Control device for a mobile body, method for controlling a mobile body, and program

Abstract

This control device for a mobile body, which recognizes the periphery of the mobile body and moves, comprises: an object recognition unit that recognizes an object present in the periphery of the mobile body on the basis of observation values of a sensor provided to the mobile body; and an object tracking unit that tracks the object recognized by the object recognition unit, on the basis of time-series data of the observation values. The object tracking unit determines whether or not the object is a tracking target, on the basis of signal intensities of the observation values and the dispersion of the observation values.

Description

Technical Field

[0001] The present invention relates to a control device for a moving body, a control method for a moving body, and a program.

Background Art

[0002] In recent years, efforts to achieve a low-carbon society or a decarbonized society have been activated, and in vehicles as well, research and development have been conducted on electric moving bodies that can move on both sidewalks and roads in order to reduce CO2 emissions and improve energy efficiency. For example, Patent Document 1 discloses a technique for considering the influence of multipath in an occupancy grid map in order to accurately estimate the surrounding environment of a moving body from observation data obtained using sensing equipment. Also, for example, Patent Document 2 discloses that when determining whether a target detected by a camera and an ultrasonic sonar is a control target based on the detection reliability of the sensor, the threshold value is set to a different value depending on whether the target is a fusion target or a non-fusion target.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the above sensor fusion, object tracking using a Kalman filter may be performed. However, when this is simply applied to the prior art, there are cases where the situation of misrecognizing noise as a tracking target continues, and the noise may continue to be tracked.

[0005] This invention has been made in consideration of these circumstances, and one of its objectives is to provide a control device for a moving body, a control method for a moving body, and a program that can suppress the continued misidentification of noise as the tracking target when tracking objects around the moving body. This will ultimately contribute to improving energy efficiency. [Means for solving the problem]

[0006] The control device for a mobile body, the control method for a mobile body, and the program according to this invention employ the following configuration. (1) A control device for a moving body according to one aspect of the present invention is a control device for a moving body that recognizes its surroundings and moves, comprising: an object recognition unit that recognizes objects present in the surroundings of the moving body based on observed values ​​from sensors provided on the moving body; and an object tracking unit that tracks the objects recognized by the object recognition unit based on time-series data of the observed values, wherein the object tracking unit determines whether or not the object is a target for tracking based on the signal strength of the observed values ​​and the variance of the observed values.

[0007] (2) In the embodiment of (1) above, the object tracking unit determines that the object is not a target for tracking if the reliability of the observed value based on the signal strength is below a predetermined standard and the variance of the observed value is greater than a threshold.

[0008] (3) In the embodiment of (1) or (2) above, the object tracking unit starts monitoring the variance of the observed values ​​when the reliability changes from a state higher than the standard to a state lower than the standard, and determines that the object is no longer a target for tracking when the variance of the observed values ​​becomes greater than a threshold.

[0009] (4): In any embodiment of (1) to (3) above, the object tracking unit determines that the reliability is recognized based on the error between the signal intensity of the observed value and the estimated result of the signal intensity of the observed value, and that the object is not a target for tracking if the error is greater than a first threshold and the variance is greater than a second threshold.

[0010] (5) In any of the embodiments described in (1) to (4) above, the object tracking unit learns the signal intensity of the observed value using an autoencoder that compresses and then restores the signal intensity of the observed value, and compares the restoration error between the signal intensity before compression by the autoencoder and the signal intensity after restoration with the first threshold.

[0011] (6) In the embodiments of (1) to (5) above, the object tracking unit tracks the object by applying the observed values ​​to a Kalman filter, and initializes the Kalman filter when it is determined that the object is not the object to be tracked.

[0012] (7): In the embodiments of (1) to (6) above, the object tracking unit performs a gating process when the variance of the observed values ​​is greater than the third threshold, and performs an update process of the Kalman filter when there are observed values ​​that have passed the gating process. 。

[0013] (8) A method for controlling a moving body according to another aspect of the present invention is a method for controlling a moving body that moves while recognizing its surroundings, wherein the control device of the moving body performs an object recognition process that recognizes an object present in the surroundings of the moving body based on observed values ​​from a sensor provided on the moving body, and an object tracking process that tracks the object recognized by the object recognition process based on time-series data of the observed values, wherein in the object tracking process, it is determined whether or not the object is the object to be tracked based on the signal strength of the observed values ​​and the variance of the observed values.

[0014] (9): A program according to another aspect of the present invention causes a control device of a moving body that recognizes its surroundings and moves to perform an object recognition process that recognizes objects present in the vicinity of the moving body based on observed values ​​from sensors provided on the moving body, and an object tracking process that tracks the objects recognized in the object recognition process based on time-series data of the observed values, wherein the program causes the control device of the moving body to determine whether or not the object is a target for tracking based on the signal strength of the observed values ​​and the variance of the observed values. [Effects of the Invention]

[0015] According to embodiments (1) to (9), it is possible to suppress the continued misidentification of noise as the tracking target when tracking objects around a moving object. [Brief explanation of the drawing]

[0016] [Figure 1] This figure shows an example of the configuration of a mobile body and control device according to the embodiment. [Figure 2] This is a perspective view of a moving object seen from above. [Figure 3] This flowchart shows an example of the process flow in which a control device tracks dynamic objects around a moving object. [Figure 4] This diagram illