WO2022144010A1 - 无人驾驶设备的控制 - Google Patents

无人驾驶设备的控制 Download PDF

Info

Publication number
WO2022144010A1
WO2022144010A1 PCT/CN2021/144012 CN2021144012W WO2022144010A1 WO 2022144010 A1 WO2022144010 A1 WO 2022144010A1 CN 2021144012 W CN2021144012 W CN 2021144012W WO 2022144010 A1 WO2022144010 A1 WO 2022144010A1
Authority
WO
WIPO (PCT)
Prior art keywords
error
control
control amount
moment
unmanned
Prior art date
Application number
PCT/CN2021/144012
Other languages
English (en)
French (fr)
Inventor
王志超
颜诗涛
李小卿
ru 田润
任冬淳
樊明宇
夏华夏
Original Assignee
北京三快在线科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 北京三快在线科技有限公司 filed Critical 北京三快在线科技有限公司
Publication of WO2022144010A1 publication Critical patent/WO2022144010A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
    • G05D1/0238—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using obstacle or wall sensors
    • G05D1/024—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using obstacle or wall sensors in combination with a laser
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0214—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory in accordance with safety or protection criteria, e.g. avoiding hazardous areas
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of the vehicle
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
    • G05D1/0246—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means
    • G05D1/0253—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means extracting relative motion information from a plurality of images taken successively, e.g. visual odometry, optical flow
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0257—Control of position or course in two dimensions specially adapted to land vehicles using a radar
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0259—Control of position or course in two dimensions specially adapted to land vehicles using magnetic or electromagnetic means
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle

