WO2019047595A1 - 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 - Google Patents
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 Download PDFInfo
- Publication number
- WO2019047595A1 WO2019047595A1 PCT/CN2018/093356 CN2018093356W WO2019047595A1 WO 2019047595 A1 WO2019047595 A1 WO 2019047595A1 CN 2018093356 W CN2018093356 W CN 2018093356W WO 2019047595 A1 WO2019047595 A1 WO 2019047595A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- jerk
- sequence
- average
- driving system
- acceleration
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0208—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the configuration of the monitoring system
- G05B23/0213—Modular or universal configuration of the monitoring system, e.g. monitoring system having modules that may be combined to build monitoring program; monitoring system that can be applied to legacy systems; adaptable monitoring system; using different communication protocols
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24065—Real time diagnostics
Definitions
- the present invention relates to the field of computers, and in particular, to a method and apparatus for evaluating the comfort 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 comfort of an automated driving system.
- a method for evaluating comfort based on an end-to-end automatic driving system including:
- the comfort of the automatic navigation system is evaluated based on the average jerk value of the jerk sequence.
- an apparatus for evaluating comfort based on an end-to-end automatic driving system including:
- the jerk refers to the ratio of the amount of change in acceleration to the time based on the acceleration.
- the embodiment determines the average jerk value in the predetermined mileage according to the speed sequence and the time series output by the predetermined automatic driving system, and estimates the lateral control model and the longitudinal control model of the automatic driving system according to the average jerk value, thereby being able to Objectively and realistically assess the comfort of the autonomous driving system, thereby improving the learning efficiency of deep learning in the field of automatic driving.
- FIG. 1 shows a flow chart of a method for evaluating the comfort of an end-to-end automated driving system in accordance with an embodiment of the present invention.
- FIG. 2 is a flow chart showing a method for evaluating the comfort 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 for evaluating the comfort of an end-to-end automatic driving system according to a second embodiment of the present invention.
- FIG. 4 is a block diagram showing an apparatus for evaluating the comfort 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 apparatus for evaluating the comfort of an end-to-end automatic driving system according to a third embodiment of the present invention.
- Fig. 6 is a block diagram showing an apparatus for evaluating the comfort of an end-to-end automatic driving system according to a fourth embodiment 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 for evaluating the comfort of an end-to-end automated driving system in accordance with one embodiment of the present invention.
- the method for evaluating the comfort of the end-to-end automatic driving system includes the following steps:
- the jerk refers to the ratio of the amount of change in acceleration to the time based on the acceleration.
- step S110 the speed sequence and time series output by the predetermined automatic driving system are first extracted.
- the speed sequence includes, but is not limited to, extracting a predetermined number of speed values and time values corresponding to the speed values from the lateral control model and the longitudinal control model of the automatic driving system, respectively.
- the speed values therein include, but are not limited to, synthesizing the velocity component in the east-east direction of the current time and the velocity in the north-direction direction by using the coordinated world time as a reference.
- the process of synthesizing the velocity component in the east-east direction and the velocity component in the north-direction direction is obtained by the following calculation formula:
- v is the velocity value
- v east is the velocity component in the east direction
- v north is the velocity component in the north direction.
- the speed values may be stored as a speed sequence, and the time corresponding to each speed value is also stored as a time series, and then the amount of change in the speed value in a period of time corresponds to The ratio of times can obtain acceleration values over the time period and store these acceleration values as a sequence of accelerations.
- step S120 after obtaining the acceleration sequence, the embodiment may also convert the acceleration sequence into a jerk sequence.
- the jerk value is determined by the following calculation formula:
- j represents the jerk value
- ⁇ a represents the amount of acceleration change
- t represents time
- step S130 the average jerk value of the jerk sequence is determined by the following calculation formula:
- n the number of evaluable data
- j 1 , j 2 , j 3 , ..., j n the jerk value in the jerk sequence.
- the magnitude relationship between the average jerk value and the evaluation threshold may be determined. If the average jerk value is less than the evaluation threshold, the average jerk value corresponds to the automatic driving system.
- the evaluation results of the lateral control model and/or the longitudinal control model are better.
- the average jerk value greater than the evaluation threshold may be returned to the lateral control model and/or the longitudinal control model for the lateral control model and/or The longitudinal control model is further optimized.
- the average jerk value in the predetermined mileage is determined according to the speed sequence and the time series output by the predetermined automatic driving system, and the lateral control model and the vertical direction of the automatic driving system are evaluated according to the average jerk value.
- the control model can objectively and realistically evaluate the comfort of the automatic driving system, thereby improving the learning efficiency of deep learning in the field of automatic driving.
- the traffic behavior of autonomous vehicles can be estimated by selecting a cost function and following the Bellman principle, but the theoretical and computational processes of the system are still lacking. In fact, how to evaluate the comfort of the automatic driving system is more complicated than the traffic behavior, so the traditional method can not meet the requirements for evaluating the comfort of the automatic driving system.
- this embodiment proposes another evaluation method based on the end-to-end automatic driving system comfort degree, as shown in FIG. 2, including 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 the velocity v by a calculation formula.
- the sequence of v can be obtained as [v 1 , v 2 , v 3 , v 4 ...].
- time corresponding to each speed is also stored as a time series [t 1 , t 2 , t 3 , t 4 ...].
- S220 Calculate an acceleration sequence by using a speed sequence.
- the acceleration value in the time period can be obtained by the ratio of the change amount of the speed value to the corresponding time in a period of time, for example, in the time period of t 2 - t 1 , the corresponding speed change amount ⁇ v is v 2 - v 1 , then the acceleration during this time period can be calculated by the following formula:
- the acceleration values of all the time periods obtained by the calculation are stored as the acceleration sequence [a 1 , a 2 , a 3 , a 4 ...].
- the jerk value in the time period can be obtained by the ratio of the change amount of the acceleration value to the corresponding time in a period of time, for example, in the time period of t 2 -t 1 , the corresponding acceleration change amount ⁇ a is a 2 -a 1 , then the jerk in this time period can be calculated by the following formula:
- the jerk values of all the time periods obtained by the calculation are stored as a jerk sequence [j 1 , j 2 , j 3 , j 4 ...].
