WO2021081816A1 - 一种数据处理方法、装置和可移动平台 - Google Patents

一种数据处理方法、装置和可移动平台 Download PDF

Info

Publication number
WO2021081816A1
WO2021081816A1 PCT/CN2019/114391 CN2019114391W WO2021081816A1 WO 2021081816 A1 WO2021081816 A1 WO 2021081816A1 CN 2019114391 W CN2019114391 W CN 2019114391W WO 2021081816 A1 WO2021081816 A1 WO 2021081816A1
Authority
WO
WIPO (PCT)
Prior art keywords
estimate
observation
state
current moment
estimates
Prior art date
Application number
PCT/CN2019/114391
Other languages
English (en)
French (fr)
Inventor
吴显亮
余瑞
陈进
Original Assignee
深圳市大疆创新科技有限公司
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 深圳市大疆创新科技有限公司 filed Critical 深圳市大疆创新科技有限公司
Priority to CN201980031650.3A priority Critical patent/CN112119413A/zh
Priority to PCT/CN2019/114391 priority patent/WO2021081816A1/zh
Publication of WO2021081816A1 publication Critical patent/WO2021081816A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • 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
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
    • 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
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/048—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators using a predictor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/02—Services making use of location information
    • H04W4/021—Services related to particular areas, e.g. point of interest [POI] services, venue services or geofences
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/30—Services specially adapted for particular environments, situations or purposes
    • H04W4/40—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P]