Definitions

  • the present application relates to the field of unmanned driving, and in particular, to the control of unmanned equipment.
  • unmanned devices will be gradually applied to people's daily life, bringing more convenient services to people's lives.
  • the application provides a control of an unmanned device, and the application adopts the following technical solutions:
  • the application provides a control method for an unmanned device, including:
  • the unmanned device is controlled according to the actual control amount.
  • control variable error prediction model inputting the state data into a control variable error prediction model to obtain the actual error of the control variable for the unmanned device at the next moment, including:
  • the state data and the reference control amount error are input into the control amount error prediction model to obtain the control amount error for the next moment.
  • control variable error prediction model includes a first sub-model and a second sub-model
  • the state data is input into the control variable error prediction model, and the actual error of the control variable for the unmanned device at the next moment is obtained, including:
  • the first sub-model is used to determine the error offset of the control variable for the unmanned device at the next moment, as the first offset, and according to the first offset and the basic error of the first control amount, determine the actual error of the control amount for the unmanned device at the next moment;
  • the second sub-model is used to determine the error offset of the control amount for the unmanned device at the next moment, as the second offset, and according to the second offset and the basic error of the first control variable, determine the actual error of the control variable for the unmanned vehicle at the next moment.
  • the change frequency of the control quantity corresponding to the unmanned device at the current moment is determined.
  • control variable error offset for the unmanned device at the next moment as the first offset, including:
  • the control variable error offset of the unmanned vehicle at the current moment is used as the second reference offset, and the control variable for the unmanned device at the current moment determined at the previous moment is obtained.
  • the basic error as the basic error of the second control quantity
  • the second control variable basic error, the first reference offset and the second reference offset are input into the first sub-model to obtain the first offset.
  • control variable error offset for the unmanned device at the next moment as the second offset, including:
  • the first reference offset and the second reference offset are input into the second sub-model to obtain the second offset.
  • determining the actual control amount for the unmanned device at the next moment according to the actual error of the control amount and the reference control amount including:
  • the actual error of the control amount is adjusted by the first adjustment method to obtain the first adjusted control amount error, and the first adjusted control amount error is obtained according to the first adjustment method. and the reference control amount, to determine the actual control amount for the unmanned device at the next moment;
  • the actual error of the control amount is adjusted by a second adjustment method to obtain a second adjusted control amount error, and according to the second adjusted control amount the error and the reference control amount, to determine the actual control amount for the unmanned device at the next moment;
  • the actual control amount for the unmanned device at the next moment determined based on the first adjustment method and the actual control amount for the unmanned device at the current moment determined at the previous moment The amount of change in the control amount between the actual control amounts of the equipment is greater than the actual control amount for the unmanned device at the next moment determined based on the second adjustment method and the amount determined at the previous moment.
  • the change amount of the control amount between the actual control amounts of the unmanned device at the current moment is greater than the actual control amount for the unmanned device at the next moment determined based on the second adjustment method and the amount determined at the previous moment.
  • the application provides a control device for an unmanned device, including:
  • the acquisition module is used to acquire the status data of the unmanned equipment at the current moment
  • a prediction module configured to predict the control amount for the unmanned device at the next moment according to the state data, as a reference control amount
  • an input module configured to input the state data into a control variable error prediction model to obtain the actual error of the control variable for the unmanned device at the next moment;
  • a determining module configured to determine the actual control amount for the unmanned device at the next moment according to the actual error of the control amount and the reference control amount
  • the control module is used for controlling the unmanned equipment according to the actual control quantity.
  • the present application provides a non-transitory computer-readable storage medium, where the non-transitory computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, implements the above-mentioned control method for an unmanned vehicle.
  • the present application provides an unmanned device, including a memory, a processor, and a computer program stored on the memory and running on the processor, and the processor implements the control method for the unmanned device when the processor executes the program. .
  • the present application provides a computer program product or computer program, the computer program product or computer program including computer instructions, the computer instructions being stored in a non-transitory computer-readable storage medium, the processor of the computer device can be obtained from the non-transitory computer.
  • the computer instruction is read from the storage medium, and the processor executes the computer instruction, so that the computer device implements the above-mentioned control method for the unmanned vehicle.
  • FIG. 1 is a schematic flowchart of a control method for an unmanned device in an embodiment of the application
  • FIG. 2 is a schematic diagram of training a control quantity error prediction model, a first sub-model and a second sub-model provided by an embodiment of the present application;
  • FIG. 3 is a schematic diagram of a control device of an unmanned device provided by an embodiment of the present application.
  • FIG. 4 is a schematic diagram of the unmanned device corresponding to FIG. 1 provided by an embodiment of the present application;
  • FIG. 5 is another schematic diagram of training a control variable error prediction model, a first sub-model, and a second sub-model according to an embodiment of the present application.
  • FIG. 1 is a schematic flowchart of a control method for an unmanned device provided by an embodiment of the present application, comprising the following steps:
  • S101 Acquire state data of the unmanned device at the current moment.
  • the state data of the unmanned device at the current moment can be obtained, wherein the state mentioned here is
  • the data can refer to the speed, acceleration, angular velocity, geographic location, steering, etc. of the unmanned device at the current moment.
  • the state data can be obtained through various types of sensors set on the unmanned vehicle.
  • the point cloud data at the current moment can be obtained through the set lidar, and then the point cloud data at the current moment and the point cloud data at the previous moment can be obtained.
  • point cloud data to determine the steering of the unmanned vehicle.
  • the sensors set on the unmanned device not only include lidar, but also include inertial measurement units (Inertial Measurement Unit, IMU), cameras, etc., and how other types of sensors obtain status data, I will not give a detailed example here. explained.
  • the executing subject of the method for controlling an unmanned device may be the unmanned device itself, and of course, the server may also control the unmanned device, that is, the unmanned device can The state data at the moment is uploaded to the server, and the server can determine the actual control amount for the unmanned device at the next moment according to the obtained state data, and send the actual control amount to the unmanned device, so that the unmanned The driving device controls itself by this actual control quantity.
  • the server may also control the unmanned device, that is, the unmanned device can The state data at the moment is uploaded to the server, and the server can determine the actual control amount for the unmanned device at the next moment according to the obtained state data, and send the actual control amount to the unmanned device, so that the unmanned The driving device controls itself by this actual control quantity.
  • the following description only takes the unmanned device as the execution subject.
  • the unmanned devices mentioned in the embodiments of the present application may refer to devices that can realize automatic driving, such as unmanned vehicles, robots, and automatic distribution devices. Based on this, an unmanned device to which the control method for an unmanned device provided by the embodiments of the present application is applied can be used to perform distribution tasks in the field of distribution, for example, a business scenario in which the unmanned device is used for express delivery, logistics, takeaway, etc. .
  • the unmanned device After obtaining the above state data, the unmanned device can predict the control amount required to control itself at the next moment through the state data.
  • This control amount may be referred to as a reference control amount.
  • the control amount refers to the accelerator force, wheel steering angle, etc. that the unmanned device needs to control during the driving process.
  • the reason why the above-mentioned control quantity is called the reference control quantity is because the determined control quantity is inaccurate, that is, if the unmanned equipment at the next moment is directly controlled by the reference control quantity, Then, there will be a deviation between the actual control amount in the control process of the unmanned device at the next moment and the reference control amount.
  • the main reason for this deviation is that in the process of driving, the driverless device usually makes a decision every once in a while, and usually makes a decision first, and then controls itself through the decision, which leads to In the process of controlling itself, the unmanned equipment often cannot fully execute the determined control amount, and it has reached the next time period when it needs to make a decision. That is to say, because of the time delay, the control amount actually executed by the unmanned device does not reach the control amount that needs to be executed, so a deviation in the control amount occurs.
  • the wheel steering angle determined by the unmanned vehicle in the previous time period is 30% to the left, but in the actual execution, due to the time delay, it just turns the wheel to the left by 20%, and then enters the next step. decision over a period of time.
  • the actual control quantity determined by the unmanned device subsequently can be understood as, taking into account the time delay, etc., after the unmanned device controls itself through the actual control quantity, and finally
  • the control amount actually executed at the next moment is the control amount determined by the unmanned device at the current moment and needs to be executed at the next moment.
  • S103 Input the state data into a control variable error prediction model to obtain the actual error of the control variable for the unmanned vehicle at the next moment.
  • the unmanned device can input the above state data into the control variable error prediction model to predict the actual error of the control variable for the unmanned device at the next moment.
  • the control amount error is continuous, that is, the control amount error at each previous moment will have an impact on the subsequent control amount error. Therefore, the unmanned vehicle can first determine the control amount prediction.
  • the control amount error determined by the model for the current moment at the previous moment is used as the reference control amount error, and then the state data and the reference control amount error are input into the control amount error prediction model, and the control amount for the next moment is obtained. actual error. That is to say, the unmanned device needs to refer to the actual error of the control amount determined at the previous moment for the current moment to determine the actual error of the control amount determined at the current moment for the next moment.
  • the unmanned equipment is in different situations, and the determined control amount or the frequency of determining the control amount may be different.
  • the unmanned vehicle may not need to determine the control quantity frequently, and the deviation between the determined control quantities may not be very large. Therefore, in this case, the unmanned The frequency with which the device determines the control quantity or the difference between the determined control quantities for adjacent time intervals may be low.
  • the unmanned device may need to respond quickly, so it may be necessary to determine each control quantity more frequently, and the difference between the determined control quantities may also be large.
  • the control variable error prediction model may be obtained through offline training before executing the control method for the unmanned device provided by the embodiment of the present application, and the embodiment of the present application does not limit the type and structure of the control variable error prediction model.
  • the control variable error prediction model may include two models, a first sub-model and a second sub-model, and the first sub-model may correspond to the unmanned When the driving device is in a situation that requires a quick response, such as a curve, the second sub-model can correspond to a situation where the unmanned device is in a situation such as going straight, which does not require a fast response.
  • the unmanned device can determine the frequency of change of the control amount corresponding to the unmanned device at the current moment according to the historical state data and historical control amount of the time period set by itself before the current moment, and according to the frequency of change of the control amount. , choose which sub-model to use to determine the actual error of the control variable for the next moment.
  • the change frequency of the control variable mentioned above is used to indicate the speed of the change of the control variable in the past period of time.
  • the human-driven device is likely to be in a situation that requires a quick response, such as a curve, and if the frequency of changes in the control variable is low, it means that the driverless device does not need to frequently determine the control variable in the past. It is likely to be in a situation that does not require a quick response, such as driving straight, for a period of time.