- the average jerk value of the jerk sequence can be determined by the following formula:
- n the number of evaluable data
- j 1 , j 2 , j 3 , ..., j n the jerk value in the jerk sequence.
- the average jerk value obtained by the calculation can be used as the average jerk within the mileage to be evaluated.
- An evaluation threshold may be first set and the calculated average jerk is compared to the evaluation threshold. If the average jerk is less than the evaluation threshold, the evaluation result of the horizontal control model and/or the vertical control model corresponding to the average jerk is better, and can be used as a learning object for deep learning; if the average jerk is greater than the evaluation threshold The evaluation result of the horizontal control model and/or the vertical control model corresponding to the average jerk is poor, and is not recommended as a learning object for deep learning.
- the two speed components in the lateral control model or the longitudinal control model of the automatic driving system are taken as outputs, thereby obtaining the jerk of the automatic driving system, thereby obtaining the average jerk of the entire driving journey, and thus can be used.
- This average jerk is evaluated as an indicator.
- the evaluation of the comfort of the end-to-end automatic driving system can be realized.
- the traffic behavior of the self-driving car can be estimated by selecting the cost function and following the Bellman principle, but the theoretical and computational process of the system is still lacking, and the comfort of the automatic driving system cannot be objectively determined. Evaluation, it is even more difficult to optimize the lateral control model and the vertical control model based on the data collected by the automated driving system.
- the present embodiment proposes an evaluation method based on the end-to-end automatic driving system comfort degree, as shown in FIG. 3, comprising the following steps:
- S310 Determine an acceleration sequence according to a speed sequence and a time sequence output by the automatic driving system.
- the lateral control system is taken as an example.
- the velocity component in the east direction and the velocity component in the north direction can be extracted first.
- the following table shows a set of velocity components in the east direction and velocity components in the north direction:
- the velocity components shown in the above table can then be synthesized by formula calculation into velocity values as shown in the following table.
- the corresponding time series is then stored.
- the time in the predetermined navigation system is usually GPS week and week second. Since the data required for calculating acceleration and jerk in this embodiment is a time period, it is not necessary to convert the intra-week second to the UTC time.
- the table below shows a set of GPS seconds corresponding to the velocity component in the east direction and the velocity component in the north direction of the above table:
- an acceleration sequence in three time periods can be calculated, and the acceleration sequence is as shown in the following table:
- Time period 1 Time period 2
- Time period 3 Acceleration value 0.0026 0.001 0.0026
- the comfort of an automated driving system can be measured by the amount of impact or the amount of impact.
- Jerk refers to the rate of change of acceleration versus time, that is, jerk. The lower the jerk index, the stronger the comfort.
- the maximum lateral acceleration that the human body can tolerate is about 0.4 to 1.0 ms 3 . Therefore, 0.1 ms 3 can be used as the evaluation threshold threshold in this embodiment.
- the average jerk value of the jerk sequence obtained in step S320 is 0.00016 ms 3 , and the average jerk value is smaller than the evaluation threshold.
- the comparison result in step S330 is that the average jerk value is smaller than the evaluation threshold, so the evaluation result of the lateral control model can be determined to be good, which is suitable as a learning object for deep learning.
- the average jerk value of a certain time period is greater than the evaluation threshold, the average jerk value may be returned to the lateral control model, so that the lateral control model can be optimized according to the average jerk value to avoid An average jerk value greater than the evaluation threshold is generated.
- the lateral control model can not output a control amount greater than the average jerk value during normal driving to meet the passenger's need for comfort.
- FIG. 4 is a block diagram of an apparatus for evaluating the comfort of an end-to-end automatic driving system, in accordance with one embodiment of the present invention.
- the end-to-end automatic driving system comfort evaluation device (hereinafter referred to as “evaluation device”) according to the embodiment includes the following devices:
- Means for determining an acceleration sequence according to a speed sequence and a time series output by the automatic driving system (hereinafter referred to as "sequence extraction device” 410);
- Means for determining a jerk sequence according to the acceleration sequence (hereinafter referred to as "a jerk sequence determining device") 420;
- Means for evaluating the comfort of the automatic navigation system based on the average jerk value of the jerk sequence (hereinafter referred to as "evaluation device") 430.
- the speed sequence and time series output by the predetermined automatic driving system are first extracted by the sequence extracting means 410.
- the speed sequence includes, but is not limited to, extracting a predetermined number of speed values and time values corresponding to the speed values from the lateral control model and the longitudinal control model of the automatic driving system, respectively.
- the speed values therein include, but are not limited to, synthesizing the velocity component in the east-east direction of the current time and the velocity in the north-direction direction by using the coordinated world time as a reference.
- the process of synthesizing the velocity component in the east-east direction and the velocity component in the north-direction may be obtained by the sequence extraction device 410 by the following calculation formula:
- v is the velocity value
- v east is the velocity component in the east direction
- v north is the velocity component in the north direction.
- the speed values may be stored as a speed sequence by the sequence extracting means 410, and the time corresponding to each speed value is also stored as a time series, and then passed through the sequence extracting means 410 for a period of time.
- the ratio of the amount of change in the internal velocity value to the corresponding time obtains the acceleration values during the time period and stores the acceleration values as an acceleration sequence.
- the present embodiment can also convert the acceleration sequence into a jerk sequence by the jerk sequence determining device 420.
- the jerk value is determined by the following calculation formula:
- j represents the jerk value
- ⁇ a represents the amount of acceleration change
- t represents time
- the average jerk value of the jerk sequence can be determined by the evaluation device 430 by the following calculation formula:
- n the number of evaluable data
- j 1 , j 2 , j 3 , ..., j n the jerk value in the jerk sequence.
- the evaluation device 430 may determine the magnitude relationship between the average jerk value and the evaluation threshold. If the average jerk value is less than the evaluation threshold, the average jerk value corresponds to the automatic The evaluation results of the lateral control model and/or the longitudinal control model of the driving system are better.
- the average jerk value greater than the evaluation threshold may be returned by the evaluation device 430 to the lateral control model and/or the longitudinal control model for the lateral control model. And/or longitudinal control models for further optimization.
