible starting state is found.

[0125] It should be noted that the forward-looking starting state optimization is implemented in the ModifyProblem function, and the starting state of the path search is dynamically adjusted according to the current position, speed and surrounding environment of the vehicle: 1) the ideal starting state is calculated based on the current vehicle position and driving direction; 2) the feasibility of the starting state is ensured considering the vehicle dynamics constraints; 3) the starting state is adjusted to avoid traffic-intensive areas according to the traffic flow conditions.

[0126] This method generates a set of forward-looking path candidates by continuously searching for target state space 1.5 seconds in the future during vehicle motion. These candidate paths maintain dynamic continuity with the current execution path, ensuring seamless switching between new and old paths when the real-time path refresh mechanism in embodiment 4 is triggered, avoiding sudden changes and shocks in vehicle motion.

[0127] Through the above-mentioned manner, the environment changes can be quickly responded to, the real-time path planning can be realized, and the vehicle passing efficiency can be improved. The historical path is combined to realize smooth transition of new and old paths, and the ride comfort and driving stability are improved.

[0128] According to the above embodiments of the present application, the U-turn control method further comprises: during the process of controlling the target automatic driving vehicle to perform the U-turn operation according to the target U-turn strategy, evaluating the U-turn path corresponding to the target U-turn strategy to obtain a U-turn path evaluation result; when the U-turn path evaluation result indicates that the U-turn path is in a failure state, re-planning the U-turn path according to the path planning information corresponding to the U-turn path and the current traffic flow to obtain a new U-turn path; and controlling the target automatic driving vehicle to perform the U-turn operation according to the new target U-turn strategy.

[0129] In this embodiment, the above-mentioned technology is applied to the U-turn scene in complex urban road conditions, and is particularly suitable for vehicles such as Robotaxi that need to frequently U-turn in busy traffic environments. By comprehensively using traffic flow identification, soft constraint generation, multi-level degradation strategy and other technologies, the intelligent level of U-turn operation is improved.

[0130] Here, the characteristics of the current U-turn scene, such as traffic density, intersection complexity, etc., are identified, and the corresponding scene is matched. Ensure that each technical module (such as prediction, path planning, state optimization) works together to form a complete and effective U-turn operation control method.

[0131] For example, the system periodically checks the validity of the current path and evaluates the changes in traffic conditions. When a significant change in traffic conditions is detected, such as the appearance of a new obstacle or a change in traffic flow patterns, the system triggers path re-planning. During the refresh process, the system makes full use of historical planning results to avoid unnecessary repeated calculations, improving the efficiency of path updating. At the same time, the system will dynamically adjust the refresh interval according to the frequency of changes in traffic conditions, reducing the refresh frequency when the traffic conditions are stable to reduce the computational burden, and increasing the refresh frequency when the traffic conditions change rapidly to ensure the real-time and safety of the path.

[0132] Specifically, in the FindSolution function of reversible_solver_general_base.txt, the path refresh mechanism is implemented. The system periodically checks the validity of the current path and evaluates the changes in traffic conditions. When a significant change in traffic conditions is detected, such as the appearance of a new obstacle or a change in traffic flow patterns, the system triggers path re-planning. During the refresh process, the system makes full use of historical planning results to avoid unnecessary repeated calculations, improving the efficiency of path updating. At the same time, the system will dynamically adjust the refresh interval according to the frequency of changes in traffic conditions, reducing the refresh frequency when the traffic conditions are stable to reduce the computational burden, and increasing the refresh frequency when the traffic conditions change rapidly to ensure the real-time and safety of the path.

[0133] It should be noted that the real-time path refreshing mechanism herein can achieve real-time path refreshing through path updating logic in the FindSolution function: 1) periodically check the validity of the current path, and evaluate the traffic condition changes; 2) trigger path re-planning when significant changes in traffic conditions are detected; 3) combine historical planning results to improve path continuity and comfort.

[0134] By combining with the above-mentioned proactive initial state optimization, the system can quickly select the optimal solution from the pre-computed proactive path candidate set during path refreshing, and realize smooth transition of new and old paths. This cooperative mechanism controls the impact (jerk) of path switching within 0.5 m / s³, ensuring ride comfort and driving stability.