  • the unmanned device may determine the frequency of the above-mentioned control variable change in various ways.
  • the quantity is shifted to the right by ⁇ t time, and the mean square error between the historical control quantity and the control quantity actually executed by the unmanned vehicle is determined for each translation, and the translation time with the smallest mean square error is found as the delay constant ⁇ , And through the time delay constant ⁇ , the change frequency of the control quantity is determined.
  • the way to determine the delay constant ⁇ can refer to the following formula:
  • u(t- ⁇ t) is used to represent the control quantity determined by the unmanned vehicle after translation ⁇ t
  • a(t) is used to represent the control quantity actually executed by the unmanned vehicle.
  • the unmanned device may determine the basic error of the control amount for the unmanned device at the next moment according to the above state data, as the first basic error of the control amount, and then, if the unmanned device determines It is found that the change frequency of the above-mentioned control quantity is greater than the first set frequency, then through the first sub-model, the control quantity error bias for the unmanned device at the next moment is determined as the first bias, and according to the first bias quantity and the basic error of the first control quantity to determine the actual error of the control quantity for the unmanned device itself at the next moment.
  • the second sub-model is used to determine the error offset of the control variable for the unmanned device itself at the next moment, as the second Offset, and according to the second offset and the above-mentioned basic error of the first control quantity, determine the actual error of the control quantity for the unmanned device at the next moment.
  • the above-mentioned basic error of the first control variable refers to the reference error of the control variable at the next moment determined by the unmanned device after applying the above state data to the control variable error prediction model, and the subsequent determination by the unmanned device.
  • the actual error of the control variable at the next moment is determined by combining the first offset or the second offset on the basis of the basic error of the first control variable. Since the first sub-model and the second sub-model correspond to different situations in which the unmanned device is located, the actual control amount determined by the first offset can ensure that the unmanned device is in a curve, such as a curve, which requires a fast response. Safe driving can be carried out in the case of the second offset, and the actual control amount determined by the second offset can ensure that the unmanned vehicle can drive smoothly when it is in a situation such as straight driving that does not require a rapid response.
  • the unmanned device may determine its own control variable error offset at the current moment output by the first sub-model at the previous moment, as the first reference Offset, and determine the control quantity error offset of the second sub-model at the current moment output at the previous moment, as the second reference offset, and obtain the unmanned driving at the current moment determined at the last moment.
  • the basic error of the control variable of the equipment is used as the basic error of the second control variable, and then the second basic error of the control variable, the first reference offset and the second reference offset are input into the first sub-model to obtain the above-mentioned first offset.
  • the unmanned vehicle when determining the above-mentioned first offset, the unmanned vehicle needs to refer to the control variable error offset determined by the first sub-model at time t at t-1 (that is, the first offset reference offset), the second sub-model at t-1 for the control variable error offset (ie, the second reference offset) determined by the unmanned device at time t, and the unmanned device at time t-1
  • the basic error of the control variable (the second basic error of the control variable) determined by the unmanned vehicle at time t.
  • the unmanned device determines the above-mentioned first offset
  • the following formula can be used to determine:
  • ⁇ f is the time parameter corresponding to the first sub-model
  • ⁇ f is the model parameter in the first sub-model
  • the unmanned vehicle can also first determine the control variable error offset at the current moment output by the first sub-model at the previous moment, that is, the first reference offset, and determine the second offset.
  • the unmanned device when determining the above-mentioned second offset, the unmanned device needs to refer to the control variable error offset determined by the first sub-model at time t-1 for the unmanned device at time t (ie, the first reference offset), and the control variable error offset (ie, the second reference offset) determined by the second sub-model at time t-1 for the unmanned vehicle at time t.
  • the unmanned device when determining the above-mentioned second offset, can use the following formula to determine:
  • the control variable error offset determined by the second sub-model for the current moment of the unmanned device at the last moment, that is, the second reference offset, is the control variable error bias determined by the first sub-model for the current moment of the unmanned device at the last moment, that is, the first reference bias
  • ⁇ s is the time parameter corresponding to the second sub-model
  • ⁇ s is Model parameters in the second submodel.
  • the unmanned vehicle can use the data determined by the second sub-model at the last moment as input into the first sub-model, so as to introduce the data of the second sub-model at the last moment as input. Output, lowering the output of the first submodel can cause the unmanned vehicle to over-responsive.
  • the driver may be unmanned.
  • the unmanned vehicle can use the data determined by the first sub-model at the previous moment as input into the second sub-model, so as to introduce the data of the first sub-model at the previous moment as input. Output, lowering the output of the second submodel can cause the unmanned device to respond too slowly.
  • This item is mainly used to control the variation range between the first offset output by the first sub-model and the offset output by the first sub-model at the previous moment (ie, the first reference offset).
  • the first reference offset It is used to represent the first reference offset, and when ⁇ f takes 1, this item is 0, indicating that the first sub-model does not consider the first reference output by the first sub-model at the previous moment when determining the first offset. offset, then the finally determined first offset may have a larger variation range than the first reference offset.
  • ⁇ f takes a number between 0 and 1
  • This item is mainly used to control the magnitude of change between the second offset output by the second sub-model and the offset output by the second sub-model at the previous moment (ie, the second reference offset).
  • the specific principle is basically the same as the above, and will not be described in detail here.
  • ⁇ s and ⁇ f need to satisfy the relationship: 1> ⁇ s > ⁇ f >0.
  • the specific value can be determined according to the actual demand.
  • the reason why ⁇ s > ⁇ f is guaranteed is that the first sub-model mainly deals with the situation that unmanned equipment needs to respond quickly, while the second sub-model mainly deals with If the unmanned vehicle does not need to respond quickly, then it is required that ⁇ s > ⁇ f satisfy this relationship, which can ensure that at the same moment, the first offset determined by the first sub-model and the first sub-model determined at the previous moment
  • the variation range between the first reference offsets is greater than the variation range between the second offset determined by the second sub-model and the second benchmark offset determined by the second sub-model at the last moment, so as to further satisfy the The respective applicability of the first sub-model and the second sub-model.
  • the unmanned vehicle can select one of the first sub-model and the second sub-model to determine the actual control quantity at a specific moment
  • the first sub-model and the second sub-model each The respective first offset and second offset need to be determined at each moment, so that the first sub-model and the second sub-model can provide reference data for each other at the next moment.
  • control variable error prediction model the first sub-model, and the second sub-model may be trained in advance.
  • the process is shown in Figure 2.
  • FIG. 2 is a schematic diagram of training a control quantity error prediction model, a first sub-model, and a second sub-model according to an embodiment of the present application.
  • the state data contained in the training sample may be input into the control quantity error prediction model, for example, the state data is Xt in FIG. 2 . Since the control variable error prediction model includes the first sub-model and the second sub-model, in the process of training the control variable error prediction model, the first sub-model and the second sub-model need to be combined together. train.
  • the input/output layer of the control variable error prediction model is a multi-layer perceptron structure, which can receive the output from the first sub-model or the second sub-model, and the first sub-model can receive The output from the input/output layer of the control quantity error prediction model at the last moment, and the output of the second sub-model are used as input.
  • the second submodel may only receive the output of the first submodel.
  • the output of the model for the unmanned vehicle at time t+ ⁇ can be predicted by the error of this e t+ ⁇ and the control quantity.
  • the deviation between the control variables y t+ ⁇ is optimized to train the control variable error prediction model, the first sub-model and the second sub-model.
  • e t+ ⁇ and y t+ ⁇ are regarded as two probability distributions, and determined by KL divergence The loss function between the two is used to train the above three models. Other methods are not described in detail here.
  • loss functions of e t+ ⁇ and y t+ ⁇ are determined by KL divergence as J t (t, ⁇ ):
  • IO represents the output layer
  • the deviation between this e t+ ⁇ and the output layer of each output layer of the control variable error prediction model for the control variable y t+ ⁇ of the unmanned vehicle at time t+ ⁇ is optimized to control the
  • the quantitative error prediction model, the first sub-model and the second sub-model are trained.
  • the training process shown in FIG. 2 above is only illustrated by inputting the state data Xt contained in the training samples into the control variable error prediction model.
  • the state data Xt contained in the sample is input into the control quantity error prediction model after dimensionality reduction processing.
  • the size of the entire model is controlled by reducing the dimension of the features input into the control variable error prediction model.
  • the embodiments of the present application do not limit the manner in which the state data is subjected to dimensionality reduction processing.
  • a self-organizing map (SOM) manner is used to perform dimensionality reduction on the state data. For example, see Figure 5 for the training process. Compared with Figure 2, the training process shown in Figure 5 has an additional SOM process.
  • the actual error of the control variable can be adjusted through the first adjustment method to obtain the first adjustment method.
  • An adjusted control amount error, and according to the first adjusted control amount error and the above-mentioned reference control amount error, the actual control amount for the unmanned device at the next moment is determined.
  • the actual error of the control amount can be adjusted by the second adjustment method to obtain the second adjusted control amount error, and based on the second adjusted control amount error and the above
  • the reference control amount determines the actual control amount for the unmanned device itself at the next moment.
  • the second set frequency mentioned here may be the same as or different from the above-mentioned first set frequency, and may be specifically set according to actual needs.
  • the amount of change in the control amount between the actual control amount determined based on the first adjustment method for the next moment and the actual control amount determined at the previous moment for the current moment is greater than the amount determined based on the second adjustment method.
  • the amount of change in the control amount between the actual control amount for the next moment and the actual control amount for the current moment determined at the previous moment is greater than the amount determined based on the second adjustment method.
  • the frequency of change of the control quantity is high, it means that the unmanned equipment may be in a situation such as a curve, and needs to achieve the purpose of rapid response. Therefore, it is necessary to use the first adjustment method to determine the To ensure that the unmanned equipment can realize the rapid response of the unmanned equipment through the actual control quantity for the next moment determined by the control quantity error after the first adjustment.
  • the frequency of change of the control variable is low, it means that the unmanned vehicle may be in a situation such as straight driving, and does not need to respond quickly. Therefore, it is necessary to adjust the determined error of the control variable through the second adjustment method to ensure that no one is unmanned.
  • the driving device can realize the smooth driving of the unmanned device through the actual control amount for the next moment determined by the error of the second adjusted control amount.
  • the unmanned vehicle can determine the first adjusted control variable error and the second adjusted controlled variable error by using the following formula:
  • k p is the proportional coefficient
  • k d is the differential time constant
  • k i is the integral time constant.
  • the adjustment method used here is mainly based on the angle change ⁇ of the unmanned vehicle. Therefore, w 0 is used for Indicates the set angle change. This method is basically the same as the above-mentioned determination of which adjustment method is adopted by controlling the frequency of change of the quantity. That is, if the angle change is large, the unmanned device is likely to be in a situation that requires rapid response, and if the angle change is small, the unmanned device is likely to be in a situation that does not require rapid response and drives smoothly.
  • the above-mentioned first adjustment method and second adjustment method may actually be regarded as a differential (PD) controller and a proportional integral (PI) controller.
  • the second adjustment method is a proportional-integral (PI) controller, the error of the second adjusted control amount obtained by integrating in time can ensure that the actual control amount determined by the unmanned vehicle is the same as the actual control amount at the previous moment. Compared with the control amount, the change is smoother, thereby ensuring that the unmanned vehicle drives more smoothly, and because the first adjustment method is a differential (PD) controller, the determined control amount error after the first adjustment can be prominent.
