WO2019047673A1 - 一种端到端自动驾驶系统的纵向控制模型的评估方法及装置 - Google Patents
一种端到端自动驾驶系统的纵向控制模型的评估方法及装置 Download PDFInfo
- Publication number
- WO2019047673A1 WO2019047673A1 PCT/CN2018/100246 CN2018100246W WO2019047673A1 WO 2019047673 A1 WO2019047673 A1 WO 2019047673A1 CN 2018100246 W CN2018100246 W CN 2018100246W WO 2019047673 A1 WO2019047673 A1 WO 2019047673A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- sequence
- control model
- reference standard
- longitudinal control
- mean square
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Definitions
- the invention relates to the field of computers, and in particular to a method and a device for evaluating a longitudinal control model of an end-to-end automatic driving system.
- One of the technical problems solved by the present invention is the lack of an effective method for evaluating the longitudinal control model of the automatic driving system.
- a method for evaluating a longitudinal control model of an end-to-end automatic driving system including:
- a longitudinal control model of the end-to-end automatic driving system is evaluated based on the reference standard speed sequence and the mean square error of the predicted speed sequence.
- an apparatus for evaluating a longitudinal control model of an end-to-end automatic driving system including:
- the present embodiment determines the mean square error according to the reference standard speed sequence calculated in the predetermined navigation system and the estimated speed sequence read in the longitudinal control model, and estimates the longitudinal control model according to the mean square error, thereby being objective and true The evaluation of the effect of the longitudinal control model, thereby improving the learning efficiency of deep learning in the field of automatic driving.
- FIG. 1 is a flow chart showing a method of evaluating a longitudinal control model of an end-to-end automatic driving system in accordance with an embodiment of the present invention.
- FIG. 2 is a flow chart showing a method of evaluating a longitudinal control model of an end-to-end automatic driving system according to Embodiment 1 of the present invention.
- FIG. 3 is a flow chart showing a method of evaluating a longitudinal control model of an end-to-end automatic driving system according to Embodiment 2 of the present invention.
- FIG. 4 is a block diagram showing an evaluation apparatus of a longitudinal control model of an end-to-end automatic driving system in accordance with an embodiment of the present invention.
- Fig. 5 is a block diagram showing an evaluation apparatus of a longitudinal control model of the end-to-end automatic driving system proposed in the third embodiment of the present invention.
- Fig. 6 is a block diagram showing an evaluation apparatus of a longitudinal control model of the end-to-end automatic driving system proposed in Embodiment 4 of the present invention.
- Computer device also referred to as “computer” in the context, is meant an intelligent electronic device that can perform predetermined processing, such as numerical calculations and/or logical calculations, by running a predetermined program or instruction, which can include a processor and The memory is executed by the processor to execute a predetermined process pre-stored in the memory to execute a predetermined process, or is executed by hardware such as an ASIC, an FPGA, a DSP, or the like, or a combination of the two.
- Computer devices include, but are not limited to, servers, personal computers, notebook computers, tablets, smart phones, and the like.
- the computer device includes a user device and a network device.
- the user equipment includes, but is not limited to, a computer, a smart phone, a PDA, etc.
- the network device includes but is not limited to a single network server, a server group composed of multiple network servers, or a cloud computing based computer Or a cloud composed of a network server, wherein cloud computing is a type of distributed computing, a super virtual computer composed of a group of loosely coupled computers.
- the computer device can be operated separately to implement the present invention, and can also access the network and implement the present invention by interacting with other computer devices in the network.
- the network in which the computer device is located includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a VPN network, and the like.
- the user equipment, the network equipment, the network, and the like are merely examples, and other existing or future possible computer equipment or networks, such as those applicable to the present invention, are also included in the scope of the present invention. It is included here by reference.
- FIG. 1 is a flow chart of a method of evaluating a longitudinal control model of an end-to-end automatic driving system in accordance with one embodiment of the present invention.
- the evaluation method of the longitudinal control model of the end-to-end automatic driving system described in this embodiment includes the following steps:
- step S110 the speed sequence in the predetermined navigation system is first extracted.
- the sequence of speeds includes, but is not limited to, a sequence of velocity components in the east-east direction of the current time and velocity components in the north-north direction with the coordinated world time as a reference.
- the reference standard velocity value can also be calculated by the following calculation formula, thereby storing the calculated plurality of reference velocity standard values as the reference standard velocity sequence:
- v true represents the reference standard velocity value
- v east represents the velocity component in the east direction
- v north represents the velocity component in the north direction.
- step S120 after calculating the reference standard speed sequence, the present embodiment can also read the estimated speed sequence from the pre-evaluated longitudinal control model.
- the step of reading the predicted speed sequence comprises reading a plurality of estimated speed values in chronological order from the longitudinal control model, and storing the plurality of estimated speed values as a set as an estimate Speed sequence.
- step S130 the mean square error of the reference standard speed sequence and the estimated speed sequence can be determined by the following calculation formula:
- MSE represents the mean square error
- n represents the number of evaluable data
- v true represents the reference standard speed value
- v predict represents the estimated speed value
- the magnitude relationship between the mean square error and the evaluation threshold may be determined by the value of the mean square error, and if the mean square error is smaller than the evaluation threshold, the vertical direction corresponding to the mean square error The evaluation results of the control model are better.
- the mean square error is determined according to the reference standard speed sequence calculated in the predetermined navigation system and the predicted speed sequence read in the longitudinal control model, and the longitudinal control model is evaluated according to the mean square error. Therefore, the effect of the longitudinal control model can be objectively and realistically evaluated, thereby improving the learning efficiency of deep learning in the field of automatic driving.
- the longitudinal control of a car mainly refers to speed control.
- the traditional evaluation method can only evaluate the self-driving system by calculating the time ratio of the self-driving vehicle by simulating the driver's intervention.
- the longitudinal control model of the automatic driving system is much more complicated, so the traditional method can not meet the requirements for evaluating the longitudinal control model of the automatic driving system.
- the present embodiment proposes another evaluation method of the longitudinal control model of the end-to-end automatic driving system, as shown in FIG. 2, comprising the following steps:
- the positive east direction velocity component v east and the northeast direction velocity component v north are read at the current time, and v east and v north are synthesized into a standard reference velocity v true by a calculation formula.