Definitions

  • the embodiments of the present invention relate to the field of data processing, and more specifically, to a data processing method, device, and movable platform.
  • movable platforms for example, unmanned vehicles or unmanned aerial vehicles
  • state of motion for example, position, speed, etc.
  • objects for example, pedestrians, vehicles, etc.
  • Mobile platform for control.
  • the estimation of the prediction of the motion state of the target object at the current moment is determined according to the estimation of the motion state of the target object at the previous moment. Then, obtain the observations of the sensor output configured on the movable platform for detecting the object.
  • the observations may include multiple (there may be multiple objects around the movable platform), and further, use the Hungarian distribution algorithm to divide the multiple One of the observations is allocated to the target object, and the predicted estimate is updated by using the allocated observation to obtain an estimate of the motion state of the target object at the current moment.
  • the embodiments of the present invention provide a data processing method, device, and movable platform, which can improve the robustness when determining the estimation of the motion state of an object.
  • a data processing method which includes: obtaining multiple estimates of the motion state of a target object at a previous moment, wherein the object includes a target object; according to the motion state of the target object at the previous moment To obtain multiple predicted estimates of the motion state of the target object at the current moment; obtain one or more observations output by the sensor; determine the relationship between each predicted estimate and each The matching state between observations, wherein the matching state includes a compatible state and an incompatible state; the corresponding predicted estimate is updated according to the observation that the matching state is a compatible state; the updated estimate and The estimation of the prediction that the matching state is an incompatible state is determined as multiple estimates of the motion state at the current moment.
  • a data processing device configured on a movable platform, and the movable platform includes a sensor for outputting observations of an object around the movable platform, including: an acquisition module, It is used to obtain multiple estimates of the motion state of the target object at the previous moment, wherein the object includes the target object; the processing module is used to predict according to the multiple estimates of the motion state of the target object at the previous moment to obtain The estimation of multiple predictions of the motion state of the target object at the current moment; obtaining one or more observations output by the sensor; the processing module is also used to determine the estimation of each prediction and the estimation of each The matching state between observations, wherein the matching state includes a compatible state and an incompatible state; the processing module is further configured to update the corresponding prediction estimate according to the observation that the matching state is a compatible state; The processing module is further configured to determine the updated estimate and the predicted estimate that the matching state is an incompatible state as multiple estimates of the motion state at the current moment.
  • a data processing device configured on a movable platform, and the movable platform includes a sensor for outputting observations of an object around the movable platform, including: memory and processing Wherein the memory is used to store program instructions; the processor is configured to execute the program instructions, and when the program instructions are executed, it is used to: obtain the motion state of the target object at the previous moment Multiple estimates of the target object, wherein the object includes a target object; prediction is performed according to multiple estimates of the motion state of the target object at the previous moment to obtain multiple predicted estimates of the motion state of the target object at the current moment; Obtain one or more observations output by the sensor; determine a matching state between each predicted estimate and each observation, wherein the matching state includes a compatible state and an incompatible state; according to the The observation that the matching state is a compatible state updates the corresponding predicted estimate; the updated estimate and the predicted estimate of the matching state being incompatible state are determined to be multiple of the motion state at the current moment estimate.
  • the memory is used to store program instructions
  • the processor is configured to execute the program
  • a movable platform including: a power system for providing movement power for the movable platform; a sensor for outputting observations of objects around the movable platform; and data processing as described in the third aspect installation.
  • a computer-readable medium stores program code for execution by an encoder, and the program code includes instructions for executing the method in the first aspect.
  • a computer program product containing instructions which when run on a computer, causes the computer to execute the method in the first aspect.
  • the embodiment of the present invention determines the matching state of the predicted estimate at the current moment with each of one or more observations at the current moment, and determines the motion of the object at the current moment based on the matching state between the predicted estimate and each observation Multiple estimates of the state, since the method maintains multiple estimates of the current time corresponding to each predicted estimate of the current time, the robustness can be improved when determining the estimation of the motion state of the object at the current time.
  • FIG. 1 is a schematic diagram of a scene for determining the estimation of the motion state of a target object provided by an embodiment of the present invention.
  • Fig. 2 is a schematic flowchart of a data processing method according to an embodiment of the present invention.
  • FIG. 3 is a schematic structural diagram of a data processing device 300 provided by an embodiment of the present invention.
  • Fig. 4 is a schematic structural diagram of a movable platform provided by an embodiment of the present invention.
  • a component when referred to as being "fixed to” another component, it can be directly on the other component or a central component may also exist. When a component is considered to be “connected” to another component, it can be directly connected to the other component or there may be a centered component at the same time.
  • the movable platform 101 is equipped with a sensor 1011 for detecting one or more objects around it (for example, the object 102 and the object 103), and the sensor 1011 can output the observation of the object (ie, sensing data).
  • the movable platform 101 can determine the estimation of the motion state of the object based on the observation of the object, and control the movable platform 101 according to the estimation of the motion state of the object around the movable platform 101.
  • the system model of the motion state estimation of the target object 102 at the previous moment is first used to determine the target object 102's motion state at the current moment. Estimation of the prediction of the motion state. Then, obtain the observations output by the sensor 1011 configured on the movable platform 101 for detecting the object.
  • the observations may include the observations of the target object 102 and the object 103.
  • the Hungarian allocation algorithm is used to divide the observations from the multiple observations. One observation is allocated to the target object 102, and the predicted estimate is updated using the allocated observation to obtain an estimate of the motion state of the target object 102 at the current moment. It is understandable that the target object may be any one of multiple objects around the movable platform.
  • the embodiment of the present invention provides a data processing method, which can improve the robustness when determining the estimation of the motion state of the object.
  • the following describes in detail a data processing method provided by an embodiment of the present invention with reference to FIG. 2.
  • the 2 is a data processing method 200 provided by an embodiment of the present invention.
  • the method 200 can be applied to a movable platform.
  • the movable platform can include sensors for outputting observations of objects around the movable platform.
  • the method 200 can Including steps 210-260, the steps 210-260 will be described in detail below.
  • Step 210 Obtain multiple estimates of the motion state of the target object at the previous moment.
  • the execution subject of the method may be a data processing device
  • the movable platform may include the data processing device.
  • the movable platform will be used as an unmanned vehicle for a schematic description. It is understandable that the unmanned vehicle described later can be equally replaced by the movable platform.
  • the objects can be, for example, at least one of static objects on the road and traffic participants.
  • the static objects can include, for example, traffic lights, road signs, buildings, and trees on the road.
  • Traffic participants can include pedestrians and vehicles on the road, for example.
  • the driverless car obtains multiple estimates of the motion state of the target object at the last moment
  • the target object here may be any one of the foregoing objects, and the motion state of the target object may include, for example, one or more of position, distance, speed, posture, acceleration, and angular acceleration.
  • the driverless car can obtain multiple estimates of the motion state of the target object at a moment in time
  • multiple estimates It can include anomaly estimation, and the variance corresponding to anomaly estimation is infinite.
  • the unmanned vehicle obtains the variances corresponding to the multiple estimates of the target object at the previous moment And prior probability among them,
  • the target object is a driving vehicle
  • the motion state of the target object may be the straight-line distance between the vehicle and the driverless vehicle (hereinafter referred to as "distance").
  • the driverless vehicle obtains multiple estimates of the distance of the vehicle at the previous moment.
  • the multiple estimates may be 5 meters, 10 meters, and 0 meters, where the variances corresponding to 5 meters, 10 meters, and 0 meters may be 1 meter, 2 meters and infinity, then 0 meter is an abnormal estimate.
  • the multiple estimates of the distance at the last moment of 5 meters, 10 meters, and 0 meters correspond to the prior probabilities of 0.5, 0.4, and 0.1, respectively.
  • Step 220 Perform prediction based on multiple estimates of the motion state of the target object at the previous moment, so as to obtain multiple predicted estimates of the motion state of the target object at the current moment.
  • the motion state of the target object at the current moment can be predicted according to the system model equation that satisfies the formula (1):
  • x k-1 represents the motion state of the target object at the previous time k-1
  • x k, k-1 represents the prediction of the motion state of the target object at the current time k
  • the motion state is the distance between the vehicle and the driverless vehicle as an example.
  • the driverless vehicle determines the multiple predicted estimates of the distance at the current time according to equation (1). 5.25 meters, 10.25 meters, and 0 meters.
  • multiple predicted variances corresponding to 5.25 meters, 10.25 meters, and 0 meters can be predicted based on the aforementioned formula.
  • the prior probabilities corresponding to the variances corresponding to multiple predicted estimates of 5.25 meters, 10.25 meters, and 0 meters are the prior probabilities corresponding to multiple estimates of 5 meters, 10 meters, and 0 meters.
  • Step 230 Obtain one or more observations output by the sensor.
  • the unmanned vehicle can obtain one or more observations of the vehicle at the current moment through the sensor
  • the observation here can be one or more distances of the vehicle at the current moment detected by the sensor.
  • the observation may include, for example, one or more of position, distance, speed, attitude, acceleration, and angular acceleration.
  • the driverless car can also obtain the variance of the one or more observations
  • the one or more observations It can include anomalous observations, and the variance of anomalous observations is infinite.
  • the driverless car can also obtain the prior probability of the one or more observations among them,
  • multiple predicted estimates of the distance between the vehicle and the unmanned vehicle at the current moment are 5.25 meters, 10.25 meters, and 0 meters.
  • the unmanned vehicle obtains multiple distance observations output by the sensor, such as 6 meters, 17 meters, and 0 meters.
  • the multiple estimates may be 6 meters, 17 meters, and 0 meters, and the corresponding variances may be 1 meter, 2 meters, and infinity, and 0 meter is an abnormal observation.
  • the prior probabilities corresponding to multiple observations of 6 meters, 17 meters, and 0 meters are 0.5, 0.3, and 0.2, respectively.
  • Step 240 The unmanned vehicle determines a matching state between each predicted estimate and each observation, where the matching state includes a compatible state and an incompatible state.
  • the unmanned vehicle determines the matching state between the estimation of each prediction and each of the one or more observations for the estimation of multiple predictions of the distance of the target object at the current moment, that is, the determination of each prediction
  • the estimate of and each of one or more observations are in a compatible state or an incompatible state.
  • the matching state is used to characterize whether the observation is an estimated measurement of the prediction.
  • the matching state is a compatible state, the observation is an estimated measurement of the prediction, and when the matching state is an incompatible state, the observation is not an estimated measurement of the prediction.
  • multiple predicted estimates of the distance between the vehicle and the unmanned vehicle at the current time are 5.25 meters, 10.25 meters, and 0 meters
  • the multiple observations of the vehicle at the current time are 6 meters, 17 meters, and 0 meters, respectively.
  • the unmanned vehicle respectively determines the matching state of the predicted estimated 5.25 meters and the observed 6 meters, 17 meters and 0 meters
  • the unmanned vehicle determines the predicted matching state of the estimated 10.25 meters and the observed 6 meters, 17 meters and 0 meters, respectively.
  • the unmanned vehicle determines the matching state of the predicted estimated 0 meter and the observed 6 meter, 17 meter and 0 meter respectively.
  • the driverless car can determine the matching state between each predicted estimate and each observation in the following ways:
  • the measurement residual, the estimated variance of each prediction, and the variance of each observation determine the matching state between each predicted estimate and each observation.
  • Step 250 Update the corresponding prediction estimate according to the observation that the matching state is the compatible state.
  • the unmanned vehicle can update the predicted estimation that the matching state with the observation is a compatible state based on the observation, and obtain An estimate of the above distance at the current moment.
  • the Kalman filter update algorithm may be executed according to the observation to update the predicted estimate that is compatible with the observed matching state to obtain the estimate at the current time corresponding to the predicted estimate.
  • the Kalman filter update algorithm may include various types of Kalman filter update algorithms, such as a traditional Kalman filter update algorithm, an extended Kalman filter update algorithm, or an unscented Kalman filter update algorithm.
  • the unmanned vehicle can update the predicted estimate in the following way: when the corresponding predicted estimate is an abnormal estimate, the observation that is consistent with the predicted estimate is substituted into the observation model equation to Obtain an estimate and use the estimate to replace the corresponding predicted estimate, where the variance of the anomaly estimate is infinite; or, when the corresponding predicted estimate is not an anomaly estimate, based on the observation pair that is consistent with the predicted estimate
  • the corresponding predicted estimate runs a filtering update algorithm to update the corresponding predicted estimate.
  • the unmanned vehicle can determine whether the predicted estimate that is compatible with a certain observation is an abnormal estimate, and how to update the predicted estimate according to whether the predicted estimate corresponding to the observation is an abnormal estimate.
  • an unmanned vehicle can determine the estimated variance of the prediction. If the variance of the predicted estimate is infinite, the unmanned vehicle can determine that the predicted estimate is an abnormal estimate.
  • the observation that is consistent with the predicted estimate can be substituted into the observation model equation, so that an estimate can be obtained through the observation model equation, and the estimate obtained through the observation model can be used to replace the original
  • the estimation of the prediction that is compatible with the observation that is, the update of the estimation of the prediction that is compatible with the observation is completed.
  • the unmanned vehicle when it is determined whether the estimation of a prediction that is compatible with a certain observation is an abnormal estimation, the unmanned vehicle will substitute the observation z k that is compatible with the predicted estimation into the observation model equation of equation (2) to obtain another an estimate of x k, x k using estimated replacement alternative with the corresponding predicted estimates x k as an estimate of the unmanned vehicle motion state of the current time.
  • the driverless vehicle can determine that the predicted estimate is not an abnormal estimate.
  • a filtering update algorithm can be run on the predicted estimate based on observations that are consistent with the predicted estimate, so as to complete the update of the predicted estimate.
  • Step 260 The unmanned vehicle determines the updated estimate and the predicted estimate whose matching state is an incompatible state as multiple estimates of the motion state at the current moment.
  • the driverless vehicle will determine its matching state with each observation.
  • the unmanned vehicle updates the predicted estimation according to the corresponding observations to obtain an estimation at the current moment corresponding to the predicted estimation;