  • the change of the control amount error at a specific moment ensures that the unmanned equipment can achieve an effective and rapid response when it is in a situation such as a curve that requires a rapid response.
  • control variable error prediction model adopts a multi-time scale recurrent neural network (MTRNN, Multiple Timescale Recurrent Neural Network)
  • MRNN Multiple Timescale Recurrent Neural Network
  • PID Proportion Integral Differential, Proportional Integral Differential
  • an overall network is directly used to obtain the residual control quantity of the unmanned equipment, it is necessary to construct a supervision signal of the optimal control quantity, and this supervision signal is used in the actual control of the unmanned equipment. It is often impossible to obtain, so before using the overall network to control the unmanned equipment, it is necessary to use the obtained simulation data to train the overall network, which makes the trained overall network often difficult to use. In the actual control of the driving device.
  • S104 Determine the actual control amount for the unmanned device at the next moment according to the actual error of the control amount and the reference control amount.
  • the reference control variable can be corrected by the actual error of the control variable, so as to determine the actual control variable for the unmanned device at the next moment.
  • the reference control quantity determined by the unmanned device is actually a control quantity sequence, that is, the control quantity corresponding to several consecutive times starting from the next moment, and the actual error of the control quantity determined above is Aiming at the actual error of the control quantity at the next moment, the unmanned vehicle can correct the control quantity corresponding to the next moment in the sequence of control quantities through the actual error of the control quantity, so as to obtain the actual control for itself at the next moment.
  • quantity refer to the following formula:
  • U(k) can be regarded as the above-mentioned reference control quantity. It can be seen from this formula that the reference control quantity contains multiple control quantities, that is, starting from the next time, the continuous k time The control amount, u(t) is the control amount corresponding to the next moment.
  • the unmanned device can determine the actual control amount for the next moment through the following formula:
  • u'(t) is the actual control amount for the next moment
  • u 2 (t) is the control amount for the next moment included in the reference control amount mentioned above
  • u 1 (t) is The actual error of the control amount determined by the unmanned device for the next moment. Through this formula, the actual control amount of the unmanned equipment for the next moment can be finally determined.
  • the above-mentioned reference control variable may be determined by solving the nonlinear optimization problem. Since there are various methods for solving the nonlinear optimization problem, the embodiment of the present application does not specify the specific reference control variable. Determine how to limit.
  • the unmanned device can control itself through the actual control amount, and since the actual control amount is the corrected control amount, the unmanned device uses the actual control amount to control itself.
  • the control can effectively ensure the smooth driving of the unmanned equipment, and ensure the safety of the unmanned equipment during the driving process.
  • FIG. 3 is a schematic diagram of a control device of an unmanned device provided by an embodiment of the application, including:
  • An acquisition module 301 configured to acquire the state data of the unmanned device at the current moment
  • a prediction module 302 configured to predict the control amount for the unmanned device at the next moment according to the state data, as a reference control amount
  • an input module 303 configured to input the state data into a control variable error prediction model to obtain the actual error of the control variable for the unmanned device at the next moment;
  • a determination module 304 configured to determine the actual control amount for the unmanned device at the next moment according to the actual error of the control amount and the reference control amount;
  • the control module 305 is configured to control the unmanned device according to the actual control quantity.
  • the input module 303 is configured to determine the control variable error determined by the control variable error prediction model at the previous moment for the current moment, as the reference control variable error;
  • the reference control amount error is input into the control amount error prediction model to obtain the control amount error for the next moment.
  • control variable error prediction model includes a first sub-model and a second sub-model
  • the input module 303 is configured to, according to the state data, determine the basic error of the control quantity of the unmanned device for the next moment as the first basic error of the control quantity;
  • the historical state data and historical control quantities of the set time period before the current moment determine the frequency of change of the control quantity corresponding to the unmanned device at the current moment; if it is determined that the frequency of change of the control quantity is greater than the first setting frequency, through the first sub-model, determine the control variable error offset for the unmanned device at the next moment, as the first offset, and according to the first offset and the first offset
  • the basic error of the control quantity to determine the actual error of the control quantity for the unmanned device at the next moment; if it is determined that the change frequency of the control quantity is not greater than the first set frequency, pass the second sub-model , determine the control variable error offset for the unmanned device at the next moment as the second offset, and determine the control variable based on the second offset and the first control variable basic error.
  • the input module 303 is configured to, for each historical moment included in the set time period, according to the historical control quantity corresponding to the historical moment and the historical state data corresponding to the next historical moment of the historical moment , determine the historical control variable deviation corresponding to the historical moment; according to the historical control variable deviation corresponding to each historical moment, determine the control variable variation frequency corresponding to the unmanned device at the current moment.
  • the input module 303 is configured to determine the control variable error offset of the unmanned device at the current moment output by the first sub-model at the previous moment, as the first reference offset, Determine the control variable error offset of the unmanned vehicle at the current moment output by the second sub-model at the last moment, as the second reference offset, and obtain the determined value at the last moment for the
  • the basic error of the control amount of the unmanned device at the current moment is taken as the second basic error of the control amount; the basic error of the second control amount, the first reference offset and the second reference offset are input into In the first sub-model, the first offset is obtained.
  • the input module 303 is configured to determine the control variable error offset of the unmanned device at the current moment output by the first sub-model at the previous moment, as the first reference offset, Determine the control variable error offset of the unmanned vehicle at the current moment output by the second sub-model at the previous moment, as the second reference offset; use the first reference offset and the first reference offset Two reference offsets are input into the second sub-model to obtain the second offset.
  • the determining module 304 is configured to, if it is determined that the change frequency of the control amount is greater than the second set frequency, adjust the actual error of the control amount by using a first adjustment method to obtain the first adjusted control amount error. , and according to the first adjusted control amount error and the reference control amount, determine the actual control amount for the unmanned device at the next moment; if it is determined that the change frequency of the control amount is not greater than the second Set the frequency, adjust the actual error of the control amount through the second adjustment method, obtain the second adjusted control amount error, and determine the control amount according to the second adjusted control amount error and the reference control amount The actual control amount for the unmanned device at the next moment;
  • the actual control amount for the unmanned device at the next moment determined based on the first adjustment method and the actual control amount for the unmanned device at the current moment determined at the previous moment The amount of change in the control amount between the actual control amounts of the equipment is greater than the actual control amount for the unmanned device at the next moment determined based on the second adjustment method and the amount determined at the previous moment.
  • the change amount of the control amount between the actual control amounts of the unmanned device at the current moment is greater than the actual control amount for the unmanned device at the next moment determined based on the second adjustment method and the amount determined at the previous moment.
  • Embodiments of the present application further provide a non-transitory computer-readable storage medium, where a computer program is stored in the non-transitory computer-readable storage medium, and the computer program can be used to execute the control method for an unmanned vehicle provided in FIG. 1 above.
  • the embodiment of the present application also provides a schematic structural diagram of the unmanned device shown in FIG. 4 .
  • the driverless device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and of course, it may also include hardware required by other services.
  • the processor reads the corresponding computer program from the non-volatile memory into the memory and then executes it, so as to realize the control method of the unmanned vehicle described in FIG. 1 above.
  • the embodiments of the present application do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subjects of the following processing procedures are not limited to individual logic units, and also Can be a hardware or logic device.
  • a Programmable Logic Device (such as a Field Programmable Gate Array (FPGA)) is an integrated circuit whose logic function is determined by user programming of the device.
  • HDL Hardware Description Language
  • ABEL Advanced Boolean Expression Language
  • AHDL Altera Hardware Description Language
  • HDCal JHDL
  • Lava Lava
  • Lola MyHDL
  • PALASM RHDL
  • VHDL Very-High-Speed Integrated Circuit Hardware Description Language
  • Verilog Verilog
  • the controller may be implemented in any suitable manner, for example, the controller may take the form of eg a microprocessor or processor and a computer readable medium storing computer readable program code (eg software or firmware) executable by the (micro)processor , logic gates, switches, application specific integrated circuits (ASICs), programmable logic controllers and embedded microcontrollers, examples of controllers include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicon Labs C8051F320, the memory controller can also be implemented as part of the control logic of the memory.
  • the controller may take the form of eg a microprocessor or processor and a computer readable medium storing computer readable program code (eg software or firmware) executable by the (micro)processor , logic gates, switches, application specific integrated circuits (ASICs), programmable logic controllers and embedded microcontrollers
  • ASICs application specific integrated circuits
  • controllers include but are not limited to
  • the controller in addition to implementing the controller in the form of pure computer-readable program code, the controller can be implemented as logic gates, switches, application-specific integrated circuits, programmable logic controllers and embedded devices by logically programming the method steps.
  • the same function can be realized in the form of a microcontroller, etc. Therefore, such a controller can be regarded as a hardware component, and the devices included therein for realizing various functions can also be regarded as a structure within the hardware component. Or even, the means for implementing various functions can be regarded as both a software module implementing a method and a structure within a hardware component.
  • a typical implementation device is a computer.
  • the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device Or a combination of any of these devices.
  • embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the embodiments of the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
  • computer-usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc.
  • These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions
  • the apparatus implements the functions specified in the flow or flows of the flowcharts and/or the block or blocks of the block diagrams.
  • a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
  • processors CPUs
  • input/output interfaces network interfaces
  • memory volatile and non-volatile memory
  • Memory may include forms of non-persistent memory, random access memory (RAM) and/or non-volatile memory in computer readable media, such as read only memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.
  • RAM random access memory
  • ROM read only memory
  • flash RAM flash memory
  • Computer-readable media includes both persistent and non-permanent, removable and non-removable media, and storage of information may be implemented by any method or technology.
  • Information may be computer readable instructions, data structures, modules of programs, or other data.
  • Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), Flash Memory or other memory technology, Compact Disc Read Only Memory (CD-ROM), Digital Versatile Disc (DVD) or other optical storage, Magnetic tape cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information that can be accessed by a computing device.
  • computer-readable media does not include transitory computer-readable media, such as modulated data signals and carrier waves.
  • embodiments of the present application may be provided as a method, a system or a computer program product. Accordingly, the embodiments of the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
  • computer-usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc.
  • Embodiments of the present application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer.
  • program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types.
  • Embodiments of the application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network.
  • program modules may be located in both local and remote computer storage media including storage devices.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Electromagnetism (AREA)
  • Optics & Photonics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Forklifts And Lifting Vehicles (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
  • Feedback Control In General (AREA)