- a number of estimated speed values are read from the longitudinal control model to form a predicted speed sequence [v predict1 , v predict2 , v predict3 ...].
- the mean square error of the reference standard speed sequence and the estimated speed sequence can be determined by the following calculation formula:
- MSE represents the mean square error and n represents the number of evaluable data, ie the length of the reference standard sequence.
- An evaluation threshold can be set first and the calculated mean squared error is compared to the evaluation threshold. If the mean square error is less than the evaluation threshold, the evaluation result of the vertical control model corresponding to the mean square error is better, and can be used as a learning object of deep learning; if the mean square error is greater than the evaluation threshold, Then, the evaluation result of the longitudinal control model corresponding to the mean square error is poor, and is not recommended as a learning object for deep learning.
- the two velocity components in the longitudinal control model of the automatic driving system are taken as outputs, thereby interpreting the longitudinal control model as a regression problem, and thus the mean square error can be used as an index for evaluation.
- the evaluation of the longitudinal control model of the end-to-end automatic driving system can be realized.
- the conventional evaluation method can only evaluate some single parameters, for example, by calculating the number of simulated human interventions when the simulated vehicle deviates from the center line by more than one meter, that is, the time of the automatic driving system.
- the ratio is used to estimate the steering angle. Since the longitudinal control of the vehicle mainly refers to the speed control, these methods cannot evaluate the longitudinal control model of the automatic driving system.
- the present embodiment proposes an evaluation method of the longitudinal control model of the end-to-end automatic driving system, as shown in FIG. 3, comprising the following steps:
- the time in the predetermined navigation system is usually GPS week.
- the GPS week can be converted into UTC time, and then the velocity component in the east direction and the velocity component in the north direction are extracted separately.
- the following table shows a set of velocity components in the east direction and velocity components in the north direction:
- Speed component 1 (m/s) Speed component 2 (m/s) Speed component 3 (m/s) Speed component 4 (m/s)
- the velocity components shown in the above table can be synthesized into the reference standard velocity values shown in the following table by formula calculation.
- Sequence speed 1 (m/s) Sequence speed 2 (m/s) Sequence speed 3 (m/s) Sequence speed 4 (m/s) Reference standard speed sequence 11.569 11.556 11.551 11.538
- the mean square error can be calculated by the following formula. When four sets of data are included in the speed sequence, the formula is:
- the longitudinal control model of the automatic driving system can be evaluated by obtaining the mean square error by this calculation.
- the two velocity components in the longitudinal control model of the automatic driving system are taken as outputs, thereby interpreting the longitudinal control model as a regression problem, and thus the mean square error can be used as an index for evaluation.
- FIG. 4 is a block diagram of an evaluation apparatus for a longitudinal control model of an end-to-end automatic driving system in accordance with one embodiment of the present invention.
- the evaluation device for the longitudinal control model of the end-to-end automatic driving system (hereinafter referred to as “evaluation device") according to the present embodiment includes the following devices:
- Reference standard extraction device Means for determining a reference standard speed sequence based on a sequence of speeds extracted from a predetermined navigation system (hereinafter referred to as "reference standard extraction device”) 410;
- Estimated speed reading device Means for reading the estimated speed sequence of the longitudinal control model (hereinafter referred to as "estimated speed reading device") 420;
- Means for evaluating a longitudinal control model of the end-to-end automatic driving system (hereinafter referred to as "evaluation device") 430 according to the reference standard speed sequence and the mean square error of the predicted speed sequence.
- the speed sequence in the predetermined navigation system is first extracted by reference to the standard extraction device 410.
- the speed sequence includes, but is not limited to, a sequence of velocity components in the east-east direction of the current time and velocity components in the north-south direction extracted by the reference standard extraction device 410 with the coordinated world time as a reference.
- the reference standard velocity value may also be obtained by the reference standard extraction device 410 by the following calculation formula, thereby storing the calculated plurality of reference velocity standard values as the reference standard velocity sequence:
- v true represents the reference standard velocity value
- v east represents the velocity component in the east direction
- v north represents the velocity component in the north direction.
- the present embodiment can also read the predicted speed sequence from the pre-evaluated longitudinal control model by the predicted speed reading device 420.
- the estimated speed reading device 420 includes means for reading a plurality of estimated speed values chronologically from the longitudinal control model, and storing the plurality of estimated speed values in a set form.
- a device for estimating the speed sequence is provided.
- the mean square error of the reference standard speed sequence and the estimated speed sequence can be determined by the evaluation device 430 by the following calculation formula:
- MSE represents the mean square error
- n represents the number of evaluable data
- v true represents the reference standard speed value
- v predict represents the estimated speed value
- the magnitude difference between the mean square error and the evaluation threshold may be determined by the evaluation device 430 by the value of the mean square error, and if the mean square error is smaller than the evaluation threshold, the mean square error The evaluation results of the corresponding longitudinal control model are better.
- the mean square error is determined according to the reference standard speed sequence calculated in the predetermined navigation system and the predicted speed sequence read in the longitudinal control model, and the longitudinal control model is evaluated according to the mean square error. Therefore, the effect of the longitudinal control model can be objectively and realistically evaluated, thereby improving the learning efficiency of deep learning in the field of automatic driving.
- the longitudinal control of a car mainly refers to speed control.
- the traditional evaluation method can only evaluate the self-driving system by calculating the time ratio of the self-driving vehicle by simulating the driver's intervention.
- the longitudinal control model of the automatic driving system is much more complicated, so the traditional method can not meet the requirements for evaluating the longitudinal control model of the automatic driving system.
- the present embodiment proposes another evaluation device for the longitudinal control model of the end-to-end automatic driving system, as shown in FIG. 5, including the following devices:
- Means for calculating a reference standard speed sequence (hereinafter referred to as "first computing device") 510;
- Means for reading the estimated speed sequence of the longitudinal control model (hereinafter referred to as "first reading device") 520;
- error calculation means Means for calculating a mean square error of the reference standard speed sequence and the estimated speed sequence (hereinafter referred to as "error calculation means") 530;
- Means for evaluating the longitudinal control model (hereinafter referred to as "first evaluation device") 540.
- the first east direction directional velocity component v east and the north directional velocity component v north are read by the first computing device 510 on the basis of the coordinated world time, and v east and v north are synthesized into a standard reference velocity v by a calculation formula. True .