  • the observed matching state is the predicted estimate of the incompatible state, and the driverless vehicle directly determines the predicted estimate as another estimate at the current moment.
  • the driverless car can determine multiple estimates of the aforementioned distance at the current moment.
  • one estimate can be selected from the multiple estimates at the current moment, and the movable platform can be controlled according to this estimate.
  • an unmanned vehicle can be based on the selected estimate , Control its own movement.
  • the method 200 may further include: selecting one estimate from multiple estimates at the current moment, and controlling the movable platform according to the selected estimate.
  • the unmanned vehicle can also control itself according to one of the multiple estimates at the previous time.
  • the method for unmanned vehicles to determine multiple estimates at the previous moment please refer to the above-mentioned method for determining multiple estimates at the current moment. For the sake of brevity, details are not repeated here.
  • the driverless vehicle can select one estimate from multiple estimates at the current moment in the following manner:
  • the unmanned vehicle obtains the prior probability corresponding to each of the multiple estimates of the motion state of the target object at the previous time, and obtains the prior probability corresponding to each of the one or more observations at the current time, Finally, the unmanned vehicle determines the prior probability of each estimate in the multiple estimates at the current time according to the prior probability corresponding to each of the multiple estimates at the previous moment and the prior probability corresponding to each observation.
  • the unmanned vehicle can select one estimate from the multiple estimates according to the prior probability of each of the multiple estimates at the current moment, and use the selected estimate to control itself. For example, an unmanned vehicle can select the estimate with the largest prior probability among the multiple estimates at the current moment, and control itself according to the estimate with the largest prior probability.
  • an unmanned vehicle can obtain the prior probability corresponding to each of the multiple estimates of the vehicle's distance at the previous moment, and it can also obtain one or more observations of the vehicle's distance at the current moment.
  • the prior probability corresponding to each observation in the driverless car can determine the prior probability corresponding to each observation at the current time according to the prior probability corresponding to each estimate at the previous time, and determine each of the multiple estimates at the current time.
  • An estimated prior probability can be used to determine the prior probability corresponding to each of the multiple estimates of the vehicle's distance at the previous moment.
  • the driverless vehicle can select one estimate from the multiple estimates of the distance based on the prior probability of each estimate of the vehicle's distance at the current moment, and use the estimate to control itself.
  • an unmanned vehicle can select the estimate with the largest prior probability among the multiple estimates of the distance of the vehicle at the current moment, and adjust its position with the vehicle according to the estimate with the largest prior probability.
  • the driverless vehicle can determine the prior probability of each of the multiple estimates at the current moment in the following ways:
  • an unmanned vehicle determines the combined probability of each observation and the predicted estimate at the current time based on the prior probability corresponding to each estimate and the prior probability corresponding to each observation in the multiple estimates at the previous time.
  • the unmanned vehicle obtains the conditional probability corresponding to each observation and the predicted estimation at the current moment, where the conditional probability is the observation that the matching state of the observation and the estimation at the current moment when the observation is combined with the estimation at the current moment is a certain observation
  • the estimated matching state probability with the prediction at the current moment and finally the unmanned vehicle determines the prior probability of each of the multiple estimates at the current moment according to the combined probability and the conditional probability.
  • the driverless car determines the priori of the estimation index i (denoted as the estimation #i of the previous time) among the multiple estimates of the distance of the vehicle at the previous time.
  • the probability is p i
  • the prior probability of observation #j of the driverless vehicle to determine the distance of the vehicle at the current moment is p j .
  • the unmanned vehicle can determine the current observation #j and the current prediction estimate #i according to the prior probability p i of the estimate #i at the previous time and the prior probability p j of the observation #j predicted at the current time.
  • the combination probability p j,i The combination probability p j,i .
  • an unmanned vehicle can determine the combined probability p j,i of the observation #j at the current moment and the predicted estimate #i at the current moment by the following formula:
  • the unmanned vehicle can further determine the conditional probability ⁇ j,i of the observation #j at the current moment and the estimated estimate #i at the current moment. For the conditional probability ⁇ j,i , if the matching state of the observation #j at the current moment determined in step 240 and the predicted estimate #i at the current moment is consistent, the conditional probability ⁇ j,i represents the current moment When the observation #j is combined with the predicted estimate #i at the current moment, the probability that the matching state of the observation #j at the current moment and the predicted estimate #i at the current moment is a consistent state; if the observation at the current moment determined in step 240 #j ⁇ current moment prediction estimation#i is the incompatible state of the matching state, and the conditional probability ⁇ j,i represents the current moment observation #j and current moment prediction estimation #i when combined, the current moment observation The probability that the matching state of #j and the prediction of the current moment #i is an incompatible state.
  • the unmanned vehicle may also update multiple estimates at the determined current moment in the following manner:
  • the driverless vehicle may delete estimates with a priori probability lower than a preset probability threshold from the multiple estimates at the current moment, so as to obtain multiple updated estimates at the current moment.
  • the driverless vehicle can compare the estimated prior probability at the current moment with a preset probability threshold, and through the comparison, discard the estimate corresponding to the prior probability that is less than the preset probability threshold, which will be greater than or equal to the preset probability threshold
  • the estimates corresponding to the prior probabilities of are used as the updated multiple estimates.
  • the driverless vehicle can determine the first estimate and the second estimate from multiple estimates at the current moment, where the difference between the first estimate and the second estimate is less than or equal to a preset threshold, and the driverless vehicle compares the first estimate with the second estimate.
  • the first estimate and the second estimate are merged to determine the third estimate, and the driverless vehicle can use the third estimate to replace the first estimate and the second estimate among the multiple estimates at the current moment to obtain multiple estimates updated at the current moment.
  • the driverless car can determine at least two estimates from multiple estimates of the distance of the vehicle at the current moment. For example, the driverless car determines the estimate at the current moment from multiple estimates of the distance of the vehicle at the current moment #1 (E.g., an example of the first estimate) and estimate #2 at the current time (e.g., an example of the second estimate).
  • the driverless car determines the estimate at the current moment from multiple estimates of the distance of the vehicle at the current moment #1 (E.g., an example of the first estimate) and estimate #2 at the current time (e.g., an example of the second estimate).
  • the driverless car can fuse the current time estimate #1 and the current time estimate #2 to determine the estimate #3 (for example, an example of the third estimate), for example, the driverless car can estimate the current time #1 Perform a weighting operation with the current time estimate #2 to obtain estimate #3.
  • the unmanned vehicle can replace the estimation #1 and the estimation #2 in the multiple estimates at the current moment with the estimation #3, so as to obtain the multiple estimates at the current moment after update.
  • the embodiment of the present invention also provides a computer storage medium in which program instructions are stored, and the program execution may include part or all of the steps of the above data processing method.
  • FIG. 3 is a schematic structural diagram of a data processing apparatus 300 provided by an embodiment of the present invention.
  • the data processing apparatus 300 may include: a memory 310, a processor 320, and an input/output interface 330.
  • the data processing device 300 may be configured on a movable platform, and the movable platform includes a sensor for outputting observations of objects around the movable platform.
  • the memory 310, the processor 320, and the input/output interface 330 are connected by an internal connection path.
  • the memory 310 is used to store program instructions
  • the processor 320 is used to execute the program instructions stored in the memory 310 to control the input/output interface.
  • 330 receives input data and information, and outputs data such as operation results.
  • the processor 320 may adopt a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), Application specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • the processor 320 adopts one or more integrated circuits to execute related programs to implement the technical solutions provided by the embodiments of the present invention.
  • the memory 310 may include a read-only memory and a random access memory, and provides instructions and data to the processor 320.
  • a part of the processor 320 may also include a non-volatile random access memory.
  • the processor 320 may also store device type information.
  • the memory 310 is used to store program codes, and the processor 320 is configured to execute the program instructions.
  • the program instructions are executed, they are used to: obtain multiple estimates of the motion state of the target object at the previous moment. , Wherein the object includes a target object; prediction is performed according to multiple estimates of the motion state of the target object at the previous moment to obtain multiple predicted estimates of the motion state of the target object at the current moment; acquiring the sensor Output one or more observations; determine the matching state between the estimation of each prediction and each observation, wherein the matching state includes a compatible state and an incompatible state; The observation of the tolerance state updates the corresponding predicted estimate; the updated estimate and the predicted estimate of the matching state being an incompatible state are determined as multiple estimates of the motion state at the current moment.
  • the motion state includes one or more of position, distance, speed, posture, acceleration, and angular acceleration.
  • the observation includes one or more of position, distance, velocity, attitude, acceleration, and angular acceleration.
  • one of the multiple estimates at the last moment is used to control the movable platform.
  • the processor 320 is further configured to: select one estimate from multiple estimates at the current moment; and control the movable platform according to the selected estimate.
  • the movable platform is an unmanned vehicle or an unmanned aerial vehicle.
  • the processor 320 when the processor 320 updates the corresponding prediction estimate according to the observation that the matching state is the compatible state, it is specifically configured to: when the corresponding prediction estimate is When anomaly is estimated, the observation of the compatible state is substituted into the observation model equation to obtain one, and the estimate is used to replace the corresponding predicted estimate, wherein the variance of the anomaly estimate is infinite; when the corresponding predicted When the estimation is not an abnormal estimation, a filter update algorithm is run on the corresponding predicted estimation according to the observation that the matching state is the compatible state to update the corresponding predicted estimation.
  • the filtering includes Kalman filtering.
  • the processor 320 is further configured to: obtain the variance of each estimate at the previous moment; obtain the variance of each observation; and determine the variance of each prediction.
  • the matching state with each of the observations includes: determining the measurement residual of each of the observations according to a plurality of predicted estimates and the one or more observations; and determining the measurement residual of each of the observations according to the measurement residual of each of the observations , The variance of each estimate and the variance of each observation determine the matching state between each predicted estimate and each observation.
  • the processor 320 is further configured to: obtain the prior probability corresponding to each of the multiple estimates of the motion state of the target object at the previous moment; and obtain the one or more The prior probability corresponding to each of the observations; the prior probability corresponding to each of the multiple estimates at the previous time and the prior probability corresponding to each observation are used to determine the current time The prior probability of each of the multiple estimates.
  • the processor 320 determines the prior probability corresponding to each of the multiple estimates at the previous moment and the prior probability corresponding to each observation.
  • the prior probability of each of the multiple estimates at the current moment is specifically used to: according to the prior probability corresponding to each of the multiple estimates at the previous moment and the prior probability corresponding to each observation.
  • the test probability determines the combined probability of each observation and the predicted estimate at each current moment; obtains the conditional probability corresponding to each observation and the predicted estimate at each current moment, wherein the The conditional probability is that when the observation is combined with the estimation of the prediction at the current moment, the matching state of the observation and the estimation of the prediction at the current moment is the estimation of the determined observation and the prediction at the current moment
  • the probability of the matching state; the prior probability of each of the multiple estimates at the current moment is determined according to the combined probability and the conditional probability.
  • the processor 320 is further configured to: determine an estimate with a prior probability lower than a preset probability threshold from a plurality of estimates at the current moment to obtain an updated post-modification at the current moment Multiple estimates.
  • the processor 320 is further configured to: determine a first estimate and a second estimate from a plurality of estimates at the current moment, where the difference between the first estimate and the second estimate The value is less than or equal to the preset threshold; the first estimate and the second estimate are fused to determine the third estimate; the third estimate is used to replace the first estimate and the second estimate of the multiple estimates at the current moment to obtain Multiple estimates updated at the current time.
  • the steps of the foregoing method may be completed by an integrated logic circuit of hardware in the processor 320 or instructions in the form of software.
  • the method disclosed in combination with the embodiments of the present invention may be directly embodied as being executed and completed by a hardware processor, or executed and completed by a combination of hardware and software modules in the processor.
  • the software module can be located in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, registers.
  • the storage medium is located in the memory 310, and the processor 320 reads the information in the memory 310, and completes the steps of the foregoing method in combination with its hardware. To avoid repetition, it will not be described in detail here.
  • an embodiment of the present invention also provides a movable platform, where the movable platform may include an unmanned vehicle, an unmanned aerial vehicle, an unmanned ship, etc., and the movable platform includes:
  • the power system 410 is used to provide mobile power for the movable platform
  • the sensor 420 is used to output the observation of the object around the movable platform
  • the data processing device 430 as shown in FIG. 3.
  • the computer may be implemented in whole or in part by software, hardware, firmware or any other combination.
  • software it can be implemented in the form of a computer program product in whole or in part.
  • the computer program product includes one or more computer instructions.
  • the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
  • the computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.
  • the computer instructions may be transmitted from a website, computer, server, or data center.
  • the computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center integrated with one or more available media.
  • the usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a digital video disc (DVD)), or a semiconductor medium (for example, a solid state disk (SSD)), etc.
  • the disclosed system, device, and method may be implemented in other ways.
  • the device embodiments described above are merely illustrative, for example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or It can be integrated into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional units in the various embodiments of the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Business, Economics & Management (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Signal Processing (AREA)
  • Automation & Control Theory (AREA)
  • Strategic Management (AREA)
  • Medical Informatics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Economics (AREA)
  • Data Mining & Analysis (AREA)
  • Human Resources & Organizations (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Development Economics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Navigation (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