Abstract

一种无人驾驶设备的控制方法及装置，控制方法包括以下步骤：获取无人驾驶设备在当前时刻的状态数据（S101），并根据状态数据，预测出下一时刻针对无人驾驶设备的控制量，作为基准控制量（S102），而后，将状态数据输入到控制量误差模型中，得到在下一时刻针对无人驾驶设备的控制量实际误差（S103），根据控制量实际误差以及确定出的基准控制量，确定在下一时刻针对无人驾驶设备的实际控制量（S104），进而根据实际控制量，对无人驾驶设备进行控制（S105）。

Description

无人驾驶设备的控制
本申请要求于2021年01月04日提交的申请号为202110000641.7、申请名称为“一种无人驾驶设备的控制方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及无人驾驶领域，尤其涉及一种无人驾驶设备的控制。
背景技术
当前，无人驾驶设备将逐步应用于人们的日常生活，为人们的生活带来更多的便利服务。
在无人驾驶设备行驶的过程中，需要时刻确定出无人驾驶设备的控制量，进而通过确定出的控制量对无人驾驶设备实施控制。
发明内容
本申请提供一种无人驾驶设备的控制，本申请采用下述技术方案：
本申请提供了一种无人驾驶设备的控制方法，包括：
获取无人驾驶设备在当前时刻的状态数据；
根据所述状态数据，预测在下一时刻针对所述无人驾驶设备的控制量，作为基准控制量；
将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差；
根据所述控制量实际误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
根据所述实际控制量，对所述无人驾驶设备进行控制。
可选地，将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差，包括：
确定所述控制量误差预测模型在上一时刻针对所述当前时刻所确定出的控制量误差，作为参照控制量误差；
将所述状态数据以及所述参照控制量误差输入到所述控制量误差预测模型中，得到针对所述下一时刻的控制量误差。
可选地，所述控制量误差预测模型中包含有第一子模型和第二子模型；
将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差，包括：
根据所述状态数据，确定针对所述下一时刻的所述无人驾驶设备的控制量基础误差，作为第一控制量基础误差；
根据所述无人驾驶设备在所述当前时刻之前设定时间段的历史状态数据以及历史控制量，确定所述无人驾驶设备在所述当前时刻对应的控制量变化频率；
若确定所述控制量变化频率大于第一设定频率，通过所述第一子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第一偏量，并根据所述第一偏量以及所述第一控制量基础误差，确定在所述下一时刻针对所述无人驾驶设备的控制量实际误差；
若确定所述控制量变化频率不大于所述第一设定频率，通过所述第二子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第二偏量，并根据所述第二偏量以及所述第一控制量基础误差，确定在所述下一时刻针对所述无人驾驶设备的控制量实际误差。
可选地，根据所述无人驾驶设备在所述当前时刻之前设定时间段的历史状态数据以及历史控制量，确定所述无人驾驶设备在所述当前时刻对应的控制量变化频率，包括：
针对所述设定时间段中包含的每个历史时刻，根据该历史时刻对应的历史控制量以及该历史时刻的下一历史时刻对应的历史状态数据，确定该历史时刻对应的历史控制量偏差；
根据每个历史时刻对应的历史控制量偏差，确定所述无人驾驶设备在所述当前时刻对应的控制量变化频率。
可选地，通过所述第一子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第一偏量，包括：
确定所述第一子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第一基准偏量，确定所述第二子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第二基准偏量，以及获取在上一时刻确定出的针对所述当前时刻的所述无人驾驶设备的控制量基础误差，作为第二控制量基础误差；
将所述第二控制量基础误差、所述第一基准偏量以及所述第二基准偏量输入到所述第一子模型中，得到所述第一偏量。
可选地，通过所述第二子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第二偏量，包括：
确定所述第一子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第一基准偏量，确定所述第二子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第二基准偏量；
将所述第一基准偏量以及所述第二基准偏量输入到所述第二子模型中，得到所述第二偏量。
可选地，根据所述控制量实际误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量，包括：
若确定所述控制量变化频率大于第二设定频率，通过第一调整方式对所述控制量实际误差进行调整，得到第一调整后控制量误差，并根据所述第一调整后控制量误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
若确定所述控制量变化频率不大于第二设定频率，通过第二调整方式对所述控制量实际误差进行调整，得到第二调整后控制量误差，并根据所述第二调整后控制量误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
其中，基于所述第一调整方式确定出的在所述下一时刻针对所述无人驾驶设备的实际控制量与在所述上一时刻确定出的在所述当前时刻针对所述无人驾驶设备的实际控制量之间的控制量变化量，大于基于所述第二调整方式确定出的在所述下一时刻针对所述无人驾驶设备的实际控制量与在所述上一时刻确定出的在所述当前时刻针对所述无人驾驶设备的实际控制量之间的控制量变化量。
本申请提供了一种无人驾驶设备的控制装置，包括：
获取模块，用于获取无人驾驶设备在当前时刻的状态数据；
预测模块，用于根据所述状态数据，预测在下一时刻针对所述无人驾驶设备的控制量，作为基准控制量；
输入模块，用于将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差；
确定模块，用于根据所述控制量实际误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
控制模块，用于根据所述实际控制量，对所述无人驾驶设备进行控制。
本申请提供了一种非临时性计算机可读存储介质，所述非临时性计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现上述无人驾驶设备的控制方法。
本申请提供了一种无人驾驶设备，包括存储器、处理器及存储在存储器上并可在处理器 上运行的计算机程序，所述处理器执行所述程序时实现上述无人驾驶设备的控制方法。
本申请提供了一种计算机程序产品或计算机程序，该计算机程序产品或计算机程序包括计算机指令，该计算机指令存储在非临时性计算机可读存储介质中，计算机设备的处理器从非临时性计算机可读存储介质读取该计算机指令，处理器执行该计算机指令，使得该计算机设备实现上述无人驾驶设备的控制方法。
附图说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1为本申请实施例中一种无人驾驶设备的控制方法的流程示意图；
图2为本申请实施例提供的对控制量误差预测模型、第一子模型以及第二子模型进行训练的示意图；
图3为本申请实施例提供的一种无人驾驶设备的控制装置的示意图；
图4为本申请实施例提供的对应于图1的无人驾驶设备示意图；
图5为本申请实施例提供的另一种对控制量误差预测模型、第一子模型以及第二子模型进行训练的示意图。
具体实施方式
为使本申请的目的、技术方案和优点更加清楚，下面将结合本申请具体实施例及相应的附图对本申请技术方案进行清楚、完整地描述。显然，所描述的实施例仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
在无人驾驶设备行驶的过程中，需要时刻确定出无人驾驶设备的控制量，进而通过确定出的控制量对无人驾驶设备实施控制。在实际应用中，无人驾驶设备确定出的控制量，与无人驾驶设备实际执行的控制量往往存在偏差，致使无人驾驶设备可能会出现诸如偏量原定路线，行驶过程中平稳性较差，安全性较低等情况。所以，如何能够提高无人驾驶设备确定控制量的准确性，保证无人驾驶设备的平稳行驶，则是一个亟待解决的问题。
本申请实施例提供了一种无人驾驶设备的控制方法及装置，以下结合附图，详细说明本申请各实施例提供的技术方案。
图1为本申请实施例提供的一种无人驾驶设备的控制方法的流程示意图，包括以下步骤：
S101：获取无人驾驶设备在当前时刻的状态数据。
在实际应用中，无人驾驶设备在行驶过程中需要不断地对自身进行控制，因此，在本申请实施例中，可以获取无人驾驶设备在当前时刻的状态数据，其中，这里提到的状态数据可以是指无人驾驶设备在当前时刻的速度、加速度、角速度、地理位置、转向等。而该状态数据可以通过无人驾驶设备上设置的各种类型的传感器进行获取，如，可以通过设置的激光雷达，获取当前时刻的点云数据，进而根据当前时刻的点云数据以及上一时刻的点云数据，确定出无人驾驶设备的转向。当然，无人驾驶设备上设置的传感器不仅仅包括激光雷达，还可以包括诸如惯性测量单元(Inertial Measurement Unit，IMU)、相机等，而其他类型的传感器如何获取状态数据，在此就不详细举例说明了。
本申请实施例提供的无人驾驶设备的控制方法的执行主体可以是无人驾驶设备自身，当然，也可以由服务器对无人驾驶设备实施控制，即，无人驾驶设备可以将获取到的当前时刻的状态数据上传到服务器中，服务器可以根据获取到的状态数据，确定出下一时刻针对无人驾驶设备的实际控制量，并将该实际控制量发送给无人驾驶设备，以使无人驾驶设备通过该实际控制量对自身进行控制。而为了方便说明，下面仅以无人驾驶设备为执行主体进行说明。
本申请实施例中提到的无人驾驶设备可以是指无人车、机器人、自动配送设备等能够实现自动驾驶的设备。基于此，应用本申请实施例提供的无人驾驶设备的控制方法的无人驾驶 设备可以用于执行配送领域的配送任务，如，使用无人驾驶设备进行快递、物流、外卖等配送的业务场景。
S102：根据所述状态数据，预测在下一时刻针对所述无人驾驶设备的控制量，作为基准控制量。
无人驾驶设备在获取到上述状态数据后，可以通过该状态数据，预测出下一时刻需要针对自身进行控制所需的控制量。这一控制量可以称之为是基准控制量。
在本申请实施例中，控制量是指无人驾驶设备在行驶过程中需要控制的油门力度，车轮转向角度等。而之所以将上述控制量称之为是基准控制量，是因为确定出的这一控制量是不准确的，即，若是通过该基准控制量直接对下一时刻的无人驾驶设备进行控制，则无人驾驶设备在下一时刻进行控制过程中的实际控制量将与该基准控制量之间产生偏差。
这一偏差产生的主要原因在于，无人驾驶设备在行驶过程中，通常都是每隔一段时间做一次决策，并且，通常都是先做决策，再通过决策对自身实施控制，这就导致了无人驾驶设备在对自身进行控制的过程中，往往无法将确定出的控制量完全执行完，就已经到了需要做决策的下一个时间段。也就是说，因为存在时间延时的情况，无人驾驶设备实际执行的控制量并没有到达需要执行的控制量，因此，产生了控制量上的偏差。
例如，假设无人驾驶设备在上一时间段确定出的车轮转向角度为向左30％，而在实际执行时因为时间延时的情况，只是将车轮向左转了20％，就进入了下一时间段的决策。
所以，在本申请实施例中，无人驾驶设备后续确定出的实际控制量，可以理解成是在考虑时间延时等情况下，无人驾驶设备通过该实际控制量对自身进行控制后，最终在下一时刻实际执行的控制量即为无人驾驶设备在当前时刻所确定出的需要在下一时刻执行的控制量。
S103：将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差。
无人驾驶设备可以将上述状态数据输入到控制量误差预测模型中，以预测出在下一时刻针对无人驾驶设备的控制量实际误差。其中，由于在实际应用中，控制量误差是存在连续性的，即，之前的各时刻的控制量误差会对后续的控制量误差产生影响，所以，无人驾驶设备可以先确定出控制量预测模型在上一时刻针对当前时刻所确定出的控制量误差，作为参照控制量误差，进而将该状态数据以及该参照控制量误差输入到控制量误差预测模型中，得到针对下一时刻的控制量实际误差。也就是说，无人驾驶设备需要参考前一时刻所确定出的针对当前时刻的控制量实际误差，来确定出在当前时刻针对下一时刻所确定出的控制量实际误差。
在实际应用中，无人驾驶设备处于不同情况下，所确定出的控制量或是确定控制量的频率可能会有所不同。例如，在直行时，无人驾驶设备可能不需要频繁的确定控制量，并且，每次所确定出的控制量之间的偏差可能也不是很大，所以，在这种情况下，无人驾驶设备确定控制量的频率或是确定出的各相邻时间间隔的控制量之间的差异可能较低。而在转弯时，无人驾驶设备可能需要进行快速的响应，所以，可能需要较为频繁的确定各控制量，并且，确定出的控制量在前后之间的差异可能也较大。
其中，控制量误差预测模型可以在执行本申请实施例提供的无人驾驶设备的控制方法之前通过离线训练得到，本申请实施例不对控制量误差预测模型的类型和结构进行限定。为了能够准确的确定出上述控制量误差，在本申请实施例中，控制量误差预测模型中可以包含有两个模型，第一子模型和第二子模型，第一子模型可以对应到无人驾驶设备处于如弯道等需要快速响应的情况，第二子模型可以对应到无人驾驶设备处于如直行等无需快速响应的情况。
进一步地，无人驾驶设备可以根据自身在当前时刻之前设定时间段的历史状态数据以及历史控制量，确定无人驾驶设备在当前时刻所对应的控制量变化频率，并根据该控制量变化频率，选择使用哪一子模型，来确定针对下一时刻的控制量实际误差。
上述提到的控制量变化频率用于表明过去一段时间，控制量变化的快慢程度，若是控制量变化频率越高，则表明过去一段时间，无人驾驶设备需要频繁的确定控制量，因此，无人驾驶设备很可能处于如弯道等需要快速响应的情况，而若是控制量变化频率较低，则表明过 去一段时间，无人驾驶设备无需频繁的确定控制量，因此，无人驾驶设备在过去一段时间内很可能处于如直行等无需快速响应的情况。
