Obstacle avoidance planning method, device and equipment for low-speed unmanned vehicle and medium

Through the combination of the strategic obstacle avoidance planner and the tuning loss function, the problem of difficulty in obtaining human driving trajectory on low-speed unmanned vehicles is solved, and the obstacle avoidance planning model training under the human driving trajectory is realized, the application scope is expanded and the improvement is improved. Model capability.

CN118915740APending Publication Date: 2024-11-08JIUZHI (SUZHOU) INTELLIGENT TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410965831.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The data-based planning algorithms for existing autonomous vehicles are difficult to apply in low-speed autonomous vehicles, especially due to the inability to obtain human driving trajectories and simulate authenticity problems, which leads to the inability to effectively train and apply to the categories of vehicles that humans cannot drive.

Method used

The strategy obstacle avoidance planner obtains real road condition data, imitates training the basic obstacle avoidance planning model, uses the tuning loss function to tune, and obtains the target obstacle avoidance planning model, which is used to train low-speed unmanned vehicles without human driving trajectory. Obstacle avoidance planning model.

Benefits of technology

The obstacle avoidance planning model training is achieved without human driving trajectory on low-speed unmanned vehicles, which expands the scope of application and improves the capabilities of the model, and avoids dependence on human driving data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915740A_ABST
    Figure CN118915740A_ABST
Patent Text Reader

Abstract

The invention discloses an obstacle avoidance planning method and device for a low-speed unmanned vehicle, equipment and a medium. The method comprises the following steps: acquiring an initial obstacle avoidance trajectory of a low-speed unmanned vehicle for real road condition data with at least one obstacle through a strategy obstacle avoidance planner; performing simulation training on a basic obstacle avoidance planning model according to the initial obstacle avoidance trajectory and the real road condition data to obtain a middle obstacle avoidance planning model; performing tuning training on the intermediate obstacle avoidance planning model according to a tuning loss function and the real road condition data to obtain a target obstacle avoidance planning model; wherein the target obstacle avoidance planning model is used for determining the real-time obstacle avoidance trajectory of the low-speed unmanned vehicle according to the real-time road condition data of the low-speed unmanned vehicle. According to the embodiment of the invention, obstacle avoidance planning model training of the low-speed unmanned vehicle under the condition of no human driving track can be realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and in particular to an obstacle avoidance planning method, device, equipment and medium for a low-speed unmanned vehicle. Background Art

[0002] For autonomous vehicles, the planning algorithms used in the industry are divided into two categories: policy-based planning algorithms and data-based planning algorithms. The former uses strategies designed by engineers and mathematical optimization methods to obtain vehicle driving trajectories through calculations; the latter uses a large amount of data to train a large model, and generates trajectories through the trained large model. Policy-based planning algorithms have the advantages of good interpretability and high safety, but due to the high complexity of actual vehicle driving problems, which exceeds the limit of human exhaustion, and the high-dimensional non-convex optimization calculations are extremely time-consuming (autonomous driving problems are such problems), it is impossible to run in real time on real vehicles. Therefore, data-based planning algorithms that can learn from data are generally considered to have a higher upper limit and are a more promising direction for future development. There are two implementation schemes for data-based planning algorithms.

[0003] One is a solution based on reinforcement learning. However, reinforcement learning requires a perfect simulation of the actual environment, which can be done in chess. However, in autonomous driving, it is impossible to obtain a sufficiently realistic simulator with existing technology (because it is impossible to simulate the reactions of all other vehicles / pedestrians on the real road), so effective training cannot be carried out. The problem of simulation authenticity has not been effectively solved, resulting in the inability of such solutions to be applied on a large scale.