本发明实施例提供了一种数据处理方法、装置和可移动平台，该方法包括：获取目标对象的运动状态在上一时刻的多个估计，其中，所述对象包括目标对象；根据所述目标对象上一时刻的运动状态的多个估计进行预测以得到所述目标对象在当前时刻的运动状态的多个预测的估计；获取所述传感器输出的一个或多个观测；确定所述每一个预测的估计与所述每一个观测之间的匹配状态，其中所述匹配状态包括相容状态和不相容状态；根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新；将所述更新得到的估计和所述匹配状态为不相容状态的预测的估计确定为所述运动状态在当前时刻的多个估计。这样能够在确定对象的运动状态在当前时刻的估计时提高鲁棒性。

Description

一种数据处理方法、装置和可移动平台
版权申明
本专利文件披露的内容包含受版权保护的材料。该版权为版权所有人所有。版权所有人不反对任何人复制专利与商标局的官方记录和档案中所存在的该专利文件或者该专利披露。
技术领域
本发明实施例涉及数据处理领域，并且更具体地，涉及一种数据处理的方法、装置和可移动平台。
背景技术
在可移动平台(例如，无人驾驶车或者无人飞行器)应用场景中，往往需要根据可移动平台周围的对象(例如行人、车辆等)的运动状态(例如位置、速度等)的估计对可移动平台进行控制。
目前，在确定一个目标对象的运动状态的估计时，首先根据目标对象在上一时刻的运动状态估计确定目标对象在当前时刻的运动状态的预测的估计。然后，获取可移动平台上配置的用于对对象进行探测的传感器输出的观测，所述观测可能包括多个(可移动平台周围可能存在多个对象)，进一步地，使用匈牙利分配算法将多个观测中的一个观测分配给目标对象，并利用分配的观测对所述预测的估计进行更新以获取目标对象在当前时刻的运动状态的估计。
然而，如果上述的分配是错误的，那么根据被错误分配的观测与该预测的估计将无法准确地得到对象在当前时刻的运动状态的估计，另外，一旦分配错误，就无法纠正，这种确定对象的运动状态的方式鲁棒性不高。
发明内容
本发明实施例提供一种数据处理的方法、装置和可移动平台，能够在确定对象的运动状态的估计时提高的鲁棒性。
第一方面，提供了一种数据处理的方法，包括：获取目标对象的运动状态在上一时刻的多个估计，其中，所述对象包括目标对象；根据所述目标对 象上一时刻的运动状态的多个估计进行预测以得到所述目标对象在当前时刻的运动状态的多个预测的估计；获取所述传感器输出的一个或多个观测；确定所述每一个预测的估计与所述每一个观测之间的匹配状态，其中所述匹配状态包括相容状态和不相容状态；根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新；将所述更新得到的估计和所述匹配状态为不相容状态的预测的估计确定为所述运动状态在当前时刻的多个估计。
第二方面，提供了一种数据处理的装置，所述装置配置于可移动平台，所述可移动平台包括用于输出所述可移动平台周围的对对象的观测的传感器，包括：获取模块，用于获取目标对象的运动状态在上一时刻的多个估计，其中，所述对象包括目标对象；处理模块，用于根据所述目标对象上一时刻的运动状态的多个估计进行预测以得到所述目标对象在当前时刻的运动状态的多个预测的估计；获取所述传感器输出的一个或多个观测；所述处理模块，还用于确定所述每一个预测的估计与所述每一个观测之间的匹配状态，其中所述匹配状态包括相容状态和不相容状态；所述处理模块，还用于根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新；所述处理模块，还用于将所述更新得到的估计和所述匹配状态为不相容状态的预测的估计确定为所述运动状态在当前时刻的多个估计。
第三方面，提供了一种数据处理的装置，所述装置配置于可移动平台，所述可移动平台包括用于输出所述可移动平台周围的对对象的观测的传感器，包括：存储器和处理器，其中，所述存储器，用于存储程序指令；所述处理器，被配置为执行所述程序指令，当所述程序指令被执行时，用于：获取目标对象的运动状态在上一时刻的多个估计，其中，所述对象包括目标对象；根据所述目标对象上一时刻的运动状态的多个估计进行预测以得到所述目标对象在当前时刻的运动状态的多个预测的估计；获取所述传感器输出的一个或多个观测；确定所述每一个预测的估计与所述每一个观测之间的匹配状态，其中所述匹配状态包括相容状态和不相容状态；根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新；将所述更新得到的估计和所述匹配状态为不相容状态的预测的估计确定为所述运动状态在当前时刻的多个估计。
第四方面，提供一种可移动平台，包括：动力系统，用于为可移动平台提供移动动力；传感器，用于输出可移动平台周围的对对象的观测；如第三 方面所述的数据处理的装置。
第五方面，提供一种计算机可读介质，所述计算机可读介质存储用于编码器执行的程序代码，所述程序代码包括用于执行第一方面中的方法的指令。
第六方面，提供一种包含指令的计算机程序产品，当其在计算机上运行时，使得计算机执行上述第一方面中的方法。
本发明实施例通过确定当前时刻的预测的估计与当前时刻的一个或多个观测中的每一个观测的匹配状态，并根据预测的估计与每一个观测的匹配状态，确定对象在当前时刻的运动状态的多个估计，由于该方法中维护了每一个当前时刻的预测的估计对应的当前时刻的多个估计，从而能够在确定对象的运动状态在当前时刻的估计时提高鲁棒性。
附图说明
图1是本发明实施例提供的确定目标对象的运动状态的估计的场景示意图。
图2是本发明实施例提供的一种数据处理的方法示意性流程图。
图3是本发明实施例提供的一种数据处理装置300的示意性结构图。
图4是本发明实施例提供的一种可移动平台的示意性结构图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，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，都属于本发明保护的范围。
需要说明的是，当组件被称为“固定于”另一个组件，它可以直接在另一个组件上或者也可以存在居中的组件。当一个组件被认为是“连接”另一个组件，它可以是直接连接到另一个组件或者可能同时存在居中组件。
除非另有定义，本文所使用的所有的技术和科学术语与属于本发明的技术领域的技术人员通常理解的含义相同。本文中在本发明的说明书中所使用的术语只是为了描述具体的实施例的目的，不是旨在于限制本发明。本文所使用的术语“及/或”包括一个或多个相关的所列项目的任意的和所有的组合。
下面结合附图，对本发明的一些实施方式作详细说明。在不冲突的情况下，下述的实施例及实施例中的特征可以相互组合。
如图1所示，可移动平台101配置有用于对其周围的一个或多个对象(例如对象102和对象103)进行探测的传感器1011，所述传感器1011可以输出对对象的观测(即传感数据)。可移动平台101可以根据对对象的观测确定对象的运动状态的估计，并根据可移动平台101周围的对象的运动状态的估计，对可移动平台101进行控制。
现有技术中，在确定一个目标对象(例如对象102)的运动状态的估计时，首先根据目标对象102在上一时刻的运动状态估计的系统模型通过预测的方式确定目标对象102在当前时刻的运动状态的预测的估计。然后，获取可移动平台101上配置的用于对对象进行探测的传感器1011输出的观测，所述观测可能包括目标对象102和对象103的观测，进一步地，使用匈牙利分配算法将多个观测中的一个观测分配给目标对象102，并利用分配的观测对所述预测的估计进行更新以获取目标对象102在当前时刻的运动状态的估计。可以理解的是，目标对象可以是可移动平台周围的多个对象中的任意一个。
然而，如果上述的分配是错误的，那么根据被错误分配的观测与该预测的估计将无法准确地得到对象在当前时刻的运动状态的估计，另外，一旦分配错误，就无法纠正，这种确定对象的运动状态的方式鲁棒性不高。
有鉴于此，本发明实施例提供一种数据处理的方法，能够在确定对象的运动状态的估计时提高鲁棒性。
下面结合图2，详细描述本发明实施例提供的一种数据处理的方法，通过确定当前时刻的预测的估计与当前时刻的一个或多个观测中的每一个观测的匹配状态，并根据预测的估计与每一个观测的匹配状态，确定对象在当前时刻的运动状态的多个估计，由于该方法中维护了每一个当前时刻的预测的估计对应的当前时刻的多个估计，从而能够在确定对象的运动状态在当前时刻的估计时提高鲁棒性。在得到了对象的运动状态在当前时刻的多个估计时，可以按照预定的策略来选用，另外，即便本次选用的估计是错误的，由于维护了对象的多个估计，在下一个时刻时，也可以恢复和纠正过来，不会产生连续性的错误。
图2是本发明实施例提供的一种数据处理的方法200，方法200可以应 用于可移动平台，该可移动平台可以包括用于输出可移动平台周围的对对象的观测的传感器，方法200可以包括步骤210-260，下面分别对步骤210-260进行详细描述。
步骤210：获取目标对象的运动状态在上一时刻的多个估计。
具体地，所述方法的执行主体可以是数据处理的装置，所述可移动平台可以包括所述数据处理的装置。接下来将以可移动平台为无人驾驶车来进行示意性说明，可以理解的是，后述的无人驾驶车可以被可移动平台同等地替换。
无人驾驶车周围存在多个对象，此处的对象例如可以是道路上的静态物与交通参与者中的至少一种，其中，静态物例如可以包括道路上的红绿灯、路标、建筑物与树木，交通参与者例如可以包括道路上的行人与车辆。
无人驾驶车获取目标对象上一时刻的运动状态的多个估计
Figure PCTCN2019114391-appb-000001
此处的目标对象可以是上述对象中的任意一个，目标对象的运动状态例如可以包括位置、距离、速度、姿态、加速度、角加速度中的一种或多种。在某些情况中，无人驾驶车可以获取目标对象上一时刻的运动状态的多个估计
Figure PCTCN2019114391-appb-000002
分别对应的方差
Figure PCTCN2019114391-appb-000003
可以理解的是，在某些情况中，多个估计
Figure PCTCN2019114391-appb-000004
可以包括异常估计，异常估计对应的方差为无穷大。在某些情况中无人驾驶车获取目标对象上一时刻的所述多个估计对应的方差
Figure PCTCN2019114391-appb-000005
和先验概率
Figure PCTCN2019114391-appb-000006
其中，
Figure PCTCN2019114391-appb-000007
例如，目标对象为正在行驶的车辆，目标对象的运动状态可以为车辆与无人驾驶车之间的直线距离(以下简称“距离”)。无人驾驶车获取车辆在上一时刻的距离的多个估计，例如所述多个估计可以为5米、10米和0米，其中，5米、10米和0米分别对应的方差可以是1米、2米和无穷大，则0米是异常估计。上一时刻的距离的多个估计5米、10米和0米分别对应的先验概率分别为0.5、0.4和0.1。
步骤220：根据目标对象上一时刻的运动状态的多个估计进行预测，以得到目标对象在当前时刻的运动状态的多个预测的估计。
具体地，根据目标对象在上一时刻的运动状态的多个估计，对目标对象在当前时刻的运动状态进行预测，得到车辆在当前时刻的运动状态的多个预测的估计
Figure PCTCN2019114391-appb-000008
进一步地，可以根据满足式(1)的系统模型方程对目标对象在当前时 刻的运动状态进行预测：
x k,k-1＝f k-1(x k-1)  (1)
其中，x k-1代表目标对象在上一时刻k-1的运动状态，x k,k-1代表目标对象在当前时刻k的运动状态的预测