- the average jerk value in the predetermined mileage is determined according to the speed sequence and the time series output by the predetermined automatic driving system, and the lateral control model and the vertical direction of the automatic driving system are evaluated according to the average jerk value.
- the control model can objectively and realistically evaluate the comfort of the automatic driving system, thereby improving the learning efficiency of deep learning in the field of automatic driving.
- the traffic behavior of autonomous vehicles can be estimated by selecting a cost function and following the Bellman principle, but the theoretical and computational processes of the system are still lacking. In fact, how to evaluate the comfort of the automatic driving system is more complicated than the traffic behavior, so the traditional method can not meet the requirements for evaluating the comfort of the automatic driving system.
- the present embodiment proposes another evaluation device based on the end-to-end automatic driving system comfort, which, as shown in FIG. 5, includes the following devices:
- Reading device Means for reading a speed sequence and a time series (hereinafter referred to as "reading device") 510;
- Means for calculating an acceleration sequence by a speed sequence (hereinafter referred to as "first computing device") 520;
- Means for obtaining a jerk sequence by acceleration sequence calculation (hereinafter referred to as "second computing device") 530;
- Means for averaging the jerk sequence to obtain a mileage average jerk (hereinafter referred to as "third computing device") 540;
- a device for evaluating the comfort of the automatic driving system (hereinafter referred to as "first evaluation device") 550.
- the east-direction velocity component v east and the north-north velocity component v north of the current time are read by the reading device 510, and v east and v north are synthesized into the velocity v by a calculation formula.
- the sequence of v can be obtained as [v 1 , v 2 , v 3 , v 4 ...].
- time corresponding to each speed is also stored by the reading device 510 as a time series [t 1 , t 2 , t 3 , t 4 ...].
- the acceleration value in the time period can be obtained by the ratio of the change amount of the speed value to the corresponding time in the first calculating means 520, for example, in the time period of t 2 - t 1 , the corresponding speed change amount ⁇ v is v 2 -v 1 , and the acceleration during this period can be calculated by the following formula:
- the acceleration values of all the time periods obtained by the calculation are stored by the first calculating means 520 as the acceleration sequence [a 1 , a 2 , a 3 , a 4 ...].
- the jerk value in the time period can be obtained by the ratio of the change amount of the acceleration value to the corresponding time in the second calculating means 530, for example, the corresponding acceleration change in the time period of t 2 -t 1
- the amount ⁇ a is a 2 -a 1
- the jerk in this period can be calculated by the following formula:
- the jerk values of all the time periods obtained by the calculation are stored by the second calculating means 530 as jerk sequences [j 1 , j 2 , j 3 , j 4 ...].
- the average jerk value value of the jerk sequence is determined by the third computing device 540 by the following formula:
- n the number of evaluable data
- j 1 , j 2 , j 3 , ..., j n the jerk value in the jerk sequence.
- the obtained average jerk value calculated by the third calculating means 540 can be used as the average jerk within the mileage to be evaluated.
- An evaluation threshold may be first set and compared by the first evaluation device 550 to the calculated average jerk. If the average jerk is less than the evaluation threshold, the evaluation result of the horizontal control model and/or the vertical control model corresponding to the average jerk is better, and can be used as a learning object for deep learning; if the average jerk is greater than the evaluation threshold The evaluation result of the horizontal control model and/or the vertical control model corresponding to the average jerk is poor, and is not recommended as a learning object for deep learning.
- the two speed components in the lateral control model or the longitudinal control model of the automatic driving system are taken as outputs, thereby obtaining the jerk of the automatic driving system, thereby obtaining the average jerk of the entire driving journey, and thus can be used.
- This average jerk is evaluated as an indicator.
- the evaluation of the comfort of the end-to-end automatic driving system can be realized.
- the traffic behavior of the self-driving car can be estimated by selecting the cost function and following the Bellman principle, but the theoretical and computational process of the system is still lacking, and the comfort of the automatic driving system cannot be objectively determined. Evaluation, it is even more difficult to optimize the lateral control model and the vertical control model based on the data collected by the automated driving system.
- the present embodiment proposes an evaluation device based on the end-to-end automatic driving system comfort, which, as shown in FIG. 6, includes the following devices:
- Means for determining an acceleration sequence according to a speed sequence and a time series output by the automatic driving system hereinafter referred to as "fourth computing device" 610;
- Means for determining a jerk sequence according to an acceleration sequence hereinafter referred to as "fifth computing device" 620;
- Second evaluation device Means for determining the magnitude relationship between the average jerk value and the evaluation threshold (hereinafter referred to as "second evaluation device") 630;
- Means for returning the average jerk value greater than the evaluation threshold to the lateral control model (hereinafter referred to as "optimizing device") 640.
- the lateral control system is taken as an example.
- the fourth east computing device 610 can first extract the positive east direction velocity component and the northeast direction velocity component.
- the following table shows a set of velocity components in the east direction and velocity components in the north direction:
- the velocity components shown in the above table can then be synthesized by the fourth computing device 610 by formula calculation into velocity values as shown in the following table.
- the corresponding time series is then stored by the fourth computing device 610.
- the time in the predetermined navigation system is usually GPS week and week second. Since the data required for calculating acceleration and jerk in this embodiment is a time period, it is not necessary to convert the intra-week second to the UTC time.
- the table below shows a set of GPS seconds corresponding to the velocity component in the east direction and the velocity component in the north direction of the above table:
- an acceleration sequence over three time periods can be calculated by the fourth computing device 610, the acceleration sequence being as shown in the following table:
- Time period 1 Time period 2
- Time period 3 Acceleration value 0.0026 0.001 0.0026
- the acceleration change amount between the time period 1 and the time period 2 can be calculated by the fifth calculating means 620 to be 0.0016, between the time period 2 and the time period 3.
- the amount of change in acceleration is 0.0016.
- the duration of each two time periods is 10 seconds, and the jerk sequence within the mileage is [0.00016, 0.00016].
- the comfort of an automated driving system can be measured by the amount of impact or the amount of impact.
- Jerk refers to the rate of change of acceleration versus time, that is, jerk. The lower the jerk index, the stronger the comfort.