[0135] Specifically, in order to ensure seamless switching of new and old paths when the real-time path refreshing mechanism is triggered, and to avoid sudden changes and shocks in vehicle motion, the system needs to select a future point as the starting position for searching, and from this starting point, the new solution is searched and adopted, and the old solution is still used for the distance from the current position to the future starting position to ensure smooth transition. The system calculates the ideal starting state in the future period of time according to the current position, speed and surrounding environment of the vehicle. First, the safe parking distance is calculated according to the current speed and comfortable deceleration of the vehicle, and then the ideal starting point without collision is searched within the distance range to improve the proactivity and success rate of path planning. If the ideal starting point has collision risk, the system will automatically adjust the starting position forward and backward until a safe and feasible starting state is found.

[0136] The method generates a set of proactive path candidates available to the ego vehicle by continuously searching the target state space 1.5 seconds in the future during vehicle motion. These candidate paths maintain dynamic continuity with the currently executed path, ensuring seamless switching of new and old paths when the real-time path refreshing mechanism is triggered, and avoiding sudden changes and shocks in vehicle motion.

[0137] In this way, the problem of low success rate and poor efficiency of U-turn operation in the prior art when facing complex traffic environment is solved. The method is particularly suitable for complex urban road conditions and can significantly improve the traffic capacity of Robotaxi and other autonomous vehicles.

[0138] Through the above technical solutions provided by the embodiments of the present application, when identifying traffic flow, the uncertainty of trajectory prediction needs to be considered comprehensively, and a reasonable safety margin is set. The trigger threshold of the degradation strategy can be dynamically adjusted according to the actual traffic conditions to avoid frequent degradation; in addition, the initial state optimization needs to balance the proactivity and computational complexity, and the state after 1.5 seconds is usually predicted; the path refreshing interval needs to be dynamically adjusted according to the traffic conditions, and the interval is shortened in high-speed or congested scenes.

[0139] In addition, by traffic flow recognition and adaptive degradation strategy, the success rate and execution quality of the automatic driving vehicle U-turn path planning in complex scenarios are significantly improved, which has the following beneficial effects: 1) By the eight-level progressive degradation strategy, the success rate of path planning in complex traffic environment is improved, especially in the dense traffic area. When the high-level strategy cannot find a feasible path, the system automatically degrades to a low-level strategy with more relaxed constraints, expands the search space, and increases the possibility of finding an effective path. 2) The forward-looking initial state optimization shortens the path length and makes the curvature change smoother, improves the ride comfort and driving safety. The system selects the best initial state by predicting the future position and speed of the vehicle, reduces unnecessary steering and adjustment, and generates a more optimal path trajectory. 3) The adaptive soft constraint generation mechanism enables the system to respond to traffic flow changes in real time and maintain stable performance in complex scenarios such as peak hours. By adding soft constraints to traffic-intensive areas, the system can flexibly adjust the path under the premise of safety, balancing safety and traffic efficiency. 4) Combined with historical planning results and real-time refresh mechanism, the average calculation time is reduced, meeting the real-time requirements of the automatic driving system. The system reuses existing path information and dynamically adjusts the refresh frequency to avoid unnecessary repeated calculations and improve overall efficiency.

[0140] It should be noted that, for each of the above method embodiments, in order to simply describe, each is expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the order of the described actions, because according to the present application, certain steps can be performed in other order or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0141] From the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be realized by means of software and the necessary general hardware platform, and of course it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, or network device, etc.) execute the methods described in each embodiment of the present application.

[0142] Embodiment 2

[0143] According to the embodiments of the present application, a U-turn control device of an automatic driving vehicle for implementing the U-turn control method of the automatic driving vehicle is also provided, Figure 3is a schematic diagram of a U-turn control device of an autonomous vehicle according to an embodiment of the present application, as shown in Figure 3 The U-turn control device of the autonomous vehicle includes a first acquisition unit 301, a second acquisition unit 303, an evaluation unit 305, an adjustment unit 307, and a control unit 309. The device will be described below.

[0144] The first acquisition unit 301 is configured to acquire a first predicted trajectory of a traffic participant within a predetermined range centered on a target autonomous vehicle, and to acquire a second predicted trajectory of the target autonomous vehicle performing a U-turn operation according to an original U-turn strategy.

[0145] The second acquisition unit 303 is configured to, when it is determined based on a comparison result of the first predicted trajectory and the second predicted trajectory that the first predicted trajectory and the second predicted trajectory have a coinciding trajectory, acquire a first arrival time of the traffic participant to each predicted coinciding point on the coinciding trajectory.