Figure PCTCN2019114391-appb-000009
在某些情况中，还可以根据上一时刻的估计对应的方差
Figure PCTCN2019114391-appb-000010
得到当前时刻的预测的估计的方差
Figure PCTCN2019114391-appb-000011
例如
Figure PCTCN2019114391-appb-000012
可以理解的是，所述当前时刻的预测的估计的先验概率为对应的上一时刻的估计的先验概率。
这里继续以运动状态为所述车辆与无人驾驶车之间的距离为例来进行说明。如前所述，在上一时刻的距离的多个估计5米、10米和0米，无人驾驶车根据式(1)确定当前时刻的距离的多个预测的估计5.25米、10.25米和0米。在某些情况中，可以根据如前所述的公式预测出多个预测的估计5.25米、10.25米和0米对应的方差。可以理解的是，多个预测的估计5.25米、10.25米和0米对应的方差对应的先验概率即为多个估计5米、10米和0米对应的先验概率。
步骤230：获取传感器输出的一个或多个观测。
具体地，无人驾驶车可以通过传感器获取车辆在当前时刻的一个或者多个观测
Figure PCTCN2019114391-appb-000013
此处的观测可以是传感器检测到的车辆在当前时刻的一个或者多个距离。在本发明实施例中，观测例如可以包括位置、距离、速度、姿态、加速度、角加速度中的一种或多种。在某些情况中，无人驾驶车还可以获取所述一个或多个观测的方差
Figure PCTCN2019114391-appb-000014
所述一个或者多个观测
Figure PCTCN2019114391-appb-000015
可以包括异常观测，异常观测的方差为无穷大。在某些情况中，无人驾驶车还可以获取所述一个或多个观测的先验概率
Figure PCTCN2019114391-appb-000016
其中，
Figure PCTCN2019114391-appb-000017
例如，当前时刻车辆与无人驾驶车之间的距离的多个预测的估计5.25米、10.25米和0米。无人驾驶车获取传感器输出的多个距离观测，例如6米、17米和0米。所述多个估计可以为6米、17米和0米分别对应的方差可以是1米、2米和无穷大，则0米是异常观测。多个观测6米、17米和0米分别对应的先验概率分别为0.5、0.3和0.2。
步骤240：无人驾驶车确定每一个预测的估计与每一个观测之间的匹配状态，其中匹配状态包括相容状态和不相容状态。
具体地，无人驾驶车针对目标对象在当前时刻的距离的多个预测的估计， 确定每一个预测的估计与一个或者多个观测中的每一个观测之间的匹配状态，即确定每一个预测的估计与一个或者多个观测中的每一个观测之间是相容状态或者不相容状态。其中，匹配状态用于表征所述观测是否为该预测的估计的测量。当所述匹配状态为相容状态时，则所述观测为该预测的估计的测量，当所述匹配状态为不相容状态时，则所述观测不为该预测的估计的测量。
例如，当前时刻车辆与无人驾驶车之间的距离的多个预测的估计5.25米、10.25米和0米，得到车辆在当前时刻的多个观测分别为6米、17米和0米。无人驾驶车分别确定预测的估计5.25米与观测6米、17米和0米的匹配状态，无人驾驶车分别确定预测的估计10.25米与观测6米、17米和0米的匹配状态，无人驾驶车分别确定预测的估计0米与观测6米、17米和0米的匹配状态。
作为一种实现方式，无人驾驶车可以通过以下方式确定每一个预测的估计与每一个观测之间的匹配状态：
获取在当前时刻的每一个预测的估计的方差，并且获取每一个观测的方差，根据多个预测的估计和一个或多个观测确定每一个观测的量测残差，最终根据每一个观测的量测残差、每一个预测的估计的方差和每一个观测的方差确定每一个预测的估计与每一个观测之间的匹配状态。
具体地，无人驾驶车的观测方程为z k＝h k(x k)，可以根据多个预测的估计
Figure PCTCN2019114391-appb-000018
和一个或多个观测
Figure PCTCN2019114391-appb-000019
确定每一个观测的量测残差
Figure PCTCN2019114391-appb-000020
Figure PCTCN2019114391-appb-000021
然后，可以根据所述每一个预测的估计的方差
Figure PCTCN2019114391-appb-000022
和每一个观测的方差
Figure PCTCN2019114391-appb-000023
可以确定所述量测残差对应的方差
Figure PCTCN2019114391-appb-000024
进一步地，可以根据所述量测残差和所述量测残差对应的方差获取相容性评估参数
Figure PCTCN2019114391-appb-000025
当所述相容性评估参数小于或等于预设阈值时，确定预测的估计和观测之间的匹配状态为相容状态，否则为不相容状态。
步骤250：根据匹配状态为相容状态的观测对对应的预测的估计进行更新。
具体地，如果某一个观测与某一个预测的估计的匹配状态为相容状态，则无人驾驶车可以根据该观测，对与该观测的匹配状态为相容状态的预测的估计进行更新，得到上述距离在当前时刻的一个估计。进一步地，可以根据 所述观测运行卡尔曼滤波更新算法对与该观测的匹配状态为相容状态的预测的估计进行更新以获取与该预测的估计对应的当前时刻的估计。所述卡尔曼滤波更新算法可以包括各种类型的卡尔曼滤波更新算法，例如传统卡尔曼滤波更新算法、扩展卡尔曼滤波更新算法或无迹卡尔曼滤波更新算法等。
作为一种实现方式，无人驾驶车可以通过以下方式对预测的估计进行更新：当对应的预测的估计为异常估计时，将与该预测的估计为相容状态的观测代入观测模型方程中以获取一个估计，使用该估计替换对应的预测的估计，其中，异常估计的方差为无穷大；或者，当对应的预测的估计不为异常估计时，根据与该预测的估计为相容状态的观测对对应的预测的估计运行滤波更新算法以对对应的预测的估计进行更新。
具体地，无人驾驶车可以确定与某一个观测为相容状态的预测的估计是否为异常估计，根据与该观测对应的预测的估计是否为异常估计，确定如何对该预测的估计进行更新。
例如，无人驾驶车可以确定预测的估计的方差，如果预测的估计的方差为无穷大，此时无人驾驶车可以确定该预测的估计为异常估计。在对该预测的估计进行更新时，可以将与该预测的估计为相容状态的观测代入观测模型方程中，从而通过该观测模型方程获取一个估计，使用通过观测模型获取的该估计替换掉原来的与该观测为相容状态的预测的估计，即完成对与该观测为相容状态的预测的估计的更新。
例如，当确定与某一个观测为相容状态的预测的估计是否为异常估计，无人驾驶车将与该预测的估计为相容状态的观测z k代入式(2)的观测模型方程得到另一个估计x k，并使用x k替换用该估计替换对应的预测的估计，将x k作为无人驾驶车当前时刻的运动状态的一个估计。
z k＝h k(x k)  (2)
如果该预测的估计的方差不为无穷大，此时无人驾驶车可以确定该预测的估计不为异常估计。在对该预测的估计进行更新时，可以根据与该预测的估计为相容状态的观测，对该预测的估计运行滤波更新算法，从而完成对该预测的估计的更新。
步骤260：无人驾驶车将更新得到的估计和匹配状态为不相容状态的预测的估计确定为运动状态在当前时刻的多个估计。
具体地，对于每一个预测的估计，无人驾驶车均会确定其与每一个观测 的匹配状态。对于与观测的匹配状态为相容状态的预测的估计，无人驾驶车根据相对应的观测，对该预测的估计进行更新，得到与该预测的估计对应的在当前时刻的一个估计；对于与观测的匹配状态为不相容状态的预测的估计，无人驾驶车则直接将该预测的估计确定为在当前时刻的另一个估计。最终，无人驾驶车可以确定上述距离在当前时刻的多个估计。
在确定了上述距离在当前时刻的多个估计后，还可以在当前时刻的多个估计中选出一个估计，根据该估计对可移动平台进行控制，例如，无人驾驶车可以根据选中的估计，控制其自身的运动。此时，方法200还可以包括：从当前时刻的多个估计选中一个估计，根据选中的估计对可移动平台进行控制。
需要说明的是，在本发明实施例中，无人驾驶车还可以根据上一时刻的多个估计中的一个估计对其自身进行控制。无人驾驶车确定上一时刻的多个估计的方法请参考上述确定当前时刻的多个估计的方法，为了简洁，此处不再赘述。
在本发明实施例中，无人驾驶车可以通过以下方式从当前时刻的多个估计中选择一个估计：
例如，无人驾驶车获取目标对象的运动状态在上一时刻的多个估计中每一个估计对应的先验概率，并获取当前时刻的一个或多个观测中每一个观测对应的先验概率，最终无人驾驶车根据在上一时刻的多个估计中每一个估计对应的先验概率和每一个观测对应的先验概率确定在当前时刻的多个估计中每一个估计的先验概率。
无人驾驶车可以根据当前时刻的多个估计中每一个估计的先验概率，从多个估计中选择一个估计，使用该选中的估计对其自身进行控制。例如，无人驾驶车可以选择当前时刻的多个估计中先验概率最大的估计，根据该先验概率最大的估计对其自身进行控制。
具体地，以车辆为例，无人驾驶车可以获取车辆的距离在上一时刻的多个估计中每一个估计对应的先验概率，还可以获取车辆的距离在当前时刻的一个或多个观测中每一个观测对应的先验概率，无人驾驶车可以根据上一时刻的每一个估计对应的先验概率与当前时刻的每一个观测对应的先验概率，确定当前时刻的多个估计中每一个估计的先验概率。
无人驾驶车可以根据车辆的距离在当前时刻的多个估计中每一个估计 的先验概率，从距离的多个估计中选择一个估计，使用该估计对其自身进行控制。例如，无人驾驶车可以选择车辆的距离在当前时刻的多个估计中先验概率最大的估计，根据该先验概率最大的估计调整其与车辆之间的位置。
作为一种实现方式，无人驾驶车可以通过以下方式确定当前时刻的多个估计中每一个估计的先验概率：
例如，无人驾驶车根据在上一时刻的多个估计中每一个估计对应的先验概率和每一个观测对应的先验概率确定每一个观测与每一个当前时刻的预测的估计的组合概率，无人驾驶车获取在每一个观测与每一个当前时刻的预测的估计对应的条件概率，其中，条件概率为在观测与当前时刻的估计组合时观测与当前时刻的估计的匹配状态为确定的观测与当前时刻的预测的估计的匹配状态的概率，最终无人驾驶车根据组合概率和条件概率确定当前时刻的多个估计中每一个估计的先验概率。
具体地，同样以车辆为目标对象为例，例如，无人驾驶车确定车辆的距离在上一时刻的多个估计中索引为i的估计(记为上一时刻的估计#i)的先验概率为p i，由于上一时刻的估计#i的先验概率与根据上一时刻的估计#i确定的当前时刻的预测的估计#i的先验概率相同，则当前时刻的预测的估计的先验概率为p i。无人驾驶车确定车辆的距离在当前时刻的观测#j的先验概率为p j。无人驾驶车可以根据上一时刻的估计#i的先验概率p i与当前时刻预测的观测#j的先验概率p j，确定当前时刻的观测#j与当前时刻的预测的估计#i的组合概率p j,i。
例如，无人驾驶车可以通过下式确定当前时刻的观测#j与当前时刻的预测的估计#i的组合概率p j,i：
p j,i＝p j×p i  (3)
无人驾驶车可以进一步确定当前时刻的观测#j与当前时刻的预测的估计#i的条件概率α j,i。对于所述条件概率α j,i，如果步骤240中确定的当前时刻的观测#j与当前时刻的预测的估计#i的匹配状态的相容状态，条件概率α j,i则代表当前时刻的观测#j与当前时刻的预测的估计#i组合时，当前时刻的观测#j与当前时刻的预测的估计#i的匹配状态为相容状态的概率；如果步骤240中确定的当前时刻的观测#j与当前时刻的预测的估计#i的匹配状态的不相容状态，条件概率α j,i则代表当前时刻的观测#j与当前时刻的预测的估计#i组合时，当前时刻的观测#j与当前时刻的预测的估计#i的匹配状态为不相 容状态的概率。