- the maximum lateral acceleration that the human body can tolerate is about 0.4 to 1.0 ms 3 . Therefore, 0.1 ms 3 can be used as the evaluation threshold threshold in this embodiment.
- the second estimator 630 can determine that the average jerk value of the jerk sequence is 0.00016 ms 3 , which is less than the evaluation threshold. Therefore, the evaluation result of the lateral control model can be determined to be better by the second evaluation device 630, which is suitable as a learning object for deep learning.
- the average jerk value of a certain time period is greater than the evaluation threshold, the average jerk value may be returned to the lateral control model by the optimization device 640 to enable the lateral control model to be based on the average jerk value. Optimized to avoid generating an average jerk value greater than the evaluation threshold. After a period of deep learning, the lateral control model can not output a control amount greater than the average jerk value during normal driving to meet the passenger's need for comfort.
- 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
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Control Of Driving Devices And Active Controlling Of Vehicle (AREA)
- Navigation (AREA)
Abstract
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,方法包括:根据自动驾驶系统输出的速度序列和时间序列确定加速度序列;根据加速度序列确定加加速度序列;根据加加速度序列的平均加加速度值对自动导航系统的舒适度进行评估。本发明根据平均加加速度值评估自动驾驶系统的横向控制模型和纵向控制模型,能够客观、真实的对自动驾驶系统的舒适度进行评估,进而提高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201710792218.9、申请人为百度在线网络技术(北京)有限公司、发明名称为“一种基于端到端的自动驾驶系统舒适度的评估方法及装置”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,尤其涉及一种基于端到端的自动驾驶系统舒适度的评估方法及装置。
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,通过选择成本函数并遵循贝尔曼原则对自动驾驶汽车的交通行为进行预估。但该方案仅仅是提出了使自动驾驶汽车在形式过程中避免急加速或减速的情况,并没有系统的理论和计算过程,从而无法满足对自动驾驶系统舒适度的评估要求,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舒适度进行评估。
根据本发明一方面的一个实施例,提供了一种基于端到端的自动驾驶系统舒适度的评估方法,包括:
根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列;
根据所述加速度序列确定加加速度序列;
根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估。
根据本发明另一方面的一个实施例,提供了一种基于端到端的自动驾驶系统舒适度的评估装置,包括:
用于根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列的装置;
用于根据所述加速度序列确定加加速度序列的装置;
用于根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估的装置。
所述加加速度指在加速度的基础上求加速度变化量与时间的比值。
由于本实施例根据预定自动驾驶系统输出的速度序列和时间序列确定在预定里程中的平均加加速度值,并根据该平均加加速度值评估该自动驾驶系统的横向控制模型和纵向控制模型,从而能够客观、真实的对自动驾驶系统的舒适度进行评估,进而提高的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基于端到端的自动驾驶系统舒适度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基于端到端的自动驾驶系统舒适度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基于端到端的自动驾驶系统舒适 度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基于端到端的自动驾驶系统舒适度的评估装置的框图。
图5示出了本发明的实施例三提出的基于端到端的自动驾驶系统舒适度的评估装置的框图。
图6示出了本发明的实施例四提出的基于端到端的自动驾驶系统舒适度的评估装置的框图。
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基于端到端的自动驾驶系统舒适度的评估方法的流程图。
结合图1中所示,本实施例所述的基于端到端的自动驾驶系统舒适度的评估方法包括如下步骤:
S110、根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列;
S120、根据所述加速度序列确定加加速度序列;
S130、根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估。
所述加加速度指在加速度的基础上求加速度变化量与时间的比值。
下面对各步骤做进一步详细介绍。
步骤S110中,首先提取预定自动驾驶系统输出的速度序列和时间序列。在本实施例中该速度序列包括但不限于从所述自动驾驶系统的横向控制模型和纵向控制模型中分别提取预定数量的速度值以及所述速度值对应的时间值。其中的速度值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通过合成获得。
可选的,将所述正东方向的速度分量和正北方向的速度分量进行合成的过程通过以下的计算式计算获得:
其中,v表示速度值,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。
在计算获得预定数量的速度值后,可将这些速度值存储为速度序列, 并且将每个速度值对应的时间也存储为时间序列,再通过在一个时间段内速度值的变化量与对应的时间的比值能够获得在该时间段内的加速度值,并将这些加速度值存储为加速度序列。
步骤S120中,在获得加速度序列之后,本实施例还可从将该加速度序列转换为加加速度序列。
可选的,加加速度值通过以下的计算式确定:
其中,j表示加加速度值,Δa表示加速度变化量,t表示时间。