在本申请实施例中，无人驾驶设备确定上述控制量变化频率的方式可以有多种，例如，无人驾驶设备可以基于过去一段时间的历史状态数据以及历史控制量，将确定出的历史控制量向右依次平移Δt时间，每次平移都确定出历史控制量与无人驾驶设备实际执行的控制量之间的均方误差，并找到均方误差最小的平移时间，作为时延常数τ，并通过该时延常数τ，确定该控制量变化频率。时延常数τ的确定方式可以参考如下公式：
τ＝argmin Δt‖u(t-Δt)-a(t)‖ 2
u(t-Δt)用于表示平移Δt后无人驾驶设备确定出的控制量，a(t)用于表示无人驾驶设备实际执行的控制量。而该时延常数τ越大，则说明确定出的控制量变化频率越小，若是该时延常数τ越小，则说明确定出的控制量变化频率越大。
在本申请实施例中，无人驾驶设备可以根据上述状态数据，确定出针对下一时刻的无人驾驶设备的控制量基础误差，作为第一控制量基础误差，而后，若是无人驾驶设备确定出上述控制量变化频率大于第一设定频率，则通过该第一子模型，确定出在下一时刻针对无人驾驶设备的控制量误差偏量，作为第一偏量，并根据该第一偏量以及该第一控制量基础误差，确定在下一时刻针对无人驾驶设备自身的控制量实际误差。
而若是无人驾驶设备确定出上述控制量变化频率不大于该第一设定频率，则通过该第二子模型，确定在下一时刻针对无人驾驶设备自身的控制量误差偏量，作为第二偏量，并根据该第二偏量以及上述第一控制量基础误差，确定在下一时刻针对无人驾驶设备的控制量实际误差。
上述提到的第一控制量基础误差，是指无人驾驶设备将上述状态数据到控制量误差预测模型后确定出的在下一时刻的控制量上的基准误差，无人驾驶设备后续确定出的在下一时刻的控制量实际误差是在该第一控制量基础误差的基础上，通过结合第一偏量或是第二偏量所确定出的。而由于第一子模型与第二子模型对应无人驾驶设备所处的不同情况，所以，通过第一偏量确定出的实际控制量，可以保证无人驾驶设备处于诸如弯道等需要快速响应的情况时能够进行安全行驶，而通过第二偏量确定出的实际控制量，可以保证无人驾驶设备处于诸如直行等无需快速响应的情况时能够进行平稳行驶。
为了进一步地保证无人驾驶设备的安全行驶，在本申请实施例中，无人驾驶设备可以确定第一子模型在上一时刻输出的自身在当前时刻的控制量误差偏量，作为第一基准偏量，并确定出第二子模型在上一时刻输出的自身在当前时刻的控制量误差偏量，作为第二基准偏量，同时获取在上一时刻确定出的针对当前时刻的无人驾驶设备的控制量基础误差，作为第二控制量基础误差，进而将该第二控制量基础误差、第一基准偏量以及第二基准偏量输入到第一子模型，得到上述第一偏量。
也就是说，假设当前时刻为t，无人驾驶设备在确定上述第一偏量时，需要参考第一子模型在t-1时针对t时刻所确定出的控制量误差偏量(即第一基准偏量)，第二子模型在t-1时针对无人驾驶设备在t时刻所确定出的控制量误差偏量(即第二基准偏量)，以及无人驾驶设备在t-1时刻针对t时刻的无人驾驶设备所确定出的控制量基础误差(第二控制量基础误差)。
例如，无人驾驶设备在确定上述第一偏量时，可以采用下述公式来进行确定：
Figure PCTCN2021144012-appb-000001
Figure PCTCN2021144012-appb-000002
即为确定出的第一偏量，
Figure PCTCN2021144012-appb-000003
为通过第一子模型在上一时刻针对无人驾驶设备的当前时刻所确定出的控制量误差偏量，即第一基准偏量，
Figure PCTCN2021144012-appb-000004
为通过第二子模型在上一时刻针对无人驾驶设备的当前时刻所确定出的控制量误差偏量，即第二基准偏量，
Figure PCTCN2021144012-appb-000005
即为无人驾驶设备在上一时刻针对当前时刻的无人驾驶设备所确定出的控制量基础误差，即第二控制量基础误差。σ f为第一子模型对应的时间参数，θ f为第一子模型内的模型参数。
在确定上述第二偏量时，无人驾驶设备也可以先确定出第一子模型在上一时刻输出的自 身在当前时刻的控制量误差偏量，即第一基准偏量，以及确定第二子模型在上一时刻输出的无人驾驶设备在当前时刻的控制量误差偏量，即第二基准偏量。而后，无人驾驶设备可以将该第一基准偏量以及第二基准偏量输入到第二子模型中，得到第二偏量。
也就是说，假设当前时刻为t，无人驾驶设备在确定上述第二偏量时，需要参考第一子模型在t-1时针对无人驾驶设备在t时刻所确定出的控制量误差偏量(即第一基准偏量)，以及第二子模型在t-1时针对无人驾驶设备在t时刻所确定出的控制量误差偏量(即第二基准偏量)。
例如，无人驾驶设备在确定上述第二偏量时，可以采用下述公式来进行确定：
Figure PCTCN2021144012-appb-000006
Figure PCTCN2021144012-appb-000007
即为确定出的第二偏量，
Figure PCTCN2021144012-appb-000008
通过第二子模型在上一时刻针对无人驾驶设备的当前时刻所确定出的控制量误差偏量，即第二基准偏量，
Figure PCTCN2021144012-appb-000009
为通过第一子模型在上一时刻针对无人驾驶设备的当前时刻所确定出的控制量误差偏量，即第一基准偏量，σ s为第二子模型对应的时间参数，θ s为第二子模型中的模型参数。
需要说明的是，在确定上述第一偏量时需要参考第二子模型在上一时刻确定出的数据，是因为如果单单指考虑第一子模型自身在上一时刻确定出的数据，可能会无人驾驶设备过度响应的情况，如，无人驾驶设备按照确定出的实际控制量对自身进行控制时，在弯道可能会出现转向过度的情况。而为了避免这种情况的发生，无人驾驶设备可以将第二子模型在上一时刻确定出的数据作为输入，输入到第一子模型中，以通过引入第二子模型在上一时刻的输出，降低第一子模型的输出会引起无人驾驶设备过度响应的情况的发生。
在确定上述第二偏量时需要引入第一子模型在上一时刻确定出的数据，则是因为，如果单单指考虑第二子模型自身在上一时刻确定出的数据，可能会无人驾驶设备响应过慢的情况，如，无人驾驶设备按照确定出的实际控制量对自身进行控制时，在直道上可能会出现过度直行的情况。而为了避免这种情况的发生，无人驾驶设备可以将第一子模型在上一时刻确定出的数据作为输入，输入到第二子模型中，以通过引入第一子模型在上一时刻的输出，降低第二子模型的输出会引起无人驾驶设备响应过慢的情况的发生。
进一步地，在上述确定第一偏量的公式中，
Figure PCTCN2021144012-appb-000010
这一项主要用于控制第一子模型输出的第一偏量与上一时刻第一子模型输出的偏量(即第一基准偏量)之间的变化幅度。示例性的，
Figure PCTCN2021144012-appb-000011
这一项中的
Figure PCTCN2021144012-appb-000012
用于表示第一基准偏量，而当σ f取1时，这一项为0，说明第一子模型在确定第一偏量时完全不考虑第一子模型上一时刻输出的第一基准偏量，那么最终确定出的第一偏量相较于第一基准偏量来说，变化幅度可能较大。而如果当σ f取0～1之间的数时，说明第一子模型在确定第一偏量时会考虑第一子模型上一时刻输出的第一基准偏量，而当σ f逐渐趋近0时，说明第一子模型在确定第一偏量时考虑第一基准偏量的比重也越来越大，最终导致确定出的第一偏量相较于第一基准偏量来说，变化幅度也越来越低。
同理，在上述确定第二偏量的公式中，
Figure PCTCN2021144012-appb-000013
这一项主要用于控制第二子模型输出的第二偏量与上一时刻第二子模型输出的偏量(即第二基准偏量)之间的变化幅度。具体原理与上述基本相同，在此就不进行详细说明了。
需要说明的是，在本申请实施例中，σ s和σ f需要满足：1＞σ s＞σ f＞0这一关系。具体的数值可以根据实际需求而定，之所以要保证σ s＞σ f，是因为，第一子模型主要应对的是无人驾驶设备需要快速响应的情况，而第二子模型主要应对的是无人驾驶设备无需快速响应的情况，那么要求σ s＞σ f满足这一关系，可以保证在同一时刻，第一子模型确定出的第一偏量与第一子模型上一时刻确定出的第一基准偏量之间的变化幅度，要大于第二子模型确定出的第二偏量与第二子模型上一时刻确定出的第二基准偏量之间的变化幅度，以进一步地满足第一子模型和第二子模型各自的适用情况。
还需指出的是，虽然无人驾驶设备在一个具体的时刻可以选用第一子模型和第二子模型中的一个子模型确定实际控制量，但是，第一子模型和第二子模型每个时刻都需要确定出各自的第一偏量以及第二偏量，以使第一子模型和第二子模型能够在下一时刻为彼此提供参考数据。
在本申请实施例中，可以事先对上述控制量误差预测模型、第一子模型以及第二子模型进行训练。过程如图2所示。
图2为本申请实施例提供的对控制量误差预测模型、第一子模型以及第二子模型进行训练的示意图。
在获取到训练样本后，可以将该训练样本中包含的状态数据输入到控制量误差预测模型中，例如，状态数据为图2中的Xt。而由于控制量误差预测模型中包含有第一子模型和第二子模型，所以，在对控制量误差预测模型进行训练的过程中，还需要对第一子模型和第二子模型进行一并训练。
其中，在模型训练的过程中，控制量误差预测模型的输入/输出层为多层感知机结构，可以接收来自第一子模型或是第二子模型的输出，而第一子模型则可以接收来自上一时刻控制量误差预测模型的输入/输出层的输出，以及第二子模型的输出作为输入。第二子模型可以只接收第一子模型的输出。
由于可以在训练样本中获取到t+τ时刻无人驾驶设备的真实控制量e t+τ，因此可以通过该e t+τ与控制量误差预测模型输出的针对t+τ时刻无人驾驶设备的控制量y t+τ之间的偏差进行优化，以对控制量误差预测模型、第一子模型以及第二子模型进行训练。其中，基于e t+τ与y t+τ之间的偏差进行优化的方式可以有多种，例如，将e t+τ与y t+τ视为两个概率分布，并通过KL散度确定这两者之间的损失函数，对上述三个模型进行训练。其他方式在此就不详细举说明了。
示例性地，通过KL散度确定e t+τ和y t+τ的损失函数为J t(t,θ)：
Figure PCTCN2021144012-appb-000014
其中，IO代表输出层，该e t+τ与控制量误差预测模型的各个输出层输出的针对t+τ时刻无人驾驶设备的控制量y t+τ之间的偏差进行优化，以对控制量误差预测模型、第一子模型以及第二子模型进行训练。
需要说明的是，上述图2所示的训练过程仅以将训练样本中包含的状态数据Xt输入到控制量误差预测模型中为例进行说明，在一种可能的实现方式中，还可以对训练样本中包含的状态数据Xt进行降维处理之后再输入到控制量误差预测模型中。通过对输入到控制量误差预测模型中的特征进行降维处理，控制整个模型的大小。本申请实施例不对将状态数据进行降维处理的方式进行限定，示例性地，采用自组织映射(Self-organizing map,SOM)的方式对状态数据进行降维。例如，训练过程参见图5，图5所示的训练过程与图2相比，多了SOM的过程，状态数据Xt经过SOM之后，得到Xt’，将Xt’输入到控制量误差预测模型中。图5中与图2相对应的过程可参见图2的相关描述，此处不再赘述。
为了进一步保证无人驾驶设备的平稳行驶，在本申请实施例中，若确定上述控制量变化频率大于第二设定频率，则可以通过第一调整方式对该控制量实际误差进行调整，得到第一调整后控制量误差，并根据该第一调整后控制量误差以及上述基准控制量误差，确定在下一时刻针对无人驾驶设备的实际控制量。
若是确定出的控制量变化频率不大于第二设定频率，则可以通过第二调整方式对控制量实际误差进行，得到第二调整后控制量误差，并根据第二调整后控制量误差以及上述基准控制量，确定在下一时刻针对无人驾驶设备自身的实际控制量。这里提到的第二设定频率与上述第一设定频率可以是相同的，也可以是不同的，具体可以根据实际需求进行设定。
其中，基于第一调整方式确定出的针对下一时刻的实际控制量与在上一时刻确定出的针 对当前时刻的实际控制量之间的控制量变化量，大于基于第二调整方式确定出的针对下一时刻的实际控制量与在上一时刻确定出的针对当前时刻的实际控制量之间的控制量变化量。
这是因为，在实际应用中，控制量变化频率若是较高，则说明无人驾驶设备可能处于如弯道等情况下，需要达到快速响应的目的，所以，需要通过第一调整方式对确定出的控制量误差进行调整，以保证无人驾驶设备能够通过第一调整后控制量误差所确定出的针对下一时刻的实际控制量，实现无人驾驶设备的快速响应。而若是控制量变化频率较低，则说明无人驾驶设备可能处于如直行等情况下，无需进行快速响应，所以，需要通过第二调整方式对确定出的控制量误差进行调整，以保证无人驾驶设备能够通过第二调整后控制量误差所确定出的针对下一时刻的实际控制量，实现无人驾驶设备的平稳行驶。
在本申请实施例中，无人驾驶设备可以通过下述公式确定第一调整后控制量误差以及第二调整后控制量误差：
Figure PCTCN2021144012-appb-000015
在该公式中，
Figure PCTCN2021144012-appb-000016
用于表示确定出针对下一时刻的控制量实际误差，
Figure PCTCN2021144012-appb-000017
用于表示第一调整方式，
Figure PCTCN2021144012-appb-000018
用于表示第二调整方式。其中，k p为比例系数，k d为微分时间常数，k i为积分时间常数，这里判断采用哪种调整方式，主要是通过无人驾驶设备的角度变化γ为依据，所以，w 0用于表示设定角度变化。这种方式与上述提到的通过控制量变化频率来确定采用哪种调整方式在原理上基本是相同的。即，如果角度变化较大，则无人驾驶设备很可能处于需要快速响应的情况，如果角度变化较小，则无人驾驶设备很可能处于无需快速响应，平稳行驶的情况。
需要说明的是，上述第一调整方式和第二调整方式实际上可以看作是微分(PD)控制器和比例积分(PI)控制器。而因为第二调整方式为比例积分(PI)控制器，所以通过在时间上进行积分得到的第二调整后控制量误差，可以保证无人驾驶设备确定出的实际控制量与上一时刻的实际控制量相比，在变化上更为平滑，从而保证了无人驾驶设备行驶更加平稳，而因为第一调整方式为微分(PD)控制器，所以确定出的第一调整后控制量误差能够突出具体一个时刻的控制量误差变化，从而保证了无人驾驶设备处在诸如弯道等需要快速响应的情况时，能够实现有效地快速响应。
进一步地，在上述控制量误差预测模型采用多时间尺度的循环神经网络(MTRNN，Multiple timescale recurrent neural network)的情况下，将其与PID(Proportion Integral Differential，比例积分微分)控制算法结合来预测下一时刻的控制量误差，可以有效地提高无人驾驶设备的进行决策的效率。