- the first reading device 520 reads a number of estimated speed values from the longitudinal control model to form a predicted speed sequence [v predict1 , v predict2 , v predict3 ...].
- the mean square error of the reference standard speed sequence and the estimated speed sequence can be determined by the error calculation means 530 by the following calculation formula:
- MSE represents the mean square error and n represents the number of evaluable data, ie the length of the reference standard sequence.
- an evaluation threshold can be set first and compared by the first evaluation device 510 to the calculated mean square error. If the mean square error is less than the evaluation threshold, the evaluation result of the vertical control model corresponding to the mean square error is better, and can be used as a learning object of deep learning; if the mean square error is greater than the evaluation threshold, Then, the evaluation result of the longitudinal control model corresponding to the mean square error is poor, and is not recommended as a learning object for deep learning.
- the two velocity components in the longitudinal control model of the automatic driving system are taken as outputs, thereby interpreting the longitudinal control model as a regression problem, and thus the mean square error can be used as an index for evaluation.
- the evaluation of the longitudinal control model of the end-to-end automatic driving system can be realized.
- the conventional evaluation method can only evaluate some single parameters, for example, by calculating the number of simulated human interventions when the simulated vehicle deviates from the center line by more than one meter, that is, the time of the automatic driving system.
- the ratio is used to estimate the steering angle. Since the longitudinal control of the vehicle mainly refers to the speed control, these methods cannot evaluate the longitudinal control model of the automatic driving system.
- the present embodiment proposes an evaluation device for the longitudinal control model of the end-to-end automatic driving system, as shown in FIG. 6, including the following devices:
- Second reading device Means for reading the coordinated world time, the east-east direction velocity component, and the north-nor direction velocity component in the reference standard file (hereinafter referred to as "second reading device") 610;
- Means for calculating a reference standard speed value at a coordinated world time instant and storing the reference standard speed value in a sequence form into the reference standard file (hereinafter referred to as "storage device” 620;
- Means for reading the estimated speed sequence of the longitudinal control model (hereinafter referred to as "third reading device") 630;
- Means for calculating a mean square error of the reference standard speed sequence and the estimated speed sequence (hereinafter referred to as "second computing device") 640.
- the time in the predetermined navigation system is usually GPS week, and the GPS week can be converted into UTC time for calculation, and then the east-direction speed component and the north-direction speed are respectively extracted by the second reading device 610.
- Component Component.
- the following table shows a set of velocity components in the east direction and velocity components in the north direction:
- Speed component 1 (m/s) Speed component 2 (m/s) Speed component 3 (m/s) Speed component 4 (m/s) Positive east direction velocity component 11.544 11.531 11.526 11.512 True north direction velocity component 0.766 0.757 0.763 0.771
- the velocity components shown in the above table may be synthesized by the storage device 620 into a reference standard velocity value as shown in the following table.
- Sequence speed 1 (m/s) Sequence speed 2 (m/s) Sequence speed 3 (m/s) Sequence speed 4 (m/s) Reference standard speed sequence 11.569 11.556 11.551 11.538
- the same number of estimated speed values as the reference standard speed sequence described above may be read from the longitudinal control model by the third reading device 630.
- the mean square error can be calculated by the second computing device 640 by the following formula.
- the formula is:
- the longitudinal control model of the automatic driving system can be evaluated by obtaining the mean square error by this calculation.
- the two velocity components in the longitudinal control model of the automatic driving system are taken as outputs, thereby interpreting the longitudinal control model as a regression problem, and thus the mean square error can be used as an index for evaluation.
- the present invention can be implemented in software and/or a combination of software and hardware.
- the various devices of the present invention can be implemented using an application specific integrated circuit (ASIC) or any other similar hardware device.
- the software program of the present invention may be executed by a processor to implement the steps or functions described above.
- the software program (including related data structures) of the present invention can be stored in a computer readable recording medium such as a RAM memory, a magnetic or optical drive or a floppy disk and the like.
- some of the steps or functions of the present invention may be implemented in hardware, for example, as a circuit that cooperates with a processor to perform various steps or functions.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computational Mathematics (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Computer Hardware Design (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Computation (AREA)
- Algebra (AREA)
- Geometry (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Control Of Driving Devices And Active Controlling Of Vehicle (AREA)
Abstract
一种端到端自动驾驶系统的纵向控制模型的评估方法及装置,其中的方法包括:根据从预定导航系统中提取的速度序列确定参考标准速度序列(S110);读取所述纵向控制模型的预估速度序列(S120);根据所述参考标准速度序列和所述预估速度序列的均方误差对所述端到端自动驾驶系统的纵向控制模型进行评估(S130)。根据预定导航系统中计算获得的参考标准速度序列和纵向控制模型中读取的预估速度序列确定均方误差,并根据该均方误差评估该纵向控制模型,从而能够客观、真实的评估该纵向模型的效果,提供的深度学习在自动驾驶领域的学习效率。
Description
本专利申请要求于2017年9月5日提交的、申请号为201710792453.6、申请人为百度在线网络技术(北京)有限公司、发明名称为“一种端到端自动驾驶系统的纵向控制模型的评估方法及装置”的中国专利申请的优先权,该申请的全文以引用的方式并入本申请中。
本发明涉及计算机领域,尤其涉及一种端到端自动驾驶系统的纵向控制模型的评估方法及装置。
随着深度学习的迅速发展以及人工智能的深入研究,汽车工业发生了革命性的变化,通过端到端的深度学习实现自动驾驶便是自动驾驶领域的一个主要研究方向。在现有技术中,通过不同的神经网络可以产生很多驾驶模型。例如,通过计算模拟车辆偏离中心线超过一米时所出发的模拟驾驶员干预数量,以及计算自主驾驶车辆的时间占比来评估自动驾驶系统模型。但该方案更偏向转向角的评估,无法满足对自动驾驶系统的纵向控制模型的评估要求,从而限制了深度学习在自动驾驶领域的发展。
发明内容
本发明解决的技术问题之一是缺乏有效的方法对自动驾驶系统的纵向控制模型进行评估。
根据本发明一方面的一个实施例,提供了一种端到端自动驾驶系统的纵向控制模型的评估方法,包括:
根据从预定导航系统中提取的速度序列确定参考标准速度序列;
读取所述纵向控制模型的预估速度序列;
根据所述参考标准速度序列和所述预估速度序列的均方误差对所述端到端自动驾驶系统的纵向控制模型进行评估。
根据本发明另一方面的一个实施例,提供了一种端到端自动驾驶系统的纵向控制模型的评估装置,包括:
用于根据从预定导航系统中提取的速度序列确定参考标准速度序列的装置;
用于读取所述纵向控制模型的预估速度序列的装置;
用于根据所述参考标准速度序列和所述预估速度序列的均方误差对所述端到端自动驾驶系统的纵向控制模型进行评估的装置。
由于本实施例根据预定导航系统中计算获得的参考标准速度序列和纵向控制模型中读取的预估速度序列确定均方误差,并根据该均方误差评估该纵向控制模型,从而能够客观、真实的评估该纵向控制模型的效果,进而提高深度学习在自动驾驶领域的学习效率。