步骤S130中,加加速度序列的平均加加速度值通过以下计算式确定:
在计算获得加加速度序列的平均加加速度值之后,可判断该平均加加速度值与评估阈值的大小关系,若该平均加加速度值小于该评估阈值,则该平均加加速度值对应的自动驾驶系统的横向控制模型和/或纵向控制模型的评估结果为较好。
可选的,对于该评估加加速度值大于该评估阈值的情况,可将大于该评估阈值的该平均加加速度值返回给横向控制模型和/或纵向控制模型,以供该横向控制模型和/或纵向控制模型进行进一步优化。
采用本实施例提出的技术方案,根据预定自动驾驶系统输出的速度序列和时间序列确定在预定里程中的平均加加速度值,并根据该平均加加速度值评估该自动驾驶系统的横向控制模型和纵向控制模型,从而能够客观、真实的对自动驾驶系统的舒适度进行评估,进而提高的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,通过选择成本函数并遵循贝尔曼原则可对自动驾驶汽车的交通行为进行预估,但仍然缺少系统的理论和计算过程。而实际上如何评估自动驾驶系统的舒适度要比交通行为复杂的多,所以传统 方法都无法满足对自动驾驶系统的舒适度进行评估的要求。
因此,本实施例提出了又一种基于端到端的自动驾驶系统舒适度的评估方法,结合图2中所示,包括如下步骤:
S210、读取速度序列和时间序列。
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合成为速度v。
在提取若干个时刻的v
east和v
north后,可以获得v的序列为[v
1,v
2,v
3,v
4…]。
并且还将每个速度对应的时间存储为时间序列[t
1,t
2,t
3,t
4…]。
S220、通过速度序列计算加速度序列。
通过在一个时间段内速度值的变化量与对应的时间的比值能够获得在该时间段内的加速度值,例如在t
2-t
1的时间段内,对应的速度变化量Δv为v
2-v
1,则在该时间段内的加速度可通过以下计算式计算获得:
将计算获得的所有时间段的加速度值存储为加速度序列[a
1,a
2,a
3,a
4…]。
S230、通过加速度序列计算获得加加速度序列。
通过在一个时间段内加速度值的变化量与对应的时间的比值能够获得在该时间段内的加加速度值,例如在t
2-t
1的时间段内,对应的加速度变化量Δa为a
2-a
1,则在该时间段内的加加速度可通过以下计算式计算获得:
将计算获得的所有时间段的加加速度值存储为加加速度序列[j
1,j
2,j
3,j
4…]。
S240、对加加速度序列求平均值获得里程平均加加速度。
加加速度序列的平均加加速度值可值通过以下计算式确定:
计算获得的该平均加加速度值即可作为在待评估行驶里程内的平均 加加速度。
S250、对自动驾驶系统舒适度进行评估。
可首先设定一个评估阈值,并将计算获得的平均加加速度与该评估阈值进行比较。若该平均加加速度小于该评估阈值,则该平均加加速度对应的横向控制模型和/或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;若该平均加加速度大于该评估阈值,则该平均加加速度对应的横向控制模型和/或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,将自动驾驶系统的横向控制模型或纵向控制模型中的两个速度分量作为输出,从而获得该自动驾驶系统的加加速度,进而获得整个行驶旅程的平均加加速度,因此可使用该平均加加速度作为指标进行评估。并且,通过将采集车实时回传的速度值进行计算,能够实现对基于端到端的自动驾驶系统舒适度的评估。
实施例二
在本领域的现有技术中,通过选择成本函数并遵循贝尔曼原则可对自动驾驶汽车的交通行为进行预估,但仍然缺少系统的理论和计算过程,无法对自动驾驶系统舒适度进行客观的评估,更无法根据自动驾驶系统收集的数据对横向控制模型和纵向控制模型进行优化。
根据牛顿第二定律F=ma可知加速度与力之间的这关系,而瞬时力则等于质量与加加速度的乘积,即瞬时力越小则给乘车人员带来的冲击越小,舒适度就越高。因此,本实施例提出了一种基于端到端的自动驾驶系统舒适度的评估方法,结合图3中所示,包括如下步骤:
S310、根据自动驾驶系统输出的速度序列和时间序列确定加速度序列。
本实施例以横向控制系统为例,当自动驾驶系统开始运行时,首先可以分别提取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:
然后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速度值。
再存储对应的时间序列。在预定导航系统中的时间通常为GPS周和周内秒,由于本实施例为计算加速度和加加速度需要的数据是时间段,因此可不必将周内秒转换为协调世界时间(UTC time)。下表所示的是一组与上表的正东方向速度分量和正北方向速度分量对应的GPS秒:
时间1(s) | 时间2(s) | 时间3(s) | 时间4(s) | |
周内秒 | 442531.00 | 442531.05 | 442531.10 | 442531.15 |
根据该速度序列可以计算出在三个时间段内的加速度序列,该加速度序列如下表所示:
时间段1 | 时间段2 | 时间段3 | |
加速度值 | 0.0026 | 0.001 | 0.0026 |
S320、根据加速度序列确定加加速度序列。
由上述的三个加速度值和三个时间段的结果可知,时间段1与时间段2之间的加速度变化量为0.0016,时间段2与时间段3之间的加速度变化量为0.0016。并且,每两个时间段的时长为10秒,则在该行驶里程内的加加速度序列为[0.00016,0.00016]。
S330、判断所述平均加加速度值与评估阈值的大小关系。
自动驾驶系统舒适度可以用冲击量或者冲击力大小来衡量。Jerk(急动)作为量化舒适度的一个指标,是指加速度对时间的变化率,即加加速度,急动指数越低,舒适感越强。人体可忍受的最大横向加加速度约在0.4~1.0ms
3。因此本实施例中可采用0.1ms
3作为评估阈值阈值。
在步骤S320中获得的加加速度序列的平均加加速度值为0.00016ms
3,该平均加加速度值小于该评估阈值。
S340、将大于评估阈值的平均加加速度值返回给横向控制模型。
在步骤S330中的比较结果为该平均加加速度值小于该评估阈值,因此可以将该横向控制模型的评估结果确定为较好,适合作为深度学习的学习对象。
如果当某个时间段的平均加加速度值大于该评估阈值时,则可以将该平均加加速度值返回给该横向控制模型,以使该横向控制模型能够根据该平均加加速度值进行优化,以避免产生大于该评估阈值的平均加加速度值。经过一段时间的深度学习后,该横向控制模型能够在正常行驶过程中不输出大于该平均加加速度值的控制量,以满足乘客对舒适度的需求。
图4是根据本发明一个实施例的基于端到端的自动驾驶系统舒适度的评估装置的框图。
结合图4中所示,本实施例所述的基于端到端的自动驾驶系统舒适度的评估装置(以下简称“评估装置”),包括如下装置:
用于根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列的装置(以下简称“序列提取装置”)410;
用于根据所述加速度序列确定加加速度序列的装置(以下简称“加加速度序列确定装置”)420;
用于根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估的装置(以下简称“评估装置”)430。
下面对各装置做进一步详细介绍。
首先通过序列提取装置410提取预定自动驾驶系统输出的速度序列和时间序列。在本实施例中该速度序列包括但不限于从所述自动驾驶系统的横向控制模型和纵向控制模型中分别提取预定数量的速度值以及所述速度值对应的时间值。其中的速度值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通过合成获得。
可选的,将所述正东方向的速度分量和正北方向的速度分量进行合成的过程可由序列提取装置410通过以下的计算式计算获得:
其中,v表示速度值,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。
在计算获得预定数量的速度值后,可通过序列提取装置410将这些速度值存储为速度序列,并且将每个速度值对应的时间也存储为时间序列,再通过序列提取装置410在一个时间段内速度值的变化量与对应的时间的比值获得在该时间段内的加速度值,并将这些加速度值存储为加速度序列。