[0004] The second is the supervised learning scheme that uses human driving trajectories as the true value. This type of scheme has two shortcomings. First, the upper limit of the model will be affected by human driving and cannot truly surpass humans. In addition, it will be affected by the styles of different drivers, which may cause the style of the model output results to be inconsistent and radical. Secondly, there are many types of low-speed unmanned vehicles, such as low-speed logistics low-speed unmanned vehicles. Due to the lack of a cockpit, human driving trajectories cannot actually be obtained. In the case where human driving trajectories cannot be obtained, this type of scheme cannot be applied. Even if the supervised learning scheme using human driving trajectories cannot be applied to vehicle categories that humans cannot drive (logistics low-speed unmanned vehicles, low-speed sweepers, etc.), and the supervised learning scheme using human driving trajectories will ultimately be limited by human driving trajectories, and the ability cannot truly reach a level beyond humans. In addition, the differences in styles of different drivers also have an adverse effect on model training. Summary of the invention

[0005] The present invention provides a method, device, equipment and medium for obstacle avoidance planning of a low-speed unmanned vehicle, so as to realize the obstacle avoidance planning model training of the low-speed unmanned vehicle without a human driving trajectory.

[0006] According to one aspect of the present invention, a method for obstacle avoidance planning of a low-speed unmanned vehicle is provided, comprising:

[0007] Obtaining the initial obstacle avoidance trajectory of the low-speed unmanned vehicle on real road conditions with at least one obstacle through a strategic obstacle avoidance planner;

[0008] Performing simulation training on a basic obstacle avoidance planning model according to the initial obstacle avoidance trajectory and the real road condition data to obtain an intermediate obstacle avoidance planning model;

[0009] The intermediate obstacle avoidance planning model is tuned and trained according to the tuning loss function and the real road condition data to obtain a target obstacle avoidance planning model; wherein the target obstacle avoidance planning model is used to determine the real-time obstacle avoidance trajectory of the low-speed unmanned vehicle according to the real-time road condition data of the low-speed unmanned vehicle.

[0010] According to another aspect of the present invention, there is provided an obstacle avoidance planning device for a low-speed unmanned vehicle, comprising:

[0011] A trajectory generation module, used to obtain an initial obstacle avoidance trajectory of a low-speed unmanned vehicle for real road condition data with at least one obstacle through a strategic obstacle avoidance planner;

[0012] A simulation training module, used for imitating and training the basic obstacle avoidance planning model according to the initial obstacle avoidance trajectory and the real road condition data to obtain an intermediate obstacle avoidance planning model;

[0013] A tuning training module is used to tune and train the intermediate obstacle avoidance planning model according to the tuning loss function and the real road condition data to obtain a target obstacle avoidance planning model; wherein the target obstacle avoidance planning model is used to determine the real-time obstacle avoidance trajectory of the low-speed unmanned vehicle according to the real-time road condition data of the low-speed unmanned vehicle.

[0014] According to another aspect of the present invention, a computer program product is provided, including a computer program, which, when executed by a processor, implements the obstacle avoidance planning method for a low-speed unmanned vehicle according to any embodiment of the present invention.

[0015] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0016] at least one processor; and

[0017] a memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the obstacle avoidance planning method for a low-speed unmanned vehicle described in any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the obstacle avoidance planning method for a low-speed unmanned vehicle described in any embodiment of the present invention when executed.

[0020] The training model of the embodiment of the present invention first imitates the results of the traditional planner, and then is tuned through an artificially set tuning loss function, thereby avoiding the use of human driving data in the training process. This makes the process in this solution applicable to low-speed unmanned vehicles such as low-speed unmanned delivery vehicles that cannot be driven manually, and has a theoretically wider usable range and a higher upper limit.

[0021] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 is a flow chart of an obstacle avoidance planning method for a low-speed unmanned vehicle provided according to an embodiment of the present invention;

[0024] Figure 2A is a flow chart of an obstacle avoidance planning method for a low-speed unmanned vehicle provided according to another embodiment of the present invention;

[0025] Figure 2B is a flowchart of a model training provided according to another embodiment of the present invention;

[0026] Figure 3 is a structural schematic diagram of an obstacle avoidance planning method for a low-speed unmanned vehicle provided according to another embodiment of the present invention;

[0027] Figure 4 It is a schematic diagram of the structure of an electronic device implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] Figure 1 This is a flowchart of a method for obstacle avoidance planning of a low-speed unmanned vehicle provided by an embodiment of the present invention. This embodiment can be applied to the case where an obstacle avoidance planning model is pre-trained and the trained obstacle avoidance planning is deployed on a low-speed unmanned vehicle. The method can be executed by an obstacle avoidance planning device of a low-speed unmanned vehicle. The device can be implemented in the form of hardware and / or software. The device can be configured in an electronic device with corresponding data processing capabilities, such as a model training server. Figure 1 As shown, the method includes:

[0031] S110, obtaining an initial obstacle avoidance trajectory of the low-speed unmanned vehicle for real road condition data with at least one obstacle through a strategic obstacle avoidance planner.

[0032] S120, performing imitation training on a basic obstacle avoidance planning model according to the initial obstacle avoidance trajectory and the real road condition data to obtain an intermediate obstacle avoidance planning model.

[0033] S130, performing tuning training on the intermediate obstacle avoidance planning model according to the tuning loss function and the real road condition data to obtain a target obstacle avoidance planning model.

[0034] Among them, the target obstacle avoidance planning model is used to determine the real-time obstacle avoidance trajectory of the low-speed unmanned vehicle according to the real-time road condition data of the low-speed unmanned vehicle. The low-speed unmanned vehicle includes at least one of the following: logistics low-speed unmanned vehicle, low-speed sweeper and take-out low-speed unmanned vehicle; the obstacle avoidance measures corresponding to the obstacle avoidance trajectory include at least one of the following: following the vehicle, detouring on the left, detouring on the right, overtaking on the left, overtaking on the right, passing the vehicle and giving way. The real road condition data is the real collected road condition data, and there are one or more obstacles in the corresponding road condition. In urban road scenes with many obstacles and complex road conditions, low-speed unmanned vehicles need to use obstacle avoidance trajectories to clarify the obstacle avoidance measures that can be taken for obstacles on the road. Obstacle avoidance measures include but are not limited to: following the vehicle, detouring on the left / right, overtaking on the left / right, passing the vehicle, giving way, etc.

[0035] Specifically, for real road condition data, an existing strategy planner is first used to find a feasible solution as the initial obstacle avoidance trajectory. For example, a traditional planning method can be used to perform horizontal and vertical layered mathematical modeling on the real road condition data, and the established mathematical model is solved to obtain the feasible solution.

[0036] The initial obstacle avoidance trajectory can be regarded as the human driving trajectory of a low-speed unmanned vehicle to a certain extent. This result is used as the supervision value to perform the first step of simulation training on the basic obstacle avoidance planning model, guiding the basic obstacle avoidance planning model to imitate the output results of the strategy planner under the same real road conditions data. After the training is completed, the intermediate obstacle avoidance planning model is obtained.

[0037] The intermediate obstacle avoidance planning model imitates the result of the strategy planner. It already has certain planning capabilities, but there is still a certain gap with what is desired, because it is meaningless in actual use if it only imitates the traditional method but cannot surpass it. Therefore, the second step of tuning training will be carried out later. During the tuning training, a tuning loss function based on a more sophisticated strategy will be used. Under the same real road conditions as before, it will be tuned according to the output results of the intermediate obstacle avoidance planning model, so as to obtain the final ideal target obstacle avoidance planning model.

[0038] After training, the target obstacle avoidance planning model will be deployed on the low-speed unmanned vehicle. During actual driving, the low-speed unmanned vehicle will input the real-time road condition data into the target obstacle avoidance planning model. If there is an obstacle in the real-time road condition data, the target obstacle avoidance planning model will output the real-time obstacle avoidance trajectory for the obstacle. The low-speed unmanned vehicle will determine whether it is necessary to avoid the obstacle based on the real-time obstacle avoidance trajectory based on multiple factors.

[0039] The training model of the embodiment of the present invention first imitates the results of the traditional planner, and then is tuned through an artificially set tuning loss function, thereby avoiding the use of human driving data in the training process. This makes the process in this solution applicable to low-speed unmanned vehicles such as low-speed unmanned delivery vehicles that cannot be driven manually, and has a theoretically wider usable range and a higher upper limit.

[0040] Figure 2A This is a flowchart of a low-speed unmanned vehicle obstacle avoidance planning method provided by another embodiment of the present invention. This embodiment is optimized and improved on the basis of the above embodiment. Figure 2A As shown, the method includes:

[0041] S210, obtaining an initial obstacle avoidance trajectory of the low-speed unmanned vehicle for real road condition data with at least one obstacle through a strategic obstacle avoidance planner.

[0042] S220, inputting the real road condition data into a basic obstacle avoidance planning model to obtain a first obstacle avoidance trajectory; and determining a deviation between the first obstacle avoidance trajectory and the initial obstacle avoidance trajectory as a simulation loss function.

[0043] S230. Performing feedback training on the basic obstacle avoidance planning model according to the simulation loss function to obtain an intermediate obstacle avoidance planning model.

[0044] When the industry uses supervised training to train models, it generally looks for a sufficiently ideal "supervision value" as the training target, and uses the deviation of the model output to the above "supervision value" as the criterion for judging the quality of the model. In simulation training, the initial obstacle avoidance trajectory is the supervision value, and the simulation loss function is the deviation between the first obstacle avoidance trajectory (i.e., the model output result) and the initial obstacle avoidance trajectory (e.g., the square of the difference between the two, i.e., L2 diff). The size of the deviation is the specific loss function value.

[0045] Specifically, Figure 2B As shown, the real road condition data is input into the basic obstacle avoidance planning model to obtain the first obstacle avoidance trajectory output by the model. The deviation between the first obstacle avoidance trajectory and the initial obstacle avoidance trajectory is determined as a simulation loss function, and the loss function value of this model training is calculated accordingly. If the loss function value does not meet the training stop condition, the basic obstacle avoidance planning model is retrained based on the loss function value. After retraining, the real road condition data is input into the basic obstacle avoidance planning model again, and the above steps are repeated. Until the loss function value of a certain model training meets the training stop condition, the basic obstacle avoidance planning model at this time is determined as the intermediate obstacle avoidance planning model.

[0046] S240: Input the real road condition data into the intermediate obstacle avoidance planning model to obtain a second obstacle avoidance trajectory.

[0047] S250, based on the smoothness loss function, the safety loss function and the efficiency loss function, feedback training is performed on the intermediate obstacle avoidance planning model according to the second obstacle avoidance trajectory to obtain a target obstacle avoidance planning model.

[0048] Specifically, in the tuning training, the present invention uses a more complex tuning loss function than the simulation loss function, and uses these tuning loss functions to perform more refined tuning training on the intermediate obstacle avoidance planning model. The tuning loss function includes three categories of loss functions: smoothness loss function, safety loss function and efficiency loss function.

[0049] Compared with simulation training, the initial obstacle avoidance trajectory is no longer needed as a supervision value during tuning training. The real road condition data is input into the intermediate obstacle avoidance planning model to obtain the second obstacle avoidance trajectory output by it. The second obstacle avoidance trajectory is evaluated to determine the function values ​​of the smoothness loss function, the safety loss function, and the efficiency loss function. Then, the intermediate obstacle avoidance planning model is retrained using these three function values. This process is repeated until the second obstacle avoidance trajectory output by the intermediate obstacle avoidance planning model meets the design requirements in terms of smoothness, safety, and efficiency. That is, the three function values ​​at this time all meet the training stop conditions, and the intermediate obstacle avoidance planning model at this time is determined as the target obstacle avoidance planning model.

[0050] Optionally, the security loss function is a piecewise function, expressed as follows:

[0051]

[0052] Among them, L(z) is the function value of the safety loss function, z is the distance between the low-speed unmanned vehicle and the obstacle polygon, T is the judgment threshold of the distance between the low-speed unmanned vehicle and the obstacle, and C is the penalty value in the event of collision.

[0053] Specifically, the safety loss function is difficult to express accurately in traditional mathematical optimization methods due to its complex calculation and strong nonlinearity. However, in machine learning, since it is no longer restricted by the mathematical models and mathematical optimization methods used, very precise and complex loss functions can be used to obtain a higher capability ceiling.

[0054] In addition, in the safety loss function, since a large number of geometric operations are involved to determine the safety distance, collision, etc., the present invention performs a large number of matrix processing on the geometric operations to correspond to the characteristics of GPU training, which greatly reduces the calculation time from the mathematical principle.

[0055] The smoothness loss function is quite similar to the smoothness function in traditional mathematical optimization methods. Generally, the sum of the squares of the corresponding items is used (the squares of the acceleration / jerk / angular velocity / angular acceleration of each trajectory point are summed up). However, there is an advantage in that the loss function used for model training here only needs to ensure the existence of the gradient, and does not need to meet the order requirements required by the mathematical optimization method, as well as the requirements of constrained linearization, so a slightly more precise mathematical expression method can be used.

[0056] Optionally, the efficiency loss function is a piecewise discontinuous and non-monotonic reward function, expressed as follows:

[0057]

[0058] Among them, G(x) is the function value of the efficiency loss function, g1 to gn are preset function values, x is the current obstacle avoidance situation, and x1 to xn are preset different obstacle avoidance situations.

[0059] Specifically, efficiency-related reward functions often include conditional judgments. For example, for a slow car ahead, there may be multiple obstacle avoidance situations, such as successful overtaking, failed overtaking, and continued waiting. Each obstacle avoidance situation corresponds to a specific function value. Such a loss function is piecewise discontinuous and not monotonic, so it is very difficult to use in mathematical optimization methods, but it can be used in the machine learning framework of this solution.

[0060] The embodiment of the present invention optimizes the model from three aspects of security, smoothness and efficiency, thereby improving the optimization effect of the model.

[0061] Figure 3 A schematic diagram of the structure of an obstacle avoidance planning device for a low-speed unmanned vehicle provided by another embodiment of the present invention. Figure 3 As shown, the device comprises:

[0062] The trajectory generation module 310 is used to obtain an initial obstacle avoidance trajectory of the low-speed unmanned vehicle for real road condition data with at least one obstacle through a strategic obstacle avoidance planner;

[0063] A simulation training module 320 is used to simulate and train the basic obstacle avoidance planning model according to the initial obstacle avoidance trajectory and the real road condition data to obtain an intermediate obstacle avoidance planning model;

[0064] The tuning training module 330 is used to tune and train the intermediate obstacle avoidance planning model according to the tuning loss function and the actual road condition data to obtain a target obstacle avoidance planning model; wherein the target obstacle avoidance planning model is used to determine the real-time obstacle avoidance trajectory of the low-speed unmanned vehicle according to the real-time road condition data of the low-speed unmanned vehicle.

[0065] The obstacle avoidance planning device for a low-speed unmanned vehicle provided in an embodiment of the present invention can execute the obstacle avoidance planning method for a low-speed unmanned vehicle provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0066] Optionally, the simulation training module 320 includes:

[0067] A trajectory acquisition unit, used for inputting the real road condition data into a basic obstacle avoidance planning model to obtain a first obstacle avoidance trajectory;

[0068] a function determination unit, configured to determine a deviation between the first obstacle avoidance trajectory and the initial obstacle avoidance trajectory as a simulation loss function;

[0069] A simulation training unit is used to perform feedback training on the basic obstacle avoidance planning model according to the simulation loss function to obtain an intermediate obstacle avoidance planning model.

[0070] Optionally, the tuning loss function includes a preset smoothness loss function, a safety loss function and an efficiency loss function, and the tuning training module 330 is specifically used to: input the real road condition data into the intermediate obstacle avoidance planning model to obtain a second obstacle avoidance trajectory; based on the smoothness loss function, the safety loss function and the efficiency loss function, feedback training the intermediate obstacle avoidance planning model according to the second obstacle avoidance trajectory to obtain a target obstacle avoidance planning model.

[0071] Optionally, the security loss function is a piecewise function, expressed as follows:

[0072]

[0073] Among them, L(z) is the function value of the safety loss function, z is the distance between the low-speed unmanned vehicle and the obstacle polygon, T is the judgment threshold of the distance between the low-speed unmanned vehicle and the obstacle, and C is the penalty value in the event of collision.

[0074] Optionally, the efficiency loss function is a piecewise discontinuous and non-monotonic reward function, expressed as follows:

[0075]

[0076] Among them, G(x) is the function value of the efficiency loss function, g1 to gn are preset function values, x is the current obstacle avoidance situation, and x1 to xn are preset different obstacle avoidance situations.

[0077] Optionally, the low-speed unmanned vehicle includes at least one of the following: a low-speed unmanned vehicle for logistics, a low-speed sweeping vehicle and a low-speed unmanned vehicle for food delivery; the obstacle avoidance measures corresponding to the obstacle avoidance trajectory include at least one of the following: following the vehicle, detouring on the left, detouring on the right, overtaking on the left, overtaking on the right, passing the vehicle and giving way.

[0078] The obstacle avoidance planning device for a low-speed unmanned vehicle further described can also execute the obstacle avoidance planning method for a low-speed unmanned vehicle provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0079] Figure 4 A schematic diagram of the structure of an electronic device 40 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0080] like Figure 4 As shown, the electronic device 40 includes at least one processor 41, and a memory connected to the at least one processor 41, such as a read-only memory (ROM) 42, a random access memory (RAM) 43, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 41 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 42 or the computer program loaded from the storage unit 48 to the random access memory (RAM) 43. In the RAM 43, various programs and data required for the operation of the electronic device 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other through a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0081] A number of components in the electronic device 40 are connected to the I / O interface 45, including: an input unit 46, such as a keyboard, a mouse, etc.; an output unit 47, such as various types of displays, speakers, etc.; a storage unit 48, such as a disk, an optical disk, etc.; and a communication unit 49, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0082] The processor 41 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 41 executes the various methods and processes described above, such as the obstacle avoidance planning method for a low-speed unmanned vehicle.

[0083] In some embodiments, the obstacle avoidance planning method for a low-speed unmanned vehicle may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 48. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 40 via the ROM 42 and / or the communication unit 49. When the computer program is loaded into the RAM 43 and executed by the processor 41, one or more steps of the obstacle avoidance planning method for the low-speed unmanned vehicle described above may be performed. Alternatively, in other embodiments, the processor 41 may be configured to execute the obstacle avoidance planning method for the low-speed unmanned vehicle in any other appropriate manner (e.g., by means of firmware).

[0084] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0085] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0086] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0087] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0088] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0089] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0090] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0091] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A low-speed unmanned vehicle obstacle avoidance planning method, characterized in that: The method comprises: Obtaining the initial obstacle avoidance trajectory of the low-speed unmanned vehicle on real road conditions with at least one obstacle through a strategic obstacle avoidance planner; Performing simulation training on a basic obstacle avoidance planning model according to the initial obstacle avoidance trajectory and the real road condition data to obtain an intermediate obstacle avoidance planning model; The intermediate obstacle avoidance planning model is tuned and trained according to the tuning loss function and the real road condition data to obtain a target obstacle avoidance planning model; wherein the target obstacle avoidance planning model is used to determine the real-time obstacle avoidance trajectory of the low-speed unmanned vehicle according to the real-time road condition data of the low-speed unmanned vehicle.

2. The method according to claim 1, characterized in that The imitation training of the basic obstacle avoidance planning model according to the initial obstacle avoidance trajectory and the real road condition data to obtain the intermediate obstacle avoidance planning model includes: Inputting the real road condition data into a basic obstacle avoidance planning model to obtain a first obstacle avoidance trajectory; Determine a deviation between the first obstacle avoidance trajectory and the initial obstacle avoidance trajectory as a simulation loss function; The basic obstacle avoidance planning model is feedback trained according to the simulation loss function to obtain an intermediate obstacle avoidance planning model.

3. The method according to claim 2, characterized in that The tuning loss function includes a preset smoothness loss function, a safety loss function and an efficiency loss function. The tuning training of the intermediate obstacle avoidance planning model according to the tuning loss function and the real road condition data to obtain the target obstacle avoidance planning model includes: Inputting the real road condition data into the intermediate obstacle avoidance planning model to obtain a second obstacle avoidance trajectory; Based on the smoothness loss function, the safety loss function and the efficiency loss function, feedback training is performed on the intermediate obstacle avoidance planning model according to the second obstacle avoidance trajectory to obtain a target obstacle avoidance planning model.

4. The method according to claim 3, characterized in that The security loss function is a piecewise function, which is expressed as follows: Among them, L(z) is the function value of the safety loss function, z is the distance between the low-speed unmanned vehicle and the obstacle polygon, T is the judgment threshold of the distance between the low-speed unmanned vehicle and the obstacle, and C is the penalty value in the event of collision.

5. The method according to claim 3, characterized in that: The efficiency loss function is a piecewise discontinuous and non-monotonic reward function, expressed as follows: Among them, G(x) is the function value of the efficiency loss function, g1 to gn are preset function values, x is the current obstacle avoidance situation, and x1 to xn are preset different obstacle avoidance situations.

6. The method according to any one of claims 1 to 5, characterized in that: The low-speed unmanned vehicle includes at least one of the following: a low-speed unmanned logistics vehicle, a low-speed sweeping vehicle and a low-speed unmanned delivery vehicle; the obstacle avoidance measures corresponding to the obstacle avoidance trajectory include at least one of the following: following the vehicle, detouring on the left, detouring on the right, overtaking on the left, overtaking on the right, passing the vehicle and giving way.

7. An obstacle avoidance planning device for a low-speed unmanned vehicle, characterized in that: The device comprises: A trajectory generation module, used to obtain an initial obstacle avoidance trajectory of a low-speed unmanned vehicle for real road condition data with at least one obstacle through a strategic obstacle avoidance planner; A simulation training module, used for imitating and training the basic obstacle avoidance planning model according to the initial obstacle avoidance trajectory and the real road condition data to obtain an intermediate obstacle avoidance planning model; A tuning training module is used to tune and train the intermediate obstacle avoidance planning model according to the tuning loss function and the real road condition data to obtain a target obstacle avoidance planning model; wherein the target obstacle avoidance planning model is used to determine the real-time obstacle avoidance trajectory of the low-speed unmanned vehicle according to the real-time road condition data of the low-speed unmanned vehicle.

8. The device according to claim 7, characterized in that The simulation training module includes: A trajectory acquisition unit, used for inputting the real road condition data into a basic obstacle avoidance planning model to obtain a first obstacle avoidance trajectory; a function determination unit, configured to determine a deviation between the first obstacle avoidance trajectory and the initial obstacle avoidance trajectory as a simulation loss function; A simulation training unit is used to perform feedback training on the basic obstacle avoidance planning model according to the simulation loss function to obtain an intermediate obstacle avoidance planning model.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the obstacle avoidance planning method for a low-speed unmanned vehicle described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the obstacle avoidance planning method for a low-speed unmanned vehicle according to any one of claims 1 to 6 when executed.

Citation Information

Cited By

  • Hospital multi-scene transportation robot hybrid scheduling method based on dynamic priority

    CN120895200A