[0146] The evaluation unit 305 is configured to evaluate the first arrival time to obtain an influence degree of the traffic participant on the U-turn operation.

[0147] The adjustment unit 307 is configured to adjust the original U-turn strategy according to the influence degree to obtain a target U-turn strategy.

[0148] The control unit 309 is configured to control the target autonomous vehicle to perform the U-turn operation according to the target U-turn strategy.

[0149] It should be noted that the above-mentioned first acquisition unit 301, second acquisition unit 303, evaluation unit 305, adjustment unit 307, and control unit 309 correspond to steps S202 to S210 in the above embodiment, and the five units have the same instances and application scenarios as the corresponding steps, but are not limited to the content disclosed in the above embodiment.

[0150] From the above, in the scheme described in the above embodiments of the present application, the first acquisition unit can be used to acquire the first predicted trajectory of the traffic participants in the predetermined range centered on the target autonomous vehicle, and simultaneously acquire the second predicted trajectory when the target autonomous vehicle performs the turning operation according to the original turning strategy; the second acquisition unit is used to acquire the first arrival time of the traffic participants to each predicted intersection point on the intersection trajectory when it is determined that the first predicted trajectory and the second predicted trajectory have the intersection trajectory based on the comparison result of the first predicted trajectory and the second predicted trajectory; the evaluation unit is used to evaluate the first arrival time to obtain the influence degree of the traffic participants on the turning operation; the adjustment unit is used to adjust the original turning strategy according to the influence degree to obtain the target turning strategy; and the control unit is used to control the target autonomous vehicle to perform the turning operation according to the target turning strategy, so as to achieve the purpose of predicting the potential conflict point of the target autonomous vehicle and the traffic participants in advance, evaluating the influence on the turning operation, and adjusting the original turning strategy for real-time path planning and optimization in combination with dynamic environment evaluation, thereby improving the safe driving capability of the autonomous vehicle in the complex traffic environment, reducing unnecessary waiting time, and improving the traffic efficiency of the overall traffic flow.

[0151] Therefore, by using the above technical scheme provided by the embodiments of the present application, the technical problem that the turning path planning of the autonomous vehicle in the related art has obvious deficiencies in processing the influence of dynamic traffic flow and self-adaptive adjustment, and limits the traffic capacity of the autonomous vehicle is solved.

[0152] Optionally, the second acquisition unit comprises: a first processing module, configured to collect first motion state information of the traffic participants in the predetermined range through the target autonomous vehicle, and process the motion state information by using a trajectory prediction model to obtain the first predicted trajectory, wherein the trajectory prediction model is a model trained by machine learning using a plurality of sets of training data; and a first acquisition module, configured to acquire second motion state information of the target autonomous vehicle, and generate a second predicted trajectory according to the second motion state information and the original turning strategy.

[0153] Optionally, the second acquisition unit comprises: a second acquisition module, configured to convert the first predicted trajectory and the second predicted trajectory into the same time-space coordinate system to obtain a third predicted trajectory and a fourth predicted trajectory; a first determination module, configured to determine the shortest distance between each point on the third predicted trajectory and the fourth predicted trajectory to obtain a shortest distance set; and a second determination module, configured to determine that the first predicted trajectory and the second predicted trajectory have the intersection trajectory when there is a trajectory point less than a preset safety threshold in the shortest distance set.

[0154] Optionally, the adjusting unit comprises: a third obtaining module, configured to, when it is determined that the first predicted trajectory and the second predicted trajectory have a coincident trajectory based on a comparison result of the first predicted trajectory and the second predicted trajectory, obtain a first arrival time of the traffic participant to each predicted coincident point on the coincident trajectory, and simultaneously analyze reachability of a road network corresponding to a road where the target autonomous vehicle is located and a bottleneck area of the road network by using graph theory connectivity detection; a third determining module, configured to determine a dynamic characteristic and an obstacle point of a traffic flow corresponding to the road according to the reachability and the bottleneck area; a fourth obtaining module, configured to obtain a traffic flow mode of the traffic flow according to the dynamic characteristic and the obstacle point; and a fifth obtaining module, configured to generate an adaptive soft constraint according to the traffic flow mode, and adjust the original U-turn strategy according to the adaptive soft constraint and the influence degree to obtain the target U-turn strategy.