进一步地，可以根据当前时刻的观测#j与当前时刻的预测的估计#i的组合概率p j,i和当前时刻的观测#j与当前时刻的预测的估计#i的条件概率α j,i确定当前时刻的估计#i的一个先验概率P′ j,i，例如：
P′ j,i＝p j,i×α j,i＝p j×p i×α j,i  (4)
其中，当当前时刻的观测#j不为异常观测且当前时刻的预测的估计#i不为异常估计时，若观测#j和预测的估计#i之间的匹配状态为相容状态，条件概率α j,i可以为α 0，若观测#j和预测的估计#i之间的匹配状态为不相容状态，条件概率α j,i可以为1-α 0，其中，α 0可以为较大概率值，例如α 0＝0.9。
当当前时刻的观测#j不为异常观测且当前时刻的预测的估计#i为异常估计时，条件概率α j,i可以为1-α 1，其中，α 1可以为较大概率值，例如α 1＝0.9。
当当前时刻的观测#j为异常观测且当前时刻的预测的估计#i不为异常估计时，条件概率α j,i可以为1-α 2，其中，α 2可以为较大概率值，例如α 2＝0.9。
当当前时刻的观测#j为异常观测且当前时刻的预测的估计#i为异常估计时，条件概率α j,i可以为1-α 3，其中，α 3可以为较大概率值，例如α 3＝0.9。
在本发明实施例中，无人驾驶车还可以通过以下方式对确定的当前时刻的多个估计进行更新：
方式#1
例如，无人驾驶车可以从在当前时刻的多个估计中删除先验概率低于预设概率阈值的估计，以获取当前时刻的更新后的多个估计。
具体地，无人驾驶车可以将当前时刻的估计的先验概率与预设概率阈值进行比较，通过比较，丢弃小于预设概率阈值的先验概率对应的估计，将大于或等于预设概率阈值的先验概率对应的估计作为更新后的多个估计。
方式#2
例如，无人驾驶车可以从当前时刻的多个估计中确定第一估计和第二估计，其中，第一估计和第二估计中的差值小于或等于预设阈值，无人驾驶车对第一估计和第二估计进行融合以确定第三估计，无人驾驶车可以利用第三估计替换当前时刻的多个估计中的第一估计和第二估计以获取当前时刻更新后的多个估计。
具体地，无人驾驶车可以从车辆的距离在当前时刻的多个估计确定至少 两个估计，例如，无人驾驶车从车辆的距离在当前时刻的多个估计确定出当前时刻的估计#1(例如，第一估计的一例)与当前时刻的估计#2(例如，第二估计的一例)。
无人驾驶车可以对当前时刻的估计#1与当前时刻的估计#2进行融合以确定估计#3(例如，第三估计的一例)，例如，无人驾驶车可以对当前时刻的估计#1与当前时刻的估计#2进行加权运算，从而得到估计#3.
无人驾驶车可以将当前时刻的多个估计中的估计#1与估计#2替换为估计#3，从而获取到更新后的当前时刻的多个估计。
本发明实施例还提供了一种计算机存储介质，该计算机存储介质中存储有程序指令，所述程序执行时可包括上文数据处理的方法的部分或全部步骤。
上文结合图2，详细描述了本发明实施例的方法实施例，下文结合图3与图4，详细描述本发明实施例的装置实施例，由于装置实施例可以执行上述方法，因此未详细描述的部分可以参见前面各方法实施例。
图3是本发明实施例提供的一种数据处理的装置300的示意性结构图。该数据处理的装置300可以包括：存储器310、处理器320、输入/输出接口330。该数据处理的装置300可以配置于可移动平台，所述可移动平台包括用于输出所述可移动平台周围的对对象的观测的传感器。
其中，存储器310、处理器320和输入/输出接口330通过内部连接通路相连，该存储器310用于存储程序指令，该处理器320用于执行该存储器310存储的程序指令，以控制输入/输出接口330接收输入的数据和信息，输出操作结果等数据。
应理解，在本发明实施例中，该处理器320可以采用中央处理单元(central processing unit，CPU)，该处理器还可以是其它通用处理器、数字信号处理器(digital signal processor，DSP)、专用集成电路(application specific integrated circuit，ASIC)、现成可编程门阵列(field programmable gate Array，FPGA)或者其它可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。或者该处理器320采用一个或多个集成电路，用于执行相关程序，以实现本发明实施例所提供的技术方案。
该存储器310可以包括只读存储器和随机存取存储器，并向处理器320提供指令和数据。处理器320的一部分还可以包括非易失性随机存取存储器。 例如，处理器320还可以存储设备类型的信息。
所述存储器310用于存储程序代码，所述处理器320被配置为执行所述程序指令，当所述程序指令被执行时，用于：获取目标对象的运动状态在上一时刻的多个估计，其中，所述对象包括目标对象；根据所述目标对象上一时刻的运动状态的多个估计进行预测以得到所述目标对象在当前时刻的运动状态的多个预测的估计；获取所述传感器输出的一个或多个观测；确定所述每一个预测的估计与所述每一个观测之间的匹配状态，其中所述匹配状态包括相容状态和不相容状态；根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新；将所述更新得到的估计和所述匹配状态为不相容状态的预测的估计确定为所述运动状态在当前时刻的多个估计。
可选地，在一些实施例中，所述运动状态包括位置、距离、速度、姿态、加速度、角加速度中的一种或多种。
可选地，在一些实施例中，所述观测包括位置、距离、速度、姿态、加速度、角加速度中的一种或多种。
可选地，在一些实施例中，所述上一时刻的多个估计中的一个估计用于控制可移动平台。
可选地，在一些实施例中，所述处理器320还用于：从所述当前时刻的多个估计选中一个估计；根据所述选中的估计控制可移动平台。
可选地，在一些实施例中，所述可移动平台为无人驾驶车或无人飞行器。
可选地，在一些实施例中，所述处理器320在根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新时，具体用于：当所述对应的预测的估计为异常估计时，将所述相容状态的观测代入观测模型方程中以获取一个，使用所述估计替换所述对应的预测的估计，其中，异常估计的方差为无穷大；当所述对应的预测的估计不为异常估计时，根据所述匹配状态为相容状态的观测对对应的预测的估计运行滤波更新算法以对所述对应的预测的估计进行更新。
可选地，在一些实施例中，所述滤波包括卡尔曼滤波。
可选地，在一些实施例中，所述处理器320还用于：获取在上一时刻的每一个估计的方差；获取所述每一个观测的方差；所述确定所述每一个预测的估计与所述每一个观测之间的匹配状态包括：根据多个预测的估计和所述一个或多个观测确定所述每一个观测的量测残差；根据所述每一个观测的量 测残差、所述每一个估计的方差和所述每一个观测的方差确定所述每一个预测的估计与所述每一个观测之间的匹配状态。
可选地，在一些实施例中，所述处理器320还用于：获取所述目标对象运动状态在上一时刻的多个估计中每一个估计对应的先验概率；获取所述一个或多个观测中每一个观测对应的先验概率；根据所述在上一时刻的多个估计中每一个估计对应的先验概率和所述每一个观测对应的先验概率确定所述在当前时刻的多个估计中每一个估计的先验概率。
可选地，在一些实施例中，所述处理器320在根据所述在上一时刻的多个估计中每一个估计对应的先验概率和所述每一个观测对应的先验概率确定所述在当前时刻的多个估计中每一个估计的先验概率时，具体用于：根据所述在上一时刻的多个估计中每一个估计对应的先验概率和所述每一个观测对应的先验概率确定所述每一个观测与所述每一个当前时刻的预测的估计的组合概率；获取在所述每一个观测与所述每一个当前时刻的预测的估计对应的条件概率，其中，所述条件概率为在所述观测与所述当前时刻的预测的估计组合时所述观测与所述当前时刻的预测的估计的匹配状态为所述确定的所述观测与所述当前时刻的预测的估计的匹配状态的概率；根据所述组合概率和所述条件概率确定所述在当前时刻的多个估计中每一个估计的先验概率。
可选地，在一些实施例中，所述处理器320还用于：从所述当前时刻的多个估计中确定先验概率低于预设概率阈值的估计以获取所述当前时刻的更新后的多个估计。
可选地，在一些实施例中，所述处理器320还用于：从所述当前时刻的多个估计中确定第一估计和第二估计，其中，第一估计和第二估计中的差值小于或等于预设阈值；对第一估计和第二估计进行融合以确定第三估计；利用所述第三估计替换所述当前时刻的多个估计中的第一估计和第二估计以获取当前时刻更新后的多个估计。
在实现过程中，上述方法的各步骤可以通过处理器320中的硬件的集成逻辑电路或者软件形式的指令完成。结合本发明实施例所公开的方法可以直接体现为硬件处理器执行完成，或者用处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储 介质位于存储器310，处理器320读取存储器310中的信息，结合其硬件完成上述方法的步骤。为避免重复，这里不再详细描述。
如图4所示，本发明实施例还提供一种可移动平台，其中，所述可移动平台可以包括无人驾驶车、无人飞行器、无人船等等，所述可移动平台包括：
动力系统410，用于为可移动平台提供移动动力；
传感器420，用于输出可移动平台周围的对对象的观测；
如图3所述的数据处理的装置430。
在上述实施例中，可以全部或部分地通过软件、硬件、固件或者其他任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。所述计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行所述计算机程序指令时，全部或部分地产生按照本发明实施例所述的流程或功能。所述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输，例如，所述计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(digital subscriber line，DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。所述计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。所述可用介质可以是磁性介质(例如，软盘、硬盘、磁带)、光介质(例如数字视频光盘(digital video disc，DVD))、或者半导体介质(例如固态硬盘(solid state disk，SSD))等。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本发明实施例的范围。
在本发明实施例所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到 另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本发明实施例各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
以上所述，仅为本发明实施例的具体实施方式，但本发明实施例的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明实施例揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本发明实施例的保护范围之内。因此，本发明实施例的保护范围应以所述权利要求的保护范围为准。