示例性的，如果直接采用一个整体网络，获取到无人驾驶设备的残差控制量，那就需要构建一个最优控制量的监督信号，而这种监督信号在无人驾驶设备的实际控制中往往是无法获取，因此是使用该整体网络对无人驾驶设备进行控制之前，需要使用获取到的仿真数据对这一整体网络进行训练，而这样导致训练后的该整体网络往往难以用到无人驾驶设备的实际控制中。
S104：根据所述控制量实际误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量。
在确定出上述控制量实际误差后，可以通过该控制量实际误差对上述基准控制量进行修正，从而确定出在下一时刻针对无人驾驶设备的实际控制量。其中，无人驾驶设备所确定出的基准控制量实际上是一个控制量序列，即，从下一时刻开始，连续若干个时刻所对应的控制量，而由于上述确定出的控制量实际误差是针对下一时刻的控制量实际误差，所以，无人驾驶设备可以通过该控制量实际误差，将该控制量序列中下一时刻所对应的控制量进行修正， 得到下一时刻针对自身的实际控制量，参考如下公式：
U(k)＝[u(t),u(t+1),…,u(t+k)]
在该公式中，U(k)可以看作是上述基准控制量，从该公式中可以看出，基准控制量中包含有多个控制量，即，从下一时刻开始，连续k个时刻的控制量，u(t)即为下一时刻对应的控制量。
基于此，无人驾驶设备可以通过以下公式，确定出针对下一时刻的实际控制量：
u′(t)＝u 1(t)+u 2(t)
其中，u′(t)即为针对下一时刻的实际控制量，u 2(t)即为上面提到的基准控制量中包含的针对下一时刻的控制量，u 1(t)即为无人驾驶设备确定出的针对下一时刻的控制量实际误差。通过该公式，可以最终确定出无人驾驶设备针对下一时刻的实际控制量。
在本申请实施例中，上述基准控制量可以采用求解非线性优化问题的方式来进行确定，由于求解非线性优化问题所采用的方式有多种，所以，本申请实施例不对基准控制量的具体确定方式进行限定。
S105：根据所述实际控制量，对所述无人驾驶设备进行控制。
在确定出实际控制量后，无人驾驶设备可以通过该实际控制量对自身进行控制，而由于该实际控制量是通过修正后的控制量，所以无人驾驶设备采用该实际控制量对自身进行控制，可以有效地保证无人驾驶设备的平稳行驶，保证了无人驾驶设备在行驶过程中的安全性。
以上为本申请的一个或多个实施例提供的无人驾驶设备的控制方法，基于同样的思路，本申请实施例还提供了相应的无人驾驶设备的控制装置，如图3所示。图3为本申请实施例提供的一种无人驾驶设备的控制装置示意图，包括：
获取模块301，用于获取无人驾驶设备在当前时刻的状态数据；
预测模块302，用于根据所述状态数据，预测在下一时刻针对所述无人驾驶设备的控制量，作为基准控制量；
输入模块303，用于将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差；
确定模块304，用于根据所述控制量实际误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
控制模块305，用于根据所述实际控制量，对所述无人驾驶设备进行控制。
可选地，所述输入模块303用于，确定所述控制量误差预测模型在上一时刻针对所述当前时刻所确定出的控制量误差，作为参照控制量误差；将所述状态数据以及所述参照控制量误差输入到所述控制量误差预测模型中，得到针对所述下一时刻的控制量误差。
可选地，所述控制量误差预测模型中包含有第一子模型和第二子模型；
所述输入模块303用于，根据所述状态数据，确定针对所述下一时刻的所述无人驾驶设备的控制量基础误差，作为第一控制量基础误差；根据所述无人驾驶设备在所述当前时刻之前设定时间段的历史状态数据以及历史控制量，确定所述无人驾驶设备在所述当前时刻对应的控制量变化频率；若确定所述控制量变化频率大于第一设定频率，通过所述第一子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第一偏量，并根据所述第一偏量以及所述第一控制量基础误差，确定在所述下一时刻针对所述无人驾驶设备的控制量实际误差；若确定所述控制量变化频率不大于所述第一设定频率，通过所述第二子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第二偏量，并根据所述第二偏量以及所述第一控制量基础误差，确定在所述下一时刻针对所述无人驾驶设备的控制量实际误差。
可选地，所述输入模块303用于，针对所述设定时间段中包含的每个历史时刻，根据该历史时刻对应的历史控制量以及该历史时刻的下一历史时刻对应的历史状态数据，确定该历史时刻对应的历史控制量偏差；根据每个历史时刻对应的历史控制量偏差，确定所述无人驾驶设备在所述当前时刻对应的控制量变化频率。
可选地，所述输入模块303用于，确定所述第一子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第一基准偏量，确定所述第二子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第二基准偏量，以及获取在上一时刻确定出的针对所述当前时刻的所述无人驾驶设备的控制量基础误差，作为第二控制量基础误差；将所述第二控制量基础误差、所述第一基准偏量以及所述第二基准偏量输入到所述第一子模型中，得到所述第一偏量。
可选地，所述输入模块303用于，确定所述第一子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第一基准偏量，确定所述第二子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第二基准偏量；将所述第一基准偏量以及所述第二基准偏量输入到所述第二子模型中，得到所述第二偏量。
可选地，所述确定模块304用于，若确定所述控制量变化频率大于第二设定频率，通过第一调整方式对所述控制量实际误差进行调整，得到第一调整后控制量误差，并根据所述第一调整后控制量误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；若确定所述控制量变化频率不大于第二设定频率，通过第二调整方式对所述控制量实际误差进行调整，得到第二调整后控制量误差，并根据所述第二调整后控制量误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
其中，基于所述第一调整方式确定出的在所述下一时刻针对所述无人驾驶设备的实际控制量与在所述上一时刻确定出的在所述当前时刻针对所述无人驾驶设备的实际控制量之间的控制量变化量，大于基于所述第二调整方式确定出的在所述下一时刻针对所述无人驾驶设备的实际控制量与在所述上一时刻确定出的在所述当前时刻针对所述无人驾驶设备的实际控制量之间的控制量变化量。
本申请实施例还提供了一种非临时性计算机可读存储介质，该非临时性计算机可读存储介质存储有计算机程序，计算机程序可用于执行上述图1提供的无人驾驶设备的控制方法。
本申请实施例还提供了图4所示的无人驾驶设备的示意结构图。如图4所述，在硬件层面，该无人驾驶设备包括处理器、内部总线、网络接口、内存以及非易失性存储器，当然还可能包括其他业务所需要的硬件。处理器从非易失性存储器中读取对应的计算机程序到内存中然后运行，以实现上述图1所述的无人驾驶设备的控制方法。当然，除了软件实现方式之外，本申请实施例并不排除其他实现方式，比如逻辑器件抑或软硬件结合的方式等等，也就是说以下处理流程的执行主体并不限定于各个逻辑单元，也可以是硬件或逻辑器件。
在20世纪90年代，对于一个技术的改进可以很明显地区分是硬件上的改进(例如，对二极管、晶体管、开关等电路结构的改进)还是软件上的改进(对于方法流程的改进)。然而，随着技术的发展，当今的很多方法流程的改进已经可以视为硬件电路结构的直接改进。设计人员几乎都通过将改进的方法流程编程到硬件电路中来得到相应的硬件电路结构。因此，不能说一个方法流程的改进就不能用硬件实体模块来实现。例如，可编程逻辑器件(Programmable Logic Device,PLD)(例如现场可编程门阵列(Field Programmable Gate Array，FPGA))就是这样一种集成电路，其逻辑功能由用户对器件编程来确定。由设计人员自行编程来把一个数字系统“集成”在一片PLD上，而不需要请芯片制造厂商来设计和制作专用的集成电路芯片。而且，如今，取代手工地制作集成电路芯片，这种编程也多半改用“逻辑编译器(logic compiler)”软件来实现，它与程序开发撰写时所用的软件编译器相类似，而要编译之前的原始代码也得用特定的编程语言来撰写，此称之为硬件描述语言(Hardware Description Language，HDL)，而HDL也并非仅有一种，而是有许多种，如ABEL(Advanced Boolean Expression Language)、AHDL(Altera Hardware Description Language)、Confluence、CUPL(Cornell University Programming Language)、HDCal、JHDL(Java Hardware Description Language)、Lava、Lola、MyHDL、PALASM、RHDL(Ruby Hardware Description Language)等，目前最普遍使用的是VHDL(Very-High-Speed Integrated Circuit Hardware Description  Language)与Verilog。本领域技术人员也应该清楚，只需要将方法流程用上述几种硬件描述语言稍作逻辑编程并编程到集成电路中，就可以很容易得到实现该逻辑方法流程的硬件电路。
控制器可以按任何适当的方式实现，例如，控制器可以采取例如微处理器或处理器以及存储可由该(微)处理器执行的计算机可读程序代码(例如软件或固件)的计算机可读介质、逻辑门、开关、专用集成电路(Application Specific Integrated Circuit，ASIC)、可编程逻辑控制器和嵌入微控制器的形式，控制器的例子包括但不限于以下微控制器：ARC 625D、Atmel AT91SAM、Microchip PIC18F26K20以及Silicone Labs C8051F320，存储器控制器还可以被实现为存储器的控制逻辑的一部分。本领域技术人员也知道，除了以纯计算机可读程序代码方式实现控制器以外，完全可以通过将方法步骤进行逻辑编程来使得控制器以逻辑门、开关、专用集成电路、可编程逻辑控制器和嵌入微控制器等的形式来实现相同功能。因此这种控制器可以被认为是一种硬件部件，而对其内包括的用于实现各种功能的装置也可以视为硬件部件内的结构。或者甚至，可以将用于实现各种功能的装置视为既可以是实现方法的软件模块又可以是硬件部件内的结构。
上述实施例阐明的系统、装置、模块或单元，可以由计算机芯片或实体实现，或者由具有某种功能的产品来实现。一种典型的实现设备为计算机。示例性的，计算机例如可以为个人计算机、膝上型计算机、蜂窝电话、相机电话、智能电话、个人数字助理、媒体播放器、导航设备、电子邮件设备、游戏控制台、平板计算机、可穿戴设备或者这些设备中的任何设备的组合。
为了描述的方便，描述以上装置时以功能分为各种单元分别描述。当然，在实施本申请实施例时可以把各单元的功能在同一个或多个软件和/或硬件中实现。
本领域内的技术人员应明白，本申请的实施例可提供为方法、系统、或计算机程序产品。因此，本申请实施例可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本申请实施例可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本申请实施例是参照根据本申请实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
在一个典型的配置中，计算设备包括一个或多个处理器(CPU)、输入/输出接口、网络接口和内存。
内存可能包括计算机可读介质中的非永久性存储器，随机存取存储器(RAM)和/或非易失性内存等形式，如只读存储器(ROM)或闪存(flash RAM)。内存是计算机可读介质的示例。
计算机可读介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括，但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随 机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带，磁带磁磁盘存储或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。按照本文中的界定，计算机可读介质不包括暂存电脑可读媒体(transitory media)，如调制的数据信号和载波。
还需要说明的是，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、商品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、商品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、商品或者设备中还存在另外的相同要素。
本领域技术人员应明白，本申请的实施例可提供为方法、系统或计算机程序产品。因此，本申请实施例可采用完全硬件实施例、完全软件实施例或结合软件和硬件方面的实施例的形式。而且，本申请实施例可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本申请实施例可以在由计算机执行的计算机可执行指令的一般上下文中描述，例如程序模块。一般地，程序模块包括执行特定任务或实现特定抽象数据类型的例程、程序、对象、组件、数据结构等等。也可以在分布式计算环境中实践本申请实施例，在这些分布式计算环境中，由通过通信网络而被连接的远程处理设备来执行任务。在分布式计算环境中，程序模块可以位于包括存储设备在内的本地和远程计算机存储介质中。
本申请中的各个实施例均采用递进的方式描述，各个实施例之间相同相似的部分互相参见即可，每个实施例重点说明的都是与其他实施例的不同之处。尤其，对于系统实施例而言，由于其基本相似于方法实施例，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
以上所述仅为本申请的实施例而已，并不用于限制本申请。对于本领域技术人员来说，本申请可以有各种更改和变化。凡在本申请的原理之内所作的任何修改、等同替换、改进等，均应包含在本申请的权利要求范围之内。