[0155] Optionally, the adjusting unit comprises: a sixth obtaining module, configured to obtain environmental change information of the target autonomous vehicle during driving of the target autonomous vehicle; an updating module, configured to update a weight of the traffic flow mode and the adaptive soft constraint according to the environmental change information to obtain an updated traffic flow mode and an updated adaptive soft constraint; and a first adjusting module, configured to adjust the original U-turn strategy according to the updated traffic flow mode and the updated adaptive soft constraint to obtain the target U-turn strategy.

[0156] Optionally, the evaluating unit comprises: a seventh obtaining module, configured to obtain a starting time point at which the target autonomous vehicle starts to perform the U-turn operation, and compare the starting time point with the first arrival time to calculate a time difference between the starting time point and the first arrival time; an eighth obtaining module, configured to obtain a type and a behavior of the traffic participant; and a fourth determining module, configured to determine the influence degree according to the type, the behavior and the time difference.

[0157] Optionally, the adjusting unit comprises: a second processing module, configured to, when it is determined that the influence degree is greater than an influence degree threshold, postpone the U-turn operation or adjust the second predicted trajectory.

[0158] Optionally, the adjusting unit comprises: a second adjusting module, configured to adopt a multi-level progressive degradation strategy to dynamically adjust a search constraint condition to adjust the original U-turn strategy to obtain the target U-turn strategy, wherein the search constraint condition is used to constrain a U-turn path of the target autonomous vehicle, and the multi-level progressive degradation strategy is used to adjust a margin of the search constraint condition.

[0159] Optionally, the second adjusting module comprises: a defining submodule, configured to define the multi-level progressive degradation strategy; and a determining submodule, configured to determine the U-turn path from the current position to the destination by using a search algorithm from a highest strategy of the multi-level progressive degradation strategy until the U-turn path meets a safety constraint condition.

[0160] Optionally, the adjusting unit comprises: a prediction module, configured to predict a traffic flow in a predetermined time window in a future time period to obtain a traffic flow prediction result; a third adjusting module, configured to dynamically adjust the turning start point of the target autonomous vehicle according to the traffic flow prediction result and the adaptive software constraint to obtain an adjusted turning start point; and a fourth adjusting module, configured to adjust the original turning strategy according to the adjusted turning start point to obtain the target turning strategy.

[0161] Optionally, the turning control device further comprises: an evaluation unit, configured to evaluate the turning path corresponding to the target turning strategy in a process in which the target autonomous vehicle performs the turning operation according to the target turning strategy to obtain a turning path evaluation result; a planning unit, configured to perform turning path planning again according to the path planning information corresponding to the turning path and the current traffic flow to obtain a new turning path when the turning path evaluation result indicates that the turning path is in a failure state; and a control unit, further configured to control the target autonomous vehicle to perform the turning operation according to the new target turning strategy.

[0162] According to another aspect of the embodiments of the present application, there is also provided an autonomous vehicle, which uses the turning control method of the autonomous vehicle according to any one of the above.

[0163] According to another aspect of the embodiments of the present application, there is also provided a processor, which is used to run a program, wherein the program performs the turning control method of the autonomous vehicle according to any one of the above when the program is run.

[0164] According to another aspect of the embodiments of the present application, there is also provided a computer program product, comprising computer instructions, which perform the turning control method of the autonomous vehicle according to any one of the above when the computer instructions are executed by a processor.

[0165] According to another aspect of the embodiments of the present application, there is also provided a computer readable storage medium, comprising a stored program, wherein the program performs the turning control method of the autonomous vehicle according to any one of the above.

[0166] Optionally, in the present embodiment, the computer readable storage medium can be located in any one of computer terminals in a computer terminal group in a computer network, or in any one of communication devices in a communication device group.

[0167] Optionally, in the embodiment, the computer readable storage medium is configured to store program codes for performing the following steps: obtaining a first predicted trajectory of a traffic participant within a predetermined range centered on the target autonomous vehicle, and obtaining a second predicted trajectory when the target autonomous vehicle performs a U-turn operation according to an original U-turn strategy; when it is determined that the first predicted trajectory and the second predicted trajectory have overlapping trajectories based on a comparison result of the first predicted trajectory and the second predicted trajectory, obtaining a first arrival time of the traffic participant at each predicted overlapping point on the overlapping trajectories; evaluating the first arrival time to obtain an influence degree of the traffic participant on the U-turn operation; adjusting the original U-turn strategy according to the influence degree to obtain a target U-turn strategy; and controlling the target autonomous vehicle to perform the U-turn operation according to the target U-turn strategy.