Claims (28)

  1. 一种数据处理的方法，应用于可移动平台，所述可移动平台包括用于输出可移动平台周围的对对象的观测的传感器，其特征在于，所述方法包括：
    获取目标对象的运动状态在上一时刻的多个估计，其中，所述对象包括目标对象；
    根据所述目标对象上一时刻的运动状态的多个估计进行预测以得到所述目标对象在当前时刻的运动状态的多个预测的估计；
    获取所述传感器输出的一个或多个观测；
    确定所述每一个预测的估计与所述每一个观测之间的匹配状态，其中所述匹配状态包括相容状态和不相容状态；
    根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新；
    将所述更新得到的估计和所述匹配状态为不相容状态的预测的估计确定为所述运动状态在当前时刻的多个估计。
  2. 根据权利要求1所述的方法，其特征在于，所述运动状态包括位置、距离、速度、姿态、加速度、角加速度中的一种或多种。
  3. 根据权利要求1或2所述的方法，其特征在于，所述观测包括位置、距离、速度、姿态、加速度、角加速度中的一种或多种。
  4. 根据权利要求1-3任一项所述的方法，其特征在于，所述上一时刻的多个估计中的一个估计用于控制可移动平台。
  5. 根据权利要求1-4任一项所述的方法，其特征在于，所述方法还包括：
    从所述当前时刻的多个估计选中一个估计；
    根据所述选中的估计控制可移动平台。
  6. 根据权利要求1-5任一项所述的方法，其特征在于，所述可移动平台为无人驾驶车或无人飞行器。
  7. 根据权利要求1-6任一项所述的方法，其特征在于，所述根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新，包括：
    当所述对应的预测的估计为异常估计时，将所述相容状态的观测代入观测模型方程中以获取一个估计，使用所述获取的估计替换所述对应的预测的估计，其中，异常估计的方差为无穷大；
    当所述对应的预测的估计不为异常估计时，根据所述匹配状态为相容状态的观测对对应的预测的估计运行滤波更新算法以对所述对应的预测的估计进行更新。
  8. 根据权利要求7所述的方法，其特征在于，所述滤波包括卡尔曼滤波。
  9. 根据权利要1-8任一项所述的方法，其特征在于，所述方法还包括：
    获取当前时刻的每一个预测的估计的方差；
    获取所述每一个观测的方差；
    所述确定所述每一个预测的估计与所述每一个观测之间的匹配状态包括：
    根据多个预测的估计和所述一个或多个观测确定所述每一个观测的量测残差；
    根据所述每一个观测的量测残差、所述每一个预测的估计的方差和所述每一个观测的方差确定所述每一个预测的估计与所述每一个观测之间的匹配状态。
  10. 根据权利要求1-9任一项所述的方法，其特征在于，所述方法还包括：
    获取所述目标对象运动状态在上一时刻的多个估计中每一个估计对应的先验概率；
    获取所述一个或多个观测中每一个观测对应的先验概率；
    根据所述在上一时刻的多个估计中每一个估计对应的先验概率和所述每一个观测对应的先验概率确定所述在当前时刻的多个估计中每一个估计的先验概率。
  11. 根据权利要求10所述的方法，其特征在于，
    所述根据所述在上一时刻的多个估计中每一个估计对应的先验概率和所述每一个观测对应的先验概率确定所述在当前时刻的多个估计中每一个估计的先验概率，包括：
    根据所述在上一时刻的多个估计中每一个估计对应的先验概率和所述每一个观测对应的先验概率确定所述每一个观测与所述每一个当前时刻的预测的估计的组合概率；
    获取在所述每一个观测与所述每一个当前时刻的估计对应的条件概率， 其中，所述条件概率为在所述观测与所述当前时刻的预测的估计组合时所述观测与所述当前时刻的预测的估计的匹配状态为所述确定的所述观测与所述当前时刻的预测的估计的匹配状态的概率；
    根据所述组合概率和所述条件概率确定所述在当前时刻的多个估计中每一个估计的先验概率。
  12. 根据权利要求10或11所述的方法，其特征在于，所述方法还包括：从所述当前时刻的多个估计中删除先验概率低于预设概率阈值的估计以获取所述当前时刻的更新后的多个估计。
  13. 根据权利要求1-12任一项所述的方法，其特征在于，所述方法还包括：
    从所述当前时刻的多个估计中确定第一估计和第二估计，其中，第一估计和第二估计中的差值小于或等于预设阈值；
    对第一估计和第二估计进行融合以确定第三估计；
    利用所述第三估计替换所述当前时刻的多个估计中的第一估计和第二估计以获取当前时刻更新后的多个估计。
  14. 一种数据处理的装置，所述装置配置于可移动平台，所述可移动平台包括用于输出所述可移动平台周围的对对象的观测的传感器，其特征在于，包括：存储器和处理器，其中，
    所述存储器，用于存储程序指令；
    所述处理器，被配置为执行所述程序指令，当所述程序指令被执行时，用于：
    获取目标对象的运动状态在上一时刻的多个估计，其中，所述对象包括目标对象；
    根据所述目标对象上一时刻的运动状态的多个估计进行预测以得到所述目标对象在当前时刻的运动状态的多个预测的估计；
    获取所述传感器输出的一个或多个观测；
    确定所述每一个预测的估计与所述每一个观测之间的匹配状态，其中所述匹配状态包括相容状态和不相容状态；
    根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新；
    将所述更新得到的估计和所述匹配状态为不相容状态的预测的估计确定为所述运动状态在当前时刻的多个估计。
  15. 根据权利要求14所述的装置，其特征在于，所述运动状态包括位置、距离、速度、姿态、加速度、角加速度中的一种或多种。
  16. 根据权利要求14或15所述的装置，其特征在于，所述观测包括位置、距离、速度、姿态、加速度、角加速度中的一种或多种。
  17. 根据权利要求14-16任一项所述的装置，其特征在于，所述上一时刻的多个估计中的一个估计用于控制可移动平台。
  18. 根据权利要求14-17任一项所述的装置，其特征在于，所述处理器还用于：
    从所述当前时刻的多个估计选中一个估计；
    根据所述选中的估计控制可移动平台。
  19. 根据权利要求14-18任一项所述的装置，其特征在于，所述可移动平台为无人驾驶车或无人飞行器。
  20. 根据权利要求14-19任一项所述的装置，其特征在于，所述处理器在根据所述匹配状态为相容状态的观测对对应的预测的估计进行更新时，具体用于：
    当所述对应的预测的估计为异常估计时，将所述相容状态的观测代入观测模型方程中以获取一个估计，使用所述获取的估计替换所述对应的预测的估计，其中，异常估计的方差为无穷大；
    当所述对应的预测的估计不为异常估计时，根据所述匹配状态为相容状态的观测对对应的预测的估计运行滤波更新算法以对所述对应的预测的估计进行更新。
  21. 根据权利要求20所述的装置，其特征在于，所述滤波包括卡尔曼滤波。
  22. 根据权利要14-21任一项所述的装置，其特征在于，所述处理器还用于：
    获取当前时刻的每一个预测的估计的方差；
    获取所述每一个观测的方差；
    所述确定所述每一个预测的估计与所述每一个观测之间的匹配状态包括：
    根据多个预测的估计和所述一个或多个观测确定所述每一个观测的量测残差；
    根据所述每一个观测的量测残差、所述每一个预测的估计的方差和所述每一个观测的方差确定所述每一个预测的估计与所述每一个观测之间的匹配状态。
  23. 根据权利要求14-22任一项所述的装置，其特征在于，所述处理器还用于：
    获取所述目标对象运动状态在上一时刻的多个估计中每一个估计对应的先验概率；
    获取所述一个或多个观测中每一个观测对应的先验概率；
    根据所述在上一时刻的多个估计中每一个估计对应的先验概率和所述每一个观测对应的先验概率确定所述在当前时刻的多个估计中每一个估计的先验概率。
  24. 根据权利要求23所述的装置，其特征在于，所述处理器在根据所述在上一时刻的多个估计中每一个估计对应的先验概率和所述每一个观测对应的先验概率确定所述在当前时刻的多个估计中每一个估计的先验概率时，具体用于：
    根据所述在上一时刻的多个估计中每一个估计对应的先验概率和所述每一个观测对应的先验概率确定所述每一个观测与所述每一个当前时刻的预测的估计的组合概率；
    获取在所述每一个观测与所述每一个当前时刻的预测的估计对应的条件概率，其中，所述条件概率为在所述观测与所述当前时刻的预测的估计组合时所述观测与所述当前时刻的预测的估计的匹配状态为所述确定的所述观测与所述当前时刻的预测的估计的匹配状态的概率；
    根据所述组合概率和所述条件概率确定所述在当前时刻的多个估计中每一个估计的先验概率。
  25. 根据权利要求23或24所述的装置，其特征在于，所述处理器还用于：
    从所述当前时刻的多个估计中删除先验概率低于预设概率阈值的估计以获取所述当前时刻的更新后的多个估计。
  26. 根据权利要求14-25任一项所述的装置，其特征在于，所述处理器还用于：
    从所述当前时刻的多个估计中确定第一估计和第二估计，其中，第一估 计和第二估计中的差值小于或等于预设阈值；
    对第一估计和第二估计进行融合以确定第三估计；
    利用所述第三估计替换所述当前时刻的多个估计中的第一估计和第二估计以获取当前时刻更新后的多个估计。
  27. 一种可移动平台，其特征在于，包括：
    动力系统，用于为可移动平台提供移动动力；
    传感器，用于输出可移动平台周围的对对象的观测；
    如权利要求14至26中任一项所述的数据处理的装置。
  28. 根据权利要求27所述的可移动平台，其特征在于，所述可移动平台为无人驾驶车或无人飞行器。
PCT/CN2019/114391 2019-10-30 2019-10-30 一种数据处理方法、装置和可移动平台 WO2021081816A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201980031650.3A CN112119413A (zh) 2019-10-30 2019-10-30 一种数据处理方法、装置和可移动平台
PCT/CN2019/114391 WO2021081816A1 (zh) 2019-10-30 2019-10-30 一种数据处理方法、装置和可移动平台