本领域普通技术人员将了解,虽然下面的详细说明将参考图示实施例、附图进行,但本发明并不仅限于这些实施例。而是,本发明的范围是广泛的,且意在仅通过后附的权利要求限定本发明的范围。
通过阅读参照以下附图所作的对非限制性实施例所作的详细描述,本发明的其它特征、目的和优点将会变得更明显:
图1示出了根据本发明一实施例中的端到端自动驾驶系统的纵向控制模型的评估方法的流程图。
图2示出了本发明的实施例一提出的端到端自动驾驶系统的纵向控制模型的评估方法的流程图。
图3示出了本发明的实施例二提出的端到端自动驾驶系统的纵向控制模型的评估方法的流程图。
图4示出了根据本发明一实施例中的端到端自动驾驶系统的纵向控制模型的评估装置的框图。
图5示出了本发明的实施例三提出的端到端自动驾驶系统的纵向控制模型的评估装置的框图。
图6示出了本发明的实施例四提出的端到端自动驾驶系统的纵向控制模型的评估装置的框图。
附图中相同或相似的附图标记代表相同或相似的部件。
在更加详细地讨论示例性实施例之前应当提到的是,一些示例性实施例被描述成作为流程图描绘的处理或方法。虽然流程图将各项操作描述成顺序的处理,但是其中的许多操作可以被并行地、并发地或者同时实施。此外,各项操作的顺序可以被重新安排。当其操作完成时所述处理可以被终止,但是还可以具有未包括在附图中的附加步骤。所述处理可以对应于方法、函数、规程、子例程、子程序等等。
在上下文中所称“计算机设备”,也称为“电脑”,是指可以通过运行预定程序或指令来执行数值计算和/或逻辑计算等预定处理过程的智能电子设备,其可以包括处理器与存储器,由处理器执行在存储器中预存的存续指令来执行预定处理过程,或是由ASIC、FPGA、DSP等硬件执行预定处理过程,或是由上述二者组合来实现。计算机设备包括但不限于服务器、个人电脑、笔记本电脑、平板电脑、智能手机等。
所述计算机设备包括用户设备与网络设备。其中,所述用户设备包括但不限于电脑、智能手机、PDA等;所述网络设备包括但不限于单个网络服务器、多个网络服务器组成的服务器组或基于云计算(Cloud Computing)的由大量计算机或网络服务器构成的云,其中,云计算是分布式计算的一种,由一群松散耦合的计算机集组成的一个超级虚拟计算机。其中,所述计算机设备可单独运行来实现本发明,也可接入网络并通过与网络中的其他计算机设备的交互操作来实现本发明。其中,所述计算机设备所处的网络包括但不限于互联网、广域网、城域网、局域网、VPN网络等。
需要说明的是,所述用户设备、网络设备和网络等仅为举例,其他现有的或今后可能出现的计算机设备或网络如可适用于本发明,也应包含在 本发明保护范围以内,并以引用方式包含于此。
后面所讨论的方法(其中一些通过流程图示出)可以通过硬件、软件、固件、中间件、微代码、硬件描述语言或者其任意组合来实施。当用软件、固件、中间件或微代码来实施时,用以实施必要任务的程序代码或代码段可以被存储在机器或计算机可读介质(比如存储介质)中。(一个或多个)处理器可以实施必要的任务。
这里所公开的具体结构和功能细节仅仅是代表性的,并且是用于描述本发明的示例性实施例的目的。但是本发明可以通过许多替换形式来具体实现,并且不应当被解释成仅仅受限于这里所阐述的实施例。
应当理解的是,虽然在这里可能使用了术语“第一”、“第二”等等来描述各个单元,但是这些单元不应当受这些术语限制。使用这些术语仅仅是为了将一个单元与另一个单元进行区分。举例来说,在不背离示例性实施例的范围的情况下,第一单元可以被称为第二单元,并且类似地第二单元可以被称为第一单元。这里所使用的术语“和/或”包括其中一个或更多所列出的相关联项目的任意和所有组合。
应当理解的是,当一个单元被称为“连接”或“耦合”到另一单元时,其可以直接连接或耦合到所述另一单元,或者可以存在中间单元。与此相对,当一个单元被称为“直接连接”或“直接耦合”到另一单元时,则不存在中间单元。应当按照类似的方式来解释被用于描述单元之间的关系的其他词语(例如“处于...之间”相比于“直接处于...之间”,“与...邻近”相比于“与...直接邻近”等等)。
这里所使用的术语仅仅是为了描述具体实施例而不意图限制示例性实施例。除非上下文明确地另有所指,否则这里所使用的单数形式“一个”、“一项”还意图包括复数。还应当理解的是,这里所使用的术语“包括”和/或“包含”规定所陈述的特征、整数、步骤、操作、单元和/或组件的存在,而不排除存在或添加一个或更多其他特征、整数、步骤、操作、单元、组件和/或其组合。
还应当提到的是,在一些替换实现方式中,所提到的功能/动作可以按照不同于附图中标示的顺序发生。举例来说,取决于所涉及的功能/动作, 相继示出的两幅图实际上可以基本上同时执行或者有时可以按照相反的顺序来执行。
下面结合附图对本发明作进一步详细描述。
图1是根据本发明一个实施例的端到端自动驾驶系统的纵向控制模型的评估方法的流程图。
结合图1中所示,本实施例所述的端到端自动驾驶系统的纵向控制模型的评估方法包括如下步骤:
S110、根据从预定导航系统中提取的速度序列确定参考标准速度序列;
S120、读取所述纵向控制模型的预估速度序列;
S130、根据所述参考标准速度序列和所述预估速度序列的均方误差对所述端到端自动驾驶系统的纵向控制模型进行评估。
下面对各步骤做进一步详细介绍。
步骤S110中,首先提取预定导航系统中的速度序列。在本实施例中该速度序列包括但不限于以协调世界时间作为基准将当前时刻的正东方向的速度分量和正北方向的速度分量的序列。提取出上述两个速度分量的序列后,还可通过以下的计算式计算获得参考标准速度值,从而将计算获得的多个参考速度标准值存储为参考标准速度序列:
其中,v
true表示参考标准速度值,v
east表示正东方向的速度分量,v
north表示正北方向的速度分量。
步骤S120中,在计算获得参考标准速度序列之后,本实施例还可从预评估的纵向控制模型中读取预估速度序列。
可选的,读取预估速度序列的步骤包括从所述纵向控制模型中按时间顺序读取多个预估速度值,并将所述多个预估速度值以集合的形式存储为预估速度序列。
步骤S130中,参考标准速度序列和预估速度序列的均方误差可通过以下的计算式确定:
其中,MSE表示均方误差,n表示可评估的数据个数,v
true表示参考标准速度值,v
predict表示预估速度值。
在计算获得均方误差之后,可通过该均方误差的值判断所述均方误差与评估阈值的大小关系,若所述均方误差小于所述评估阈值,则所述均方误差对应的纵向控制模型的评估结果为较好。
采用本实施例提出的技术方案,根据预定导航系统中计算获得的参考标准速度序列和纵向控制模型中读取的预估速度序列确定均方误差,并根据该均方误差评估该纵向控制模型,从而能够客观、真实的评估该纵向控制模型的效果,进而提高深度学习在自动驾驶领域的学习效率。
实施例一
在本领域的现有技术中,汽车的纵向控制主要是指速度控制。但是传统的评估方法只能通过模拟驾驶员的干预从而通过计算自动驾驶车辆的时间占比来对自学驾驶系统进行评估。而实际上自动驾驶系统的纵向控制模型要复杂的多,所以传统方法都无法满足对自动驾驶系统的纵向控制模型进行评估的要求。
因此,本实施例提出了又一种端到端自动驾驶系统的纵向控制模型的评估方法,结合图2中所示,包括如下步骤:
S210、计算参考标准速度序列。
以协调世界时间为基准,读取当前时刻的正东方向速度分量v
east和正北方向速度分量v
north,并通过计算式将v
east和v
north合成为标准参考速度v
true。
在提取若干个时刻的v
east和v
north后,可以获得v
true的序列为[v
true1,v
true2,v
true3…]。
S220、读取纵向控制模型的预估速度序列。
基于协调世界时间时刻,从纵向控制模型中读取若干个预估速度值形成预估速度序列[v
predict1,v
predict2,v
predict3…]。
S230、计算获得参考标准速度序列和预估速度序列的均方误差。
参考标准速度序列和预估速度序列的均方误差可通过以下的计算式确定:
其中,MSE表示均方误差,n表示可评估的数据个数,即参考标准序列的长度。
S240、对纵向控制模型进行评估。
可首先设定一个评估阈值,并将计算获得的均方误差与该评估阈值进行比较。若所述均方误差小于所述评估阈值,则所述均方误差对应的纵向控制模型的评估结果为较好,可以作为深度学习的学习对象;若所述均方误差大于所述评估阈值,则所述均方误差对应的纵向控制模型的评估结果为较差,不推荐作为深度学习的学习对象。