在获得加速度序列之后,本实施例还可通过加加速度序列确定装置420从将该加速度序列转换为加加速度序列。
可选的,加加速度值通过以下的计算式确定:
其中,j表示加加速度值,Δa表示加速度变化量,t表示时间。
加加速度序列的平均加加速度值可由评估装置430通过以下计算式确定:
在计算获得加加速度序列的平均加加速度值之后,可由评估装置430判断该平均加加速度值与评估阈值的大小关系,若该平均加加速度值小于该评估阈值,则该平均加加速度值对应的自动驾驶系统的横向控制模型和/或纵向控制模型的评估结果为较好。
可选的,对于该评估加加速度值大于该评估阈值的情况,可由评估装置430将大于该评估阈值的该平均加加速度值返回给横向控制模型和/或纵向控制模型,以供该横向控制模型和/或纵向控制模型进行进一步优化。
采用本实施例提出的技术方案,根据预定自动驾驶系统输出的速度序列和时间序列确定在预定里程中的平均加加速度值,并根据该平均加加速度值评估该自动驾驶系统的横向控制模型和纵向控制模型,从而能够客观、真实的对自动驾驶系统的舒适度进行评估,进而提高的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,通过选择成本函数并遵循贝尔曼原则可对自动驾驶汽车的交通行为进行预估,但仍然缺少系统的理论和计算过程。而实际上如何评估自动驾驶系统的舒适度要比交通行为复杂的多,所以传统方法都无法满足对自动驾驶系统的舒适度进行评估的要求。
因此,本实施例提出了又一种基于端到端的自动驾驶系统舒适度的评估装置,结合图5中所示,包括如下装置:
用于读取速度序列和时间序列的装置(以下简称“读取装置”)510;
用于通过速度序列计算加速度序列的装置(以下简称“第一计算装置”)520;
用于通过加速度序列计算获得加加速度序列的装置(以下简称“第二计算装置”)530;
用于对加加速度序列求平均值获得里程平均加加速度的装置(以下简称“第三计算装置”)540;
用于对自动驾驶系统舒适度进行评估的装置(以下简称“第一评估装置”)550。
以协调世界时间为基准,通过读取装置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合成为速度v。
在提取若干个时刻的v
east和v
north后,可以获得v的序列为[v
1,v
2,v
3,v
4…]。
并且还通过读取装置510将每个速度对应的时间存储为时间序列[t
1,t
2,t
3,t
4…]。
通过第一计算装置520在一个时间段内速度值的变化量与对应的时间的比值能够获得在该时间段内的加速度值,例如在t
2-t
1的时间段内,对应的速度变化量Δv为v
2-v
1,则在该时间段内的加速度可通过以下计算式计算获得:
并通过第一计算装置520将计算获得的所有时间段的加速度值存储为加速度序列[a
1,a
2,a
3,a
4…]。
通过第二计算装置530在一个时间段内加速度值的变化量与对应的时间的比值能够获得在该时间段内的加加速度值,例如在t
2-t
1的时间段内,对应的加速度变化量Δa为a
2-a
1,则在该时间段内的加加速度可通过以下计算式计算获得:
通过第二计算装置530将计算获得的所有时间段的加加速度值存储为加加速度序列[j
1,j
2,j
3,j
4…]。
加加速度序列的平均加加速度值可值由第三计算装置540通过以下计算式确定:
通过第三计算装置540计算获得的该平均加加速度值即可作为在待评估行驶里程内的平均加加速度。
可首先设定一个评估阈值,并通过第一评估装置550将计算获得的平均加加速度与该评估阈值进行比较。若该平均加加速度小于该评估阈值,则该平均加加速度对应的横向控制模型和/或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;若该平均加加速度大于该评估阈值,则该平均加加速度对应的横向控制模型和/或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,将自动驾驶系统的横向控制模型或纵向控制模型中的两个速度分量作为输出,从而获得该自动驾驶系统的加加速度,进而获得整个行驶旅程的平均加加速度,因此可使用该平均加加速度作为指标进行评估。并且,通过将采集车实时回传的速度值进行计算,能够实现对基于端到端的自动驾驶系统舒适度的评估。
实施例四
在本领域的现有技术中,通过选择成本函数并遵循贝尔曼原则可对自动驾驶汽车的交通行为进行预估,但仍然缺少系统的理论和计算过程,无 法对自动驾驶系统舒适度进行客观的评估,更无法根据自动驾驶系统收集的数据对横向控制模型和纵向控制模型进行优化。
根据牛顿第二定律F=ma可知加速度与力之间的这关系,而瞬时力则等于质量与加加速度的乘积,即瞬时力越小则给乘车人员带来的冲击越小,舒适度就越高。因此,本实施例提出了一种基于端到端的自动驾驶系统舒适度的评估装置,结合图6中所示,包括如下装置:
用于根据自动驾驶系统输出的速度序列和时间序列确定加速度序列的装置(以下简称“第四计算装置”)610;
用于根据加速度序列确定加加速度序列的装置(以下简称“第五计算装置”)620;
用于判断所述平均加加速度值与评估阈值的大小关系的装置(以下简称“第二评估装置”)630;
用于将大于评估阈值的平均加加速度值返回给横向控制模型的装置(以下简称“优化装置”)640。
本实施例以横向控制系统为例,当自动驾驶系统开始运行时,首先可以通过第四计算装置610分别提取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:
然后可由第四计算装置610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速度值。
再通过第四计算装置610存储对应的时间序列。在预定导航系统中的时间通常为GPS周和周内秒,由于本实施例为计算加速度和加加速度需要的数据是时间段,因此可不必将周内秒转换为协调世界时间(UTC time)。 下表所示的是一组与上表的正东方向速度分量和正北方向速度分量对应的GPS秒:
时间1(s) | 时间2(s) | 时间3(s) | 时间4(s) | |
周内秒 | 442531.00 | 442531.05 | 442531.10 | 442531.15 |
根据该速度序列可以通过第四计算装置610计算出在三个时间段内的加速度序列,该加速度序列如下表所示:
时间段1 | 时间段2 | 时间段3 | |
加速度值 | 0.0026 | 0.001 | 0.0026 |
根据上述的三个加速度值和三个时间段的结果,可通过第五计算装置620计算获得时间段1与时间段2之间的加速度变化量为0.0016,时间段2与时间段3之间的加速度变化量为0.0016。并且,每两个时间段的时长为10秒,则在该行驶里程内的加加速度序列为[0.00016,0.00016]。
自动驾驶系统舒适度可以用冲击量或者冲击力大小来衡量。Jerk(急动)作为量化舒适度的一个指标,是指加速度对时间的变化率,即加加速度,急动指数越低,舒适感越强。人体可忍受的最大横向加加速度约在0.4~1.0ms
3。因此本实施例中可采用0.1ms
3作为评估阈值阈值。
通过第二评估装置630可以确定加加速度序列的平均加加速度值为0.00016ms
3,该平均加加速度值小于该评估阈值。因此可以通过第二评估装置630将该横向控制模型的评估结果确定为较好,适合作为深度学习的学习对象。
如果当某个时间段的平均加加速度值大于该评估阈值时,则可以通过优化装置640将该平均加加速度值返回给该横向控制模型,以使该横向控制模型能够根据该平均加加速度值进行优化,以避免产生大于该评估阈值的平均加加速度值。经过一段时间的深度学习后,该横向控制模型能够在正常行驶过程中不输出大于该平均加加速度值的控制量,以满足乘客对舒适度的需求。
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)
- 一种基于端到端的自动驾驶系统舒适度的评估方法,包括:根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列;根据所述加速度序列确定加加速度序列;根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估。