[0168] Optionally, in the embodiment, the computer readable storage medium is configured to store program codes for performing the following steps: collecting, by the target autonomous vehicle, first motion state information of the traffic participant within the predetermined range, and processing the motion state information by using a trajectory prediction model to obtain a first predicted trajectory, wherein the trajectory prediction model is a model trained by machine learning using a plurality of sets of training data; obtaining second motion state information of the target autonomous vehicle, and generating a second predicted trajectory according to the second motion state information and the original U-turn strategy.

[0169] Optionally, in the embodiment, the computer readable storage medium is configured to store program codes for performing the following steps: converting the first predicted trajectory and the second predicted trajectory into the same time-space coordinate system to obtain a third predicted trajectory and a fourth predicted trajectory; determining a shortest distance between each point on the third predicted trajectory and the fourth predicted trajectory to obtain a shortest distance set; and determining that the first predicted trajectory and the second predicted trajectory have overlapping trajectories when there is a trajectory point in the shortest distance set that is less than a preset safety threshold.

[0170] Optionally, in the embodiment, the computer readable storage medium is configured to store program codes for performing the following steps: when it is determined that the first predicted trajectory and the second predicted trajectory have overlapping trajectories based on a comparison result of the first predicted trajectory and the second predicted trajectory, obtaining a first arrival time of the traffic participant at each predicted overlapping point on the overlapping trajectories, and using graph theory connectivity detection to analyze reachability of a road network corresponding to a road where the target autonomous vehicle is located and a bottleneck area of the road network; determining a dynamic feature of a traffic flow and an obstacle point of the road corresponding to the road network according to the reachability and the bottleneck area; obtaining a traffic flow pattern of the traffic flow according to the dynamic feature and the obstacle point; generating an adaptive soft constraint according to the traffic flow pattern, and adjusting the original U-turn strategy according to the adaptive soft constraint and the influence degree to obtain a target U-turn strategy.

[0171] Optionally, in the embodiment, the computer readable storage medium is configured to store program code for performing the following steps: obtaining environmental change information of the target autonomous vehicle during driving of the target autonomous vehicle; updating the traffic flow pattern and the weight of the adaptive soft constraint according to the environmental change information, to obtain an updated traffic flow pattern and an updated adaptive soft constraint; and adjusting the original U-turn strategy according to the updated traffic flow pattern and the updated adaptive soft constraint, to obtain a target U-turn strategy.

[0172] Optionally, in the embodiment, the computer readable storage medium is configured to store program code for performing the following steps: obtaining a start time point at which the target autonomous vehicle starts to perform the U-turn operation, and comparing the start time point with the first arrival time to calculate a time difference between the start time point and the first arrival time; obtaining a type and a behavior of the traffic participant; and determining the influence degree according to the type, the behavior, and the time difference.

[0173] Optionally, in the embodiment, the computer readable storage medium is configured to store program code for performing the following steps: when the influence degree is greater than the influence degree threshold, postponing the U-turn operation, or adjusting the second predicted trajectory.

[0174] Optionally, in the embodiment, the computer readable storage medium is configured to store program code for performing the following steps: adopting a multi-level progressive degradation strategy to dynamically adjust a search constraint condition, to adjust the original U-turn strategy to obtain a target U-turn strategy, wherein the search constraint condition is used to constrain a U-turn path of the target autonomous vehicle, and the multi-level progressive degradation strategy is used to adjust a margin of the search constraint condition.

[0175] Optionally, in the embodiment, the computer readable storage medium is configured to store program code for performing the following steps: defining a multi-level progressive degradation strategy; and starting from a highest strategy of the multi-level progressive degradation strategy, using a search algorithm to determine a U-turn path from a current position to a destination until the U-turn path satisfies a safety constraint condition.

[0176] Optionally, in the embodiment, the computer readable storage medium is configured to store program code for performing the following steps: predicting a traffic flow of a predetermined time window in a future time period to obtain a traffic flow prediction result; dynamically adjusting a U-turn starting point of the target autonomous vehicle according to the traffic flow prediction result and the adaptive soft constraint, to obtain an adjusted U-turn starting point; and adjusting the original U-turn strategy according to the adjusted U-turn starting point, to obtain a target U-turn strategy.