在本实施例中,将自动驾驶系统的纵向控制模型中的两个速度分量作为输出,从而将纵向控制模型解释为回归问题,因此可使用均方误差作为指标进行评估。并且,通过将采集车实时回传的速度值进行计算,能够实现端到端自动驾驶系统的纵向控制模型的评估。
实施例二
在本领域的现有技术中,由于传统的评估方法只能对一些单一的参数进行评估,例如通过计算模拟车辆偏离中心线超过一米时所出发的模拟人类干预数量,即自动驾驶系统的时间占比来评估转向角。由于汽车的纵向控制主要是指速度控制,因此这些方法无法对自动驾驶系统的纵向控制模型进行评估。
因此,本实施例提出了一种端到端自动驾驶系统的纵向控制模型的评估方法,结合图3中所示,包括如下步骤:
S310、读取参考标准文件中的协调世界时间、正东方向速度分量和正北方向速度分量。
在预定导航系统中的时间通常为GPS周,为便于计算可将GPS周转换为协调世界时间(UTC time),然后再分别提取出正东方向速度分量和正北方向速度分量。下表所示的是一组正东方向速度分量和正北方向速度分量:
| 速度分量1(m/s) | 速度分量2(m/s) | 速度分量3(m/s) | 速度分量4(m/s) |
| 正东方向速度分量 | 11.544 | 11.531 | 11.526 | 11.512 |
| 正北方向速度分量 | 0.766 | 0.757 | 0.763 | 0.771 |
S320、计算在协调世界时间时刻的参考标准速度值,并将该参考标准速度值以序列的形式存储到该参考标准文件中。
在协调世界时间时刻下,可通过公式计算将上表所示的速度分量合成为如下表所示的参考标准速度值。
| 序列速度1(m/s) | 序列速度2(m/s) | 序列速度3(m/s) | 序列速度4(m/s) | |
| 参考标准速度序列 | 11.569 | 11.556 | 11.551 | 11.538 |
S330、读取所述纵向控制模型的预估速度序列。
基于协调世界时间时刻,从纵向控制模型中读取与上述参考标准速度序列包含相同数量的预估速度值。
S340、计算获得参考标准速度序列和预估速度序列的均方误差。
均方误差可通过以下公式计算获得,当速度序列中包括四组数据时,该公式则为:
通过该计算获得均方差即可对自动驾驶系统的纵向控制模型进行评估。
在本实施例中,将自动驾驶系统的纵向控制模型中的两个速度分量作为输出,从而将纵向控制模型解释为回归问题,因此可使用均方误差作为指标进行评估。
图4是根据本发明一个实施例的端到端自动驾驶系统的纵向控制模型的评估装置的框图。
结合图4中所示,本实施例所述的端到端自动驾驶系统的纵向控制模型的评估装置(以下简称“评估装置”),包括如下装置:
用于根据从预定导航系统中提取的速度序列确定参考标准速度序列的装置(以下简称“参考标准提取装置”)410;
用于读取所述纵向控制模型的预估速度序列的装置(以下简称“预估 速度读取装置”)420;
用于根据所述参考标准速度序列和所述预估速度序列的均方误差对所述端到端自动驾驶系统的纵向控制模型进行评估的装置(以下简称“评估装置”)430。
下面对各装置做进一步详细介绍。
首先通过参考标准提取装置410提取预定导航系统中的速度序列。在本实施例中,该速度序列包括但不限于通过参考标准提取装置410提取的以协调世界时间作为基准将当前时刻的正东方向的速度分量和正北方向的速度分量的序列。提取出上述两个速度分量的序列后,还可由参考标准提取装置410通过以下的计算式计算获得参考标准速度值,从而将计算获得的多个参考速度标准值存储为参考标准速度序列:
其中,v
true表示参考标准速度值,v
east表示正东方向的速度分量,v
north表示正北方向的速度分量。
在计算获得参考标准速度序列之后,本实施例还可通过预估速度读取装置420从预评估的纵向控制模型中读取预估速度序列。
可选的,在预估速度读取装置420中包括用于从所述纵向控制模型中按时间顺序读取多个预估速度值,并将所述多个预估速度值以集合的形式存储为预估速度序列的装置。
参考标准速度序列和预估速度序列的均方误差可由评估装置430通过以下的计算式确定:
其中,MSE表示均方误差,n表示可评估的数据个数,v
true表示参考标准速度值,v
predict表示预估速度值。
在计算获得均方误差之后,可由评估装置430通过该均方误差的值判断所述均方误差与评估阈值的大小关系,若所述均方误差小于所述评估阈值,则所述均方误差对应的纵向控制模型的评估结果为较好。
采用本实施例提出的技术方案,根据预定导航系统中计算获得的参考标准速度序列和纵向控制模型中读取的预估速度序列确定均方误差,并根 据该均方误差评估该纵向控制模型,从而能够客观、真实的评估该纵向控制模型的效果,进而提高深度学习在自动驾驶领域的学习效率。
实施例三
在本领域的现有技术中,汽车的纵向控制主要是指速度控制。但是传统的评估方法只能通过模拟驾驶员的干预从而通过计算自动驾驶车辆的时间占比来对自学驾驶系统进行评估。而实际上自动驾驶系统的纵向控制模型要复杂的多,所以传统方法都无法满足对自动驾驶系统的纵向控制模型进行评估的要求。
因此,本实施例提出了又一种端到端自动驾驶系统的纵向控制模型的评估装置,结合图5中所示,包括如下装置:
用于计算参考标准速度序列的装置(以下简称“第一计算装置”)510;
用于读取纵向控制模型的预估速度序列的装置(以下简称“第一读取装置”)520;
用于计算获得参考标准速度序列和预估速度序列的均方误差的装置(以下简称“误差计算装置”)530;
用于对纵向控制模型进行评估的装置(以下简称“第一评估装置”)540。
首先以协调世界时间为基准,通过第一计算装置510读取当前时刻的正东方向速度分量v
east和正北方向速度分量v
north,并通过计算式将v
east和v
north合成为标准参考速度v
true。
在提取若干个时刻的v
east和v
north后,可以获得v
true的序列为[v
true1,v
true2,v
true3…]。
然后基于协调世界时间时刻,由第一读取装置520从纵向控制模型中读取若干个预估速度值形成预估速度序列[v
predict1,v
predict2,v
predict3…]。
参考标准速度序列和预估速度序列的均方误差可由误差计算装置530通过以下的计算式确定:
其中,MSE表示均方误差,n表示可评估的数据个数,即参考标准序列的长度。
最后,可首先设定一个评估阈值,并由第一评估装置510将计算获得的均方误差与该评估阈值进行比较。若所述均方误差小于所述评估阈值,则所述均方误差对应的纵向控制模型的评估结果为较好,可以作为深度学习的学习对象;若所述均方误差大于所述评估阈值,则所述均方误差对应的纵向控制模型的评估结果为较差,不推荐作为深度学习的学习对象。
在本实施例中,将自动驾驶系统的纵向控制模型中的两个速度分量作为输出,从而将纵向控制模型解释为回归问题,因此可使用均方误差作为指标进行评估。并且,通过将采集车实时回传的速度值进行计算,能够实现端到端自动驾驶系统的纵向控制模型的评估。
实施例四
在本领域的现有技术中,由于传统的评估方法只能对一些单一的参数进行评估,例如通过计算模拟车辆偏离中心线超过一米时所出发的模拟人类干预数量,即自动驾驶系统的时间占比来评估转向角。由于汽车的纵向控制主要是指速度控制,因此这些方法无法对自动驾驶系统的纵向控制模型进行评估。
因此,本实施例提出了一种端到端自动驾驶系统的纵向控制模型的评估装置,结合图6中所示,包括如下装置:
用于读取参考标准文件中的协调世界时间、正东方向速度分量和正北方向速度分量的装置(以下简称“第二读取装置”)610;
用于计算在协调世界时间时刻的参考标准速度值,并将该参考标准速度值以序列的形式存储到该参考标准文件中的装置(以下简称“存储装置”)620;
用于读取所述纵向控制模型的预估速度序列的装置(以下简称“第三读取装置”)630;
用于计算获得参考标准速度序列和预估速度序列的均方误差的装置(以下简称“第二计算装置”)640。
在预定导航系统中的时间通常为GPS周,为便于计算可将GPS周转换为协调世界时间(UTC time),然后再通过第二读取装置610分别提取出正东方向速度分量和正北方向速度分量。下表所示的是一组正东方向速度 分量和正北方向速度分量:
| 速度分量1(m/s) | 速度分量2(m/s) | 速度分量3(m/s) | 速度分量4(m/s) | |
| 正东方向速度分量 | 11.544 | 11.531 | 11.526 | 11.512 |
| 正北方向速度分量 | 0.766 | 0.757 | 0.763 | 0.771 |
在协调世界时间时刻下,可由存储装置620通过公式计算将上表所示的速度分量合成为如下表所示的参考标准速度值。
| 序列速度1(m/s) | 序列速度2(m/s) | 序列速度3(m/s) | 序列速度4(m/s) | |
| 参考标准速度序列 | 11.569 | 11.556 | 11.551 | 11.538 |
基于协调世界时间时刻,可通过第三读取装置630从纵向控制模型中读取与上述参考标准速度序列包含相同数量的预估速度值。
均方误差可由第二计算装置640通过以下公式计算获得,当速度序列中包括四组数据时,该公式则为:
通过该计算获得均方差即可对自动驾驶系统的纵向控制模型进行评估。
在本实施例中,将自动驾驶系统的纵向控制模型中的两个速度分量作为输出,从而将纵向控制模型解释为回归问题,因此可使用均方误差作为指标进行评估。
需要注意的是,本发明可在软件和/或软件与硬件的组合体中被实施,例如,本发明的各个装置可采用专用集成电路(ASIC)或任何其他类似硬件设备来实现。在一个实施例中,本发明的软件程序可以通过处理器执行以实现上文所述步骤或功能。同样地,本发明的软件程序(包括相关的数据结构)可以被存储到计算机可读记录介质中,例如,RAM存储器,磁或光驱动器或软磁盘及类似设备。另外,本发明的一些步骤或功能可采用硬件来实现,例如,作为与处理器配合从而执行各个步骤或功能的电路。
对于本领域技术人员而言,显然本发明不限于上述示范性实施例的细节,而且在不背离本发明的精神或基本特征的情况下,能够以其他的具体 形式实现本发明。因此,无论从哪一点来看,均应将实施例看作是示范性的,而且是非限制性的,本发明的范围由所附权利要求而不是上述说明限定,因此旨在将落在权利要求的等同要件的含义和范围内的所有变化涵括在本发明内。不应将权利要求中的任何附图标记视为限制所涉及的权利要求。此外,显然“包括”一词不排除其他单元或步骤,单数不排除复数。系统权利要求中陈述的多个单元或装置也可以由一个单元或装置通过软件或者硬件来实现。第一,第二等词语用来表示名称,而并不表示任何特定的顺序。
虽然前面特别示出并且描述了示例性实施例,但是本领域技术人员将会理解的是,在不背离权利要求书的精神和范围的情况下,在其形式和细节方面可以有所变化。这里所寻求的保护在所附权利要求书中做了阐述。
Claims (17)
- 一种端到端自动驾驶系统的纵向控制模型的评估方法,包括:根据从预定导航系统中提取的速度序列确定参考标准速度序列;读取所述纵向控制模型的预估速度序列;根据所述参考标准速度序列和所述预估速度序列的均方误差对所述端到端自动驾驶系统的纵向控制模型进行评估。
- 根据权利要求1所述的方法,根据从预定导航系统中提取的速度序列确定参考标准速度序列的步骤包括:以协调世界时间作为基准将当前时刻的正东方向速度分量和正北方向的速度分量合成为参考标准速度值。
- 根据权利要求2所述的方法,根据从预定导航系统中提取的速度序列确定参考标准速度序列的步骤还包括:将计算获得的多个参考速度标准值存储为参考标准速度序列。
- 根据权利要求1所述的方法,读取所述纵向控制模型的预估速度序列的步骤包括:从所述纵向控制模型中按时间顺序读取多个预估速度值,并将所述多个预估速度值存储为预估速度序列。
- 根据权利要求1至6任意一项所述的方法,对所述端到端自动驾驶系统的纵向控制模型进行评估的步骤包括:判断所述均方误差与评估阈值的大小关系,若所述均方误差小于所述评估阈值,则所述均方误差对应的纵向控制模型的评估结果为较好。
- 一种端到端自动驾驶系统的纵向控制模型的评估装置,包括:用于根据从预定导航系统中提取的速度序列确定参考标准速度序列的装置;用于读取所述纵向控制模型的预估速度序列的装置;用于根据所述参考标准速度序列和所述预估速度序列的均方误差对所述端到端自动驾驶系统的纵向控制模型进行评估的装置。
- 根据权利要求8所述的装置,在所述用于根据从预定导航系统中提取的速度序列确定参考标准速度序列的装置中包括:用于以协调世界时间作为基准将当前时刻的正东方向速度分量和正北方向的速度分量合成为参考标准速度值的装置。
- 根据权利要求9所述的装置,在所述用于根据从预定导航系统中提取的速度序列确定参考标准速度序列的装置中还包括:用于将计算获得的多个参考速度标准值存储为参考标准速度序列的装置。
- 根据权利要求8所述的装置,在所述用于读取所述纵向控制模型的预估速度序列的装置中包括:用于从所述纵向控制模型中按时间顺序读取多个预估速度值,并将所述多个预估速度值存储为预估速度序列的装置。
- 根据权利要求8至13任意一项所述的装置,在所述用于根据所述参考标准速度序列和所述预估速度序列的均方误差对所述端到端自动驾驶系统的纵向控制模型进行评估的装置中包括:用于判断所述均方误差与评估阈值的大小,若所述均方误差小于所述评估阈值,则所述均方误差对应的纵向控制模型的评估结果为较好的装置。
- 一种计算机可读存储介质,所述计算机可读存储介质存储有计算机代码,当所述计算机代码被执行时,如权利要求1至7中任一项所述的方法被执行。
- 一种计算机程序产品,当所述计算机程序产品被计算机设备执行时,如权利要求1至7中任一项所述的方法被执行。
- 一种计算机设备,所述计算机设备包括:一个或多个处理器;存储器,用于存储一个或多个计算机程序;当所述一个或多个计算机程序被所述一个或多个处理器执行时,使得所述一个或多个处理器实现如权利要求1至7中任一项所述的方法。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710792453.6A CN107783943A (zh) | 2017-09-05 | 2017-09-05 | 一种端到端自动驾驶系统的纵向控制模型的评估方法及装置 |
| CN201710792453.6 | 2017-09-05 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2019047673A1 true WO2019047673A1 (zh) | 2019-03-14 |
Family
ID=61437619
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2018/100246 Ceased WO2019047673A1 (zh) | 2017-09-05 | 2018-08-13 | 一种端到端自动驾驶系统的纵向控制模型的评估方法及装置 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN107783943A (zh) |
| WO (1) | WO2019047673A1 (zh) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107783943A (zh) * | 2017-09-05 | 2018-03-09 | 百度在线网络技术(北京)有限公司 | 一种端到端自动驾驶系统的纵向控制模型的评估方法及装置 |
| CN109739245A (zh) * | 2019-02-19 | 2019-05-10 | 东软睿驰汽车技术(沈阳)有限公司 | 一种基于无人驾驶的端到端模型评估方法及装置 |
| US11704554B2 (en) * | 2019-05-06 | 2023-07-18 | Baidu Usa Llc | Automated training data extraction method for dynamic models for autonomous driving vehicles |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105480228A (zh) * | 2014-10-03 | 2016-04-13 | 福特全球技术公司 | 增强的车辆速度控制 |