- 根据权利要求1所述的方法,根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列的步骤包括:从所述自动驾驶系统的横向控制模型和纵向控制模型中分别提取预定数量的速度值以及所述速度值对应的时间值。
- 根据权利要求2所述的方法,根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列的步骤还包括:将计算获得的多个加速度值存储为加加速度序列。
- 根据权利要求1所述的方法,根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估的步骤包括:判断所述平均加加速度值与评估阈值的大小关系,若所述平均加加速度值小于所述评估阈值,则所述平均加加速度值对应的自动驾驶系统的横向控制模型和/或纵向控制模型的评估结果为较好。
- 根据权利要求6所述的方法,根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估的步骤还包括:将大于所述评估阈值的所述平均加加速度值返回给所述横向控制模型和/或纵向控制模型。
- 一种基于端到端的自动驾驶系统舒适度的评估装置,包括:用于根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列的装置;用于根据所述加速度序列确定加加速度序列的装置;用于根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估的装置。
- 根据权利要求8所述的装置,在所述用于根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列的装置中包括:用于从所述自动驾驶系统的横向控制模型和纵向控制模型中分别提取预定数量的速度值以及所述速度值对应的时间值的装置。
- 根据权利要求9所述的装置,在所述用于根据所述自动驾驶系统输出的速度序列和时间序列确定加速度序列的装置中还包括:用于将计算获得的多个加速度值存储为加加速度序列的装置。
- 根据权利要求8所述的装置,在所述用于根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估的装置中包括:用于判断所述平均加加速度值与评估阈值的大小关系,若所述平均加加速度值小于所述评估阈值,则所述平均加加速度值对应的自动驾驶系统的横向控制模型和/或纵向控制模型的评估结果为较好的装置。
- 根据权利要求13所述的装置,在所述用于根据所述加加速度序列的平均加加速度值对所述自动导航系统的舒适度进行评估的装置中还包括:用于将大于所述评估阈值的所述平均加加速度值返回给所述横向控制模型和/或纵向控制模型的装置。
- 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 |
---|---|---|---|
CN201710792218.9A CN107807625B (zh) | 2017-09-05 | 2017-09-05 | 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 |
CN201710792218.9 | 2017-09-05 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2019047595A1 true WO2019047595A1 (zh) | 2019-03-14 |
Family
ID=61569853
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2018/093356 WO2019047595A1 (zh) | 2017-09-05 | 2018-06-28 | 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 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN107807625B (zh) |
WO (1) | WO2019047595A1 (zh) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111829793A (zh) * | 2020-08-03 | 2020-10-27 | 广州导远电子科技有限公司 | 基于组合定位的驾驶过程舒适性评价方法、装置、系统 |
US20210170972A1 (en) * | 2019-12-06 | 2021-06-10 | Toyota Jidosha Kabushiki Kaisha | Information processing device, information processing method, and program |
CN115489535A (zh) * | 2021-06-17 | 2022-12-20 | 大众汽车股份公司 | 用于确定用于运行中的机动车的舒适度值的方法 |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107807625B (zh) * | 2017-09-05 | 2020-05-12 | 百度在线网络技术(北京)有限公司 | 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 |
CN109445426A (zh) * | 2018-09-06 | 2019-03-08 | 百度在线网络技术(北京)有限公司 | 自动驾驶模式的切换方法、装置及可读存储介质 |
CN111703423B (zh) * | 2019-03-18 | 2022-05-06 | 毫末智行科技有限公司 | 自动驾驶车辆的纵向控制安全监测方法及系统 |
CN109934511A (zh) * | 2019-03-26 | 2019-06-25 | 中国联合网络通信集团有限公司 | 公交车舒适度的评价方法和系统 |
CN110126815B (zh) * | 2019-04-10 | 2021-03-09 | 百度在线网络技术(北京)有限公司 | 用于协助交通工具的驾驶控制的方法、装置、设备和介质 |
CN114550442B (zh) * | 2021-12-31 | 2023-11-03 | 中智行(上海)交通科技有限公司 | 一种基于车路协同的自动驾驶车辆状态评估及预警方法 |
CN114235442B (zh) * | 2022-02-23 | 2022-05-10 | 国汽智控(北京)科技有限公司 | 自动驾驶车辆性能测试方法、装置、设备及存储介质 |
CN115447616B (zh) * | 2022-10-26 | 2024-05-17 | 重庆长安汽车股份有限公司 | 车辆驾驶客观指标的生成方法及装置 |
CN115973150B (zh) * | 2023-03-17 | 2023-06-23 | 中国标准化研究院 | 一种车辆自适应巡航的加速控制方法、装置及电子设备 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104648284A (zh) * | 2013-11-22 | 2015-05-27 | 福特全球技术公司 | 自主车辆模式 |
CN107000747A (zh) * | 2014-11-10 | 2017-08-01 | 马自达汽车株式会社 | 车辆加减速度控制装置 |
CN107107912A (zh) * | 2014-12-24 | 2017-08-29 | 株式会社电装 | 车辆控制装置以及车辆控制方法 |
CN107807625A (zh) * | 2017-09-05 | 2018-03-16 | 百度在线网络技术(北京)有限公司 | 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 |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101079903B1 (ko) * | 2009-08-24 | 2011-11-04 | 엘에스산전 주식회사 | 열차자동운전에서의 속도 제어장치 및 방법 |