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/114391 WO2021081816A1 (zh) 2019-10-30 2019-10-30 一种数据处理方法、装置和可移动平台

Publications (1)

Publication Number Publication Date
WO2021081816A1 true WO2021081816A1 (zh) 2021-05-06

Family

ID=73798828

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/114391 WO2021081816A1 (zh) 2019-10-30 2019-10-30 一种数据处理方法、装置和可移动平台

Country Status (2)

Country Link
CN (1) CN112119413A (zh)
WO (1) WO2021081816A1 (zh)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6502061B1 (en) * 1999-03-04 2002-12-31 Sony Corporation Method and device for information processing and program providing medium
CN104331623A (zh) * 2014-11-06 2015-02-04 西北工业大学 一种机动策略自适应的目标跟踪信息滤波算法
CN106093951A (zh) * 2016-06-06 2016-11-09 清华大学 基于超声波传感器阵列的物体跟踪方法
CN106774436A (zh) * 2017-02-27 2017-05-31 南京航空航天大学 基于视觉的旋翼无人机稳定跟踪目标的控制系统及方法
CN108074234A (zh) * 2017-12-22 2018-05-25 湖南源信光电科技股份有限公司 一种基于目标跟踪和多特征融合的大空间火焰检测方法
CN110146909A (zh) * 2018-09-06 2019-08-20 腾讯科技（深圳）有限公司 一种定位数据处理方法

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109284677B (zh) * 2018-08-16 2022-06-03 昆明理工大学 一种贝叶斯滤波目标跟踪算法
CN109738884B (zh) * 2018-12-29 2022-03-11 百度在线网络技术（北京）有限公司 对象检测方法、装置和计算机设备

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6502061B1 (en) * 1999-03-04 2002-12-31 Sony Corporation Method and device for information processing and program providing medium
CN104331623A (zh) * 2014-11-06 2015-02-04 西北工业大学 一种机动策略自适应的目标跟踪信息滤波算法
CN106093951A (zh) * 2016-06-06 2016-11-09 清华大学 基于超声波传感器阵列的物体跟踪方法
CN106774436A (zh) * 2017-02-27 2017-05-31 南京航空航天大学 基于视觉的旋翼无人机稳定跟踪目标的控制系统及方法
CN108074234A (zh) * 2017-12-22 2018-05-25 湖南源信光电科技股份有限公司 一种基于目标跟踪和多特征融合的大空间火焰检测方法
CN110146909A (zh) * 2018-09-06 2019-08-20 腾讯科技（深圳）有限公司 一种定位数据处理方法

Also Published As

Publication number Publication date
CN112119413A (zh) 2020-12-22

Similar Documents

Publication Publication Date Title
JP7440013B2 (ja) 車両環境マッピング方法、ならびに、対応するシステム、車両およびコンピュータプログラム
CN109212530B (zh) 用于确定障碍物速度的方法和装置
CN108291813B (zh) 借助姿态图的位置数据融合
CN110246182B (zh) 基于视觉的全局地图定位方法、装置、存储介质和设备
US11932274B2 (en) Electronic device and control method therefor
US20130238181A1 (en) On-board vehicle path prediction using processed sensor information
US20180126984A1 (en) Object tracking using sensor fusion within a probabilistic framework
CN108897836B (zh) 一种机器人基于语义进行地图构建的方法和装置
CN110503009B (zh) 车道线跟踪方法及相关产品
CN108509970B (zh) 学习方法、对应的系统、设备和计算机程序产品
JP7413543B2 (ja) データ伝送方法および装置
KR102168288B1 (ko) 다중 라이다를 이용한 다중 물체 추적 시스템 및 방법
CN112784060B (zh) 基于车联网的路面摩擦系数确定方法、装置和电子设备
CN110853085B (zh) 基于语义slam的建图方法和装置及电子设备
CN108885719B (zh) 基于立体视觉的随机地图生成和贝叶斯更新
US9747801B2 (en) Method and device for determining surroundings
CN112883871B (zh) 一种模型训练以及确定无人车运动策略方法及装置
CN113763425A (zh) 道路区域标定方法及电子设备
WO2021081816A1 (zh) 一种数据处理方法、装置和可移动平台
JP2020042496A (ja) 電子制御装置、ニューラルネットワーク更新システム
CN112749727A (zh) 本地服务器、影像辨识系统及其更新方法
CN111814114B (zh) 车道定位校验的方法、设备、电子设备、车辆及存储介质
JP2019207607A (ja) 移動体追跡装置
JP7031625B2 (ja) グラフ構造推定装置、ｓｌａｍ装置、グラフ構造推定方法、コンピュータプログラム、および記憶媒体
CN110244710B (zh) 自动寻迹方法、装置、存储介质及电子设备

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: 19950561

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: 19950561

Country of ref document: EP

Kind code of ref document: A1