[0177] Optionally, in the embodiment, the computer readable storage medium is configured to store program code for performing the following steps: in the process of controlling the target automatic driving vehicle to perform the turning operation according to the target turning strategy, evaluating a turning path corresponding to the target turning strategy to obtain a turning path evaluation result; when the turning path evaluation result indicates that the turning path is in a failure state, re-planning the turning path according to path planning information corresponding to the turning path and current traffic flow to obtain a new turning path; and controlling the target automatic driving vehicle to perform the turning operation according to the new target turning strategy.

[0178] The above-mentioned serial numbers of the embodiments of the application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0179] In the above-mentioned embodiments of the application, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0180] In several embodiments provided in the present application, it should be understood that the disclosed technical contents can be implemented by other manners. Among them, the above-mentioned device embodiments are only schematic, for example, the division of the units can be a logical function division, and in actual implementation, there can be another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or modules shown or discussed can be indirect coupling or communication connection through some interfaces, units or modules, which can be electrical or other forms.

[0181] Each technical feature of the above-mentioned embodiments can be combined arbitrarily, in order to make the description simple, not all possible combinations of each technical feature in the above-mentioned embodiments are described, however, as long as the combination of these technical features does not exist contradictory, it should be considered as the scope of the present application.

[0182] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed to multiple units. According to the actual needs, part or all of the units can be selected to achieve the purpose of the embodiment.

[0183] In addition, each functional unit in each embodiment of the application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be realized in the form of hardware or in the form of software functional unit.

[0184] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in other words, the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0185] The above is only the preferred embodiment of the present application, and it should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be considered as the protection scope of the present application.

Claims

1. A method for controlling the U-turn of an autonomous vehicle, characterized in that, The method comprises the following steps: acquiring a first predicted trajectory of a traffic participant within a predetermined range centered on a target autonomous vehicle, and simultaneously acquiring a second predicted trajectory of the target autonomous vehicle performing a U-turn operation according to an original U-turn strategy; when it is determined based on a comparison result of the first predicted trajectory and the second predicted trajectory that the first predicted trajectory and the second predicted trajectory have an overlapping trajectory, acquiring a first arrival time of the traffic participant reaching each predicted overlapping point on the overlapping trajectory; evaluating the first arrival time to obtain an influence degree of the traffic participant on the U-turn operation; adjusting the original U-turn strategy according to the influence degree to obtain a target U-turn strategy; controlling the target autonomous vehicle to perform the U-turn operation according to the target U-turn strategy.

2. The method of claim 1, wherein, The method comprises the following steps: collecting first motion state information of the traffic participant within the predetermined range by the target autonomous vehicle, and processing the motion state information by using a trajectory prediction model to obtain the first predicted trajectory, wherein the trajectory prediction model is a model trained by machine learning using multiple sets of training data; acquiring second motion state information of the target autonomous vehicle, and generating the second predicted trajectory according to the second motion state information and the original U-turn strategy.

3. The method of claim 1, wherein, The method comprises the following steps: converting the first predicted trajectory and the second predicted trajectory into the same time-space coordinate system to obtain a third predicted trajectory and a fourth predicted trajectory; determining the shortest distance between each point on the third predicted trajectory and the fourth predicted trajectory to obtain a shortest distance set; when it is determined that there is a trajectory point less than a preset safety threshold in the shortest distance set, it is determined that the first predicted trajectory and the second predicted trajectory have an overlapping trajectory.

4. The method of claim 1, wherein, The method comprises the following steps: when it is determined based on a comparison result of the first predicted trajectory and the second predicted trajectory that the first predicted trajectory and the second predicted trajectory have an overlapping trajectory, acquiring a first arrival time of the traffic participant reaching each predicted overlapping point on the overlapping trajectory, and simultaneously analyzing the accessibility of a road network corresponding to a road where the target autonomous vehicle is located and a bottleneck area of the road network by using graph theory connectivity detection; determining a dynamic feature and an obstacle point of a traffic flow corresponding to the road according to the accessibility and the bottleneck area; obtaining a traffic flow pattern of the traffic flow according to the dynamic feature and the obstacle point; generating an adaptive soft constraint according to the traffic flow pattern, and adjusting the original U-turn strategy according to the adaptive soft constraint and the influence degree to obtain the target U-turn strategy.