JP5414454B2 (ja) * | 2009-10-23 | 2014-02-12 | 日立オートモティブシステムズ株式会社 | 車両運動制御装置 |
KR101543156B1 (ko) * | 2014-04-10 | 2015-08-07 | 현대자동차주식회사 | 차량 속도 추정 장치 및 방법 |
CN104150301B (zh) * | 2014-07-25 | 2016-05-04 | 永大电梯设备(中国)有限公司 | 电梯运行曲线修正方法 |
KR101714276B1 (ko) * | 2015-12-14 | 2017-03-08 | 현대자동차주식회사 | 주정차시 차량 저크를 방지하기 위한 제동 제어 방법 |
CN105836560B (zh) * | 2016-03-15 | 2018-07-03 | 江苏通用电梯有限公司 | 电梯舒适度测试系统 |
-
2017
- 2017-09-05 CN CN201710792218.9A patent/CN107807625B/zh active Active
-
2018
- 2018-06-28 WO PCT/CN2018/093356 patent/WO2019047595A1/zh active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104648284A (zh) * | 2013-11-22 | 2015-05-27 | 福特全球技术公司 | 自主车辆模式 |
CN107000747A (zh) * | 2014-11-10 | 2017-08-01 | 马自达汽车株式会社 | 车辆加减速度控制装置 |
CN107107912A (zh) * | 2014-12-24 | 2017-08-29 | 株式会社电装 | 车辆控制装置以及车辆控制方法 |
CN107807625A (zh) * | 2017-09-05 | 2018-03-16 | 百度在线网络技术(北京)有限公司 | 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 |
Non-Patent Citations (1)
Title |
---|
GAO CHENXIAO: "Driving behaviour prediction and driving pattern recognition of pure electric vehicles", CHINA MASTERS`S THESES FULL-TEXT DATABASE, 30 April 2017 (2017-04-30), pages 40 - 42 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20210170972A1 (en) * | 2019-12-06 | 2021-06-10 | Toyota Jidosha Kabushiki Kaisha | Information processing device, information processing method, and program |
CN111829793A (zh) * | 2020-08-03 | 2020-10-27 | 广州导远电子科技有限公司 | 基于组合定位的驾驶过程舒适性评价方法、装置、系统 |
CN115489535A (zh) * | 2021-06-17 | 2022-12-20 | 大众汽车股份公司 | 用于确定用于运行中的机动车的舒适度值的方法 |
DE102021206233A1 (de) | 2021-06-17 | 2022-12-22 | Volkswagen Aktiengesellschaft | Verfahren zum Ermitteln eines Komfortwerts für ein betriebenes Kraftfahrzeug |
DE102021206233B4 (de) | 2021-06-17 | 2023-05-11 | Volkswagen Aktiengesellschaft | Verfahren zum Ermitteln eines Komfortwerts für ein betriebenes Kraftfahrzeug |
Also Published As
Publication number | Publication date |
---|---|
CN107807625A (zh) | 2018-03-16 |
CN107807625B (zh) | 2020-05-12 |
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 | |
CN110235148B (zh) | 训练动作选择神经网络 | |
CN109670970B (zh) | 一种驾驶行为评分方法、装置及计算机可读存储介质 | |
US20220405682A1 (en) | Inverse reinforcement learning-based delivery means detection apparatus and method | |
CN109409971A (zh) | 异常订单处理方法及装置 | |
JP2018136934A (ja) | 自動車運行に対応して運転者プロファイリングをするためのシステムおよび方法 | |
CN110209467B (zh) | 一种基于机器学习的弹性资源扩展方法和系统 | |
DE102020211472A1 (de) | System und verfahren zur verbesserung eines nicht inertialen tracking-systems mit inertialen beschränkungen | |
CN110781728B (zh) | 一种人脸朝向估计方法、装置、电子设备及存储介质 | |
CN113643329B (zh) | 一种基于孪生注意力网络的在线更新目标跟踪方法和系统 | |
CN113689699B (zh) | 一种交通流的预测方法、装置、电子设备及存储介质 | |
WO2019047673A1 (zh) | 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 | |
CN108734361A (zh) | 拼车订单处理方法和装置 | |
CN110103987B (zh) | 应用于自动驾驶车辆的决策规划方法和装置 | |
CN107450511A (zh) | 评估车辆控制模型的方法、装置、设备和计算机存储介质 | |
CN112578419B (zh) | 一种基于gru网络和卡尔曼滤波的gps数据重构方法 | |
CN116776208B (zh) | 地震波分类模型的训练方法、地震选波方法、设备及介质 | |
CN113836617A (zh) | 基于时间序列模型的掘进面前方裂隙类比预测方法及系统 | |
WO2023241388A1 (zh) | 模型训练方法及装置、补能意图识别方法及装置、设备、介质 | |
CN107808360B (zh) | 一种端到端自动驾驶系统的数据处理方法及装置 | |
WO2023050749A1 (zh) | 轨迹预测方法、装置、电子设备及存储介质 | |
CN110910191A (zh) | 拼车订单生成方法及设备 | |
CN115104114A (zh) | 评估方法、评估装置、电子设备及计算机可读存储介质 | |
CN106153046A (zh) | 一种基于自适应Kalman滤波的陀螺随机噪声AR建模方法 | |
CN115200584A (zh) | 一种路径规划方法、装置、设备及可读存储介质 |
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: 18853683 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 11/08/2020) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 18853683 Country of ref document: EP Kind code of ref document: A1 |