Claims (11)

  1. 一种无人驾驶设备的控制方法，其中，包括：
    获取无人驾驶设备在当前时刻的状态数据；
    根据所述状态数据，预测在下一时刻针对所述无人驾驶设备的控制量，作为基准控制量；
    将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差；
    根据所述控制量实际误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
    根据所述实际控制量，对所述无人驾驶设备进行控制。
  2. 如权利要求1所述的方法，其中，所述将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差，包括：
    确定所述控制量误差预测模型在上一时刻针对所述当前时刻所确定出的控制量误差，作为参照控制量误差；
    将所述状态数据以及所述参照控制量误差输入到所述控制量误差预测模型中，得到针对所述下一时刻的控制量误差。
  3. 如权利要求1所述的方法，其中，所述控制量误差预测模型中包含有第一子模型和第二子模型；
    所述将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差，包括：
    根据所述状态数据，确定针对所述下一时刻的所述无人驾驶设备的控制量基础误差，作为第一控制量基础误差；
    根据所述无人驾驶设备在所述当前时刻之前设定时间段的历史状态数据以及历史控制量，确定所述无人驾驶设备在所述当前时刻对应的控制量变化频率；
    若确定所述控制量变化频率大于第一设定频率，通过所述第一子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第一偏量，并根据所述第一偏量以及所述第一控制量基础误差，确定在所述下一时刻针对所述无人驾驶设备的控制量实际误差；
    若确定所述控制量变化频率不大于所述第一设定频率，通过所述第二子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第二偏量，并根据所述第二偏量以及所述第一控制量基础误差，确定在所述下一时刻针对所述无人驾驶设备的控制量实际误差。
  4. 如权利要求3所述的方法，其中，所述根据所述无人驾驶设备在所述当前时刻之前设 定时间段的历史状态数据以及历史控制量，确定所述无人驾驶设备在所述当前时刻对应的控制量变化频率，包括：
    针对所述设定时间段中包含的每个历史时刻，根据该历史时刻对应的历史控制量以及该历史时刻的下一历史时刻对应的历史状态数据，确定该历史时刻对应的历史控制量偏差；
    根据每个历史时刻对应的历史控制量偏差，确定所述无人驾驶设备在所述当前时刻对应的控制量变化频率。
  5. 如权利要求3所述的方法，其中，所述通过所述第一子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第一偏量，包括：
    确定所述第一子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第一基准偏量，确定所述第二子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第二基准偏量，以及获取在上一时刻确定出的针对所述当前时刻的所述无人驾驶设备的控制量基础误差，作为第二控制量基础误差；
    将所述第二控制量基础误差、所述第一基准偏量以及所述第二基准偏量输入到所述第一子模型中，得到所述第一偏量。
  6. 如权利要求3所述的方法，其中，所述通过所述第二子模型，确定在所述下一时刻针对所述无人驾驶设备的控制量误差偏量，作为第二偏量，包括：
    确定所述第一子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第一基准偏量，确定所述第二子模型在上一时刻输出的所述无人驾驶设备在所述当前时刻的控制量误差偏量，作为第二基准偏量；
    将所述第一基准偏量以及所述第二基准偏量输入到所述第二子模型中，得到所述第二偏量。
  7. 如权利要求3所述的方法，其中，所述根据所述控制量实际误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量，包括：
    若确定所述控制量变化频率大于第二设定频率，通过第一调整方式对所述控制量实际误差进行调整，得到第一调整后控制量误差，并根据所述第一调整后控制量误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
    若确定所述控制量变化频率不大于第二设定频率，通过第二调整方式对所述控制量实际误差进行调整，得到第二调整后控制量误差，并根据所述第二调整后控制量误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
    其中，基于所述第一调整方式确定出的在所述下一时刻针对所述无人驾驶设备的实际控制量与在所述上一时刻确定出的在所述当前时刻针对所述无人驾驶设备的实际控制量之间的 控制量变化量，大于基于所述第二调整方式确定出的在所述下一时刻针对所述无人驾驶设备的实际控制量与在所述上一时刻确定出的在所述当前时刻针对所述无人驾驶设备的实际控制量之间的控制量变化量。
  8. 一种无人驾驶设备的控制装置，其中，包括：
    获取模块，用于获取无人驾驶设备在当前时刻的状态数据；
    预测模块，用于根据所述状态数据，预测在下一时刻针对所述无人驾驶设备的控制量，作为基准控制量；
    输入模块，用于将所述状态数据输入到控制量误差预测模型中，得到在所述下一时刻针对所述无人驾驶设备的控制量实际误差；
    确定模块，用于根据所述控制量实际误差以及所述基准控制量，确定在所述下一时刻针对所述无人驾驶设备的实际控制量；
    控制模块，用于根据所述实际控制量，对所述无人驾驶设备进行控制。
  9. 一种非临时性计算机可读存储介质，其中，所述存储介质存储有计算机程序，所述计算机程序被处理器执行时实现上述权利要求1～7任一项所述的方法。
  10. 一种无人驾驶设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其中，所述处理器执行所述程序时实现上述权利要求1～7任一项所述的方法。
  11. 一种计算机程序产品，所述计算机程序产品包括计算机指令，所述计算机指令存储在非临时性计算机可读存储介质中，计算机设备的处理器从所述非临时性计算机可读存储介质读取所述计算机指令，所述处理器执行所述计算机指令，使得所述计算机设备实现如权利要求1～7任一项所述的方法。
PCT/CN2021/144012 2021-01-04 2021-12-31 无人驾驶设备的控制 WO2022144010A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110000641.7 2021-01-04
CN202110000641.7A CN112346467B (zh) 2021-01-04 2021-01-04 一种无人驾驶设备的控制方法及装置

Publications (1)

Publication Number Publication Date
WO2022144010A1 true WO2022144010A1 (zh) 2022-07-07

Family

ID=74427559

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/144012 WO2022144010A1 (zh) 2021-01-04 2021-12-31 无人驾驶设备的控制

Country Status (2)

Country Link
CN (1) CN112346467B (zh)
WO (1) WO2022144010A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116125796A (zh) * 2023-02-17 2023-05-16 之江实验室 一种设备控制方法、装置、存储介质及电子设备

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112346467B (zh) * 2021-01-04 2021-04-27 北京三快在线科技有限公司 一种无人驾驶设备的控制方法及装置
CN113048989B (zh) * 2021-04-06 2022-12-09 北京三快在线科技有限公司 一种无人驾驶设备的定位方法及定位装置
CN113643321A (zh) * 2021-07-30 2021-11-12 北京三快在线科技有限公司 一种无人驾驶设备的传感器数据采集方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140336787A1 (en) * 2013-05-10 2014-11-13 Honeywell International Inc. Index generation and embedded fusion for controller performance monitoring
CN108345270A (zh) * 2017-01-24 2018-07-31 欧姆龙株式会社 控制装置、控制程序及控制系统
CN110941272A (zh) * 2019-11-27 2020-03-31 深圳先进技术研究院 自动驾驶控制方法和设备
CN111152796A (zh) * 2020-04-07 2020-05-15 北京三快在线科技有限公司 一种车辆运动状态预测方法及装置
CN111208838A (zh) * 2020-04-20 2020-05-29 北京三快在线科技有限公司 一种无人驾驶设备的控制方法及装置
CN111930015A (zh) * 2020-09-16 2020-11-13 北京三快在线科技有限公司 一种无人车控制方法及装置
CN112346467A (zh) * 2021-01-04 2021-02-09 北京三快在线科技有限公司 一种无人驾驶设备的控制方法及装置

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140336787A1 (en) * 2013-05-10 2014-11-13 Honeywell International Inc. Index generation and embedded fusion for controller performance monitoring
CN108345270A (zh) * 2017-01-24 2018-07-31 欧姆龙株式会社 控制装置、控制程序及控制系统
CN110941272A (zh) * 2019-11-27 2020-03-31 深圳先进技术研究院 自动驾驶控制方法和设备
CN111152796A (zh) * 2020-04-07 2020-05-15 北京三快在线科技有限公司 一种车辆运动状态预测方法及装置
CN111208838A (zh) * 2020-04-20 2020-05-29 北京三快在线科技有限公司 一种无人驾驶设备的控制方法及装置
CN111930015A (zh) * 2020-09-16 2020-11-13 北京三快在线科技有限公司 一种无人车控制方法及装置
CN112346467A (zh) * 2021-01-04 2021-02-09 北京三快在线科技有限公司 一种无人驾驶设备的控制方法及装置

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116125796A (zh) * 2023-02-17 2023-05-16 之江实验室 一种设备控制方法、装置、存储介质及电子设备

Also Published As

Publication number Publication date
CN112346467B (zh) 2021-04-27
CN112346467A (zh) 2021-02-09

Similar Documents

Publication Publication Date Title
WO2022144010A1 (zh) 无人驾驶设备的控制
US20200324794A1 (en) Technology to apply driving norms for automated vehicle behavior prediction
WO2019242718A1 (zh) 一种车辆变道控制方法和装置
US20210001884A1 (en) Technology to generalize safe driving experiences for automated vehicle behavior prediction
WO2021190484A1 (zh) 一种障碍物的轨迹预测方法及装置
US20220324483A1 (en) Trajectory prediction method and apparatus, storage medium, and electronic device
WO2021052185A1 (zh) 确定智能驾驶车辆的行驶轨迹
JP6668375B2 (ja) 比例、積分及び微分（ｐｉｄ）コントローラを用いた自律走行車のステアリング制御方法及びシステム
CN111930015B (zh) 一种无人车控制方法及装置
US10809735B2 (en) System and method for a framework of robust and safe reinforcement learning application in real world autonomous vehicle application
JP2020083308A (ja) オブジェクト挙動のリアルタイム予測
CN111208838B (zh) 一种无人驾驶设备的控制方法及装置
CN110262486B (zh) 一种无人驾驶设备运动控制方法及装置
CN111062372B (zh) 一种预测障碍物轨迹的方法及装置
CN112896191B (zh) 轨迹处理方法、装置、电子设备和计算机可读介质
CN112180925B (zh) Agv轨迹跟随方法、装置及处理设备
US20220055651A1 (en) Data-driven warm start selection for optimization-based trajectory planning
JP2019501809A (ja) 自律走行車の速度を追従する方法及びシステム
CN113044042B (zh) 车辆预测变道图像显示方法、装置、电子设备和可读介质
CN117390585B (zh) 基于三维全连接融合的时序数据预测方法及模型训练方法
CN111522245B (zh) 用于控制无人设备的方法及装置
US20220314980A1 (en) Obstacle tracking method, storage medium and unmanned driving device
Kim et al. Vision-based uncertainty-aware lane keeping strategy using deep reinforcement learning
US20230391352A1 (en) Managing an integral part of a controller of an automated vehicle
WO2023185215A1 (zh) 数据校准

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: 21914738

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21914738

Country of ref document: EP

Kind code of ref document: A1