5. The method of claim 4, wherein, adjust the original U-turn strategy according to the influence degree, to obtain a target U-turn strategy, including: obtaining environment change information of the target autonomous vehicle during driving of the target autonomous vehicle; updating the traffic flow mode and the weight of the adaptive soft constraint according to the environment change information, to obtain an updated traffic flow mode and an updated adaptive soft constraint; adjusting the original U-turn strategy according to the updated traffic flow mode and the updated adaptive soft constraint, to obtain the target U-turn strategy.

6. The method of claim 1, wherein, evaluating the first arrival time to obtain the influence degree of the traffic participant on the U-turn operation, including: comparing the start time point at which the target autonomous vehicle starts to perform the U-turn operation with the first arrival time to calculate a time difference value between the start time point and the first arrival time; obtaining the type and behavior of the traffic participant; determining the influence degree according to the type, the behavior, and the time difference value.

7. The method of claim 1, wherein, adjusting the original U-turn strategy according to the influence degree, to obtain a target U-turn strategy, including: when it is determined that the influence degree is greater than an influence degree threshold, postponing the U-turn operation, or adjusting the second predicted trajectory.

8. The method of claim 1, wherein, adjusting the original U-turn strategy according to the influence degree, to obtain a target U-turn strategy, including: adopting a multi-level progressive degradation strategy to dynamically adjust a search constraint condition, to adjust the original U-turn strategy to obtain the target U-turn strategy, wherein the search constraint condition is used to constrain a U-turn path of the target autonomous vehicle, and the multi-level progressive degradation strategy is used to adjust a margin of the search constraint condition.

9. The method of claim 8, wherein, adopting a multi-level progressive degradation strategy to dynamically adjust a search constraint condition, to adjust the original U-turn strategy to obtain the target U-turn strategy, including: defining the multi-level progressive degradation strategy; starting from a highest strategy of the multi-level progressive degradation strategy to determine a U-turn path from a current position to a destination by using a search algorithm, until the U-turn path satisfies a safety constraint condition.

10. The method of claim 1, wherein, adjusting the original U-turn strategy according to the influence degree, to obtain a target U-turn strategy, including: predicting a traffic flow of a predetermined time window in a future time period, to obtain a traffic flow prediction result; dynamically adjusting a U-turn starting point of the target autonomous vehicle according to the traffic flow prediction result and an adaptive software constraint, to obtain an adjusted U-turn starting point; adjusting the original U-turn strategy according to the adjusted U-turn starting point, to obtain the target U-turn strategy.

11. The method of controlling a U-turn of an autonomous vehicle according to claim 1, wherein The U-turn control method further includes: evaluating a U-turn path corresponding to the target U-turn strategy to obtain a U-turn path evaluation result during control of the target autonomous vehicle to perform the U-turn operation according to the target U-turn strategy; when the U-turn path evaluation result indicates that the U-turn path is in a failure state, re-performing U-turn path planning according to path planning information corresponding to the U-turn path and a current traffic flow, to obtain a new U-turn path; The target autonomous vehicle is controlled to perform the turning operation according to the new target turning strategy.

12. A U-turn control device of an autonomous vehicle, characterized by comprising: The application comprises: a first acquisition unit configured to acquire a first predicted trajectory of a traffic participant within a predetermined range centered on a target autonomous vehicle, and to acquire a second predicted trajectory when the target autonomous vehicle performs a turning operation according to an original turning strategy; a second acquisition unit configured to, when it is determined based on a comparison result of the first predicted trajectory and the second predicted trajectory that the first predicted trajectory and the second predicted trajectory have a coincident trajectory, acquire a first arrival time of the traffic participant to each predicted coincident point on the coincident trajectory; an evaluation unit configured to evaluate the first arrival time to obtain an influence degree of the traffic participant on the turning operation; an adjustment unit configured to adjust the original turning strategy according to the influence degree to obtain a target turning strategy; a control unit configured to control the target autonomous vehicle to perform the turning operation according to the target turning strategy.

13. An autonomous vehicle, comprising: The application also provides an autonomous vehicle using the turning control method of the autonomous vehicle according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that, The application also provides a computer readable storage medium comprising a stored program, wherein the program performs the turning control method of the autonomous vehicle according to any one of claims 1 to 11.

15. A computer program product comprising computer instructions, characterized in that, The application also provides a computer instruction executed by a processor to perform the turning control method of the autonomous vehicle according to any one of claims 1 to 11.

Images