| US20160114698A1 (en) * | 2014-10-28 | 2016-04-28 | North Carolina State University | Method and apparatus for analyzing data related to vehicle range estimation |
| CN106541941A (zh) * | 2016-10-19 | 2017-03-29 | 东风汽车公司 | 基于高精导航定位系统的自动驾驶速度控制系统及方法 |
| CN106828493A (zh) * | 2017-02-20 | 2017-06-13 | 北理慧动(常熟)车辆科技有限公司 | 一种自动驾驶车辆分层式纵向规划控制系统及方法 |
| CN107123173A (zh) * | 2017-04-25 | 2017-09-01 | 百度在线网络技术(北京)有限公司 | 用于监测无人车的行驶状态的方法和装置 |
| CN107783943A (zh) * | 2017-09-05 | 2018-03-09 | 百度在线网络技术(北京)有限公司 | 一种端到端自动驾驶系统的纵向控制模型的评估方法及装置 |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6768944B2 (en) * | 2002-04-09 | 2004-07-27 | Intelligent Technologies International, Inc. | Method and system for controlling a vehicle |
| CN105892471B (zh) * | 2016-07-01 | 2019-01-29 | 北京智行者科技有限公司 | 汽车自动驾驶方法和装置 |
| CN107450511B (zh) * | 2017-06-28 | 2019-11-26 | 百度在线网络技术(北京)有限公司 | 评估车辆控制模型的方法、装置、设备和计算机存储介质 |
-
2017
- 2017-09-05 CN CN201710792453.6A patent/CN107783943A/zh active Pending
-
2018
- 2018-08-13 WO PCT/CN2018/100246 patent/WO2019047673A1/zh not_active Ceased
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105480228A (zh) * | 2014-10-03 | 2016-04-13 | 福特全球技术公司 | 增强的车辆速度控制 |
| US20160114698A1 (en) * | 2014-10-28 | 2016-04-28 | North Carolina State University | Method and apparatus for analyzing data related to vehicle range estimation |
| CN106541941A (zh) * | 2016-10-19 | 2017-03-29 | 东风汽车公司 | 基于高精导航定位系统的自动驾驶速度控制系统及方法 |
| CN106828493A (zh) * | 2017-02-20 | 2017-06-13 | 北理慧动(常熟)车辆科技有限公司 | 一种自动驾驶车辆分层式纵向规划控制系统及方法 |
| CN107123173A (zh) * | 2017-04-25 | 2017-09-01 | 百度在线网络技术(北京)有限公司 | 用于监测无人车的行驶状态的方法和装置 |
| CN107783943A (zh) * | 2017-09-05 | 2018-03-09 | 百度在线网络技术(北京)有限公司 | 一种端到端自动驾驶系统的纵向控制模型的评估方法及装置 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107783943A (zh) | 2018-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2019047595A1 (zh) | 一种基于端到端的自动驾驶系统舒适度的评估方法及装置 | |
| JP2023021994A (ja) | 自動運転車両に対するデータ処理方法及び装置、電子機器、記憶媒体、コンピュータプログラム、ならびに自動運転車両 | |
| CN113920158B (zh) | 跟踪模型的训练与交通对象跟踪方法、装置 | |
| CN112560684A (zh) | 车道线检测方法、装置、电子设备、存储介质以及车辆 | |
| CN114091515A (zh) | 障碍物检测方法、装置、电子设备和存储介质 | |
| CN112639878A (zh) | 无监督深度预测神经网络 | |
| CN112541515A (zh) | 模型训练方法、驾驶数据处理方法、装置、介质和设备 | |
| CN114547222B (zh) | 语义地图构建方法、装置及电子设备 | |
| CN111899594A (zh) | 用于自动驾驶车辆的动态模型的自动化训练数据提取方法 | |
| CN113344016A (zh) | 深度迁移学习方法、装置、电子设备及存储介质 | |
| CN113762397B (zh) | 检测模型训练、高精度地图更新方法、设备、介质及产品 | |
| CN108280843A (zh) | 一种视频目标检测跟踪方法及设备 | |
| CN108428249A (zh) | 一种基于光流跟踪和双几何模型的初始位姿估计方法 | |
| CN110297677B (zh) | 绘制方法、装置、设备以及存储介质 | |
| CN111368824B (zh) | 一种仪表识别方法、移动设备及存储介质 | |
| JP2022523312A5 (zh) | ||
| CN109410243A (zh) | 基于区域交叠与模板匹配的多目标跟踪方法 | |
| CN115447584B (zh) | 一种车道中心线的确定方法、装置、设备及存储介质 | |
| CN110717154A (zh) | 运动轨迹的特征处理方法、设备以及计算机存储介质 | |
| CN116188917B (zh) | 缺陷数据生成模型训练方法、缺陷数据生成方法及装置 | |
| CN112070071A (zh) | 视频中的对象标注方法、装置、计算机设备及存储介质 | |
| CN116295466B (zh) | 地图生成方法、装置、电子设备、存储介质、及车辆 | |
| CN109086316B (zh) | 面向工业物联网资源的知识图谱自主构建系统 | |
| US12608899B2 (en) | Interactive labeling method for 3D dynamic object based on time series data, key frames, and interpolated frames | |
| Sun et al. | Accurate deep direct geo-localization from ground imagery and phone-grade gps |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 18854255 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 23/09/2020) |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 18854255 Country of ref document: EP Kind code of ref document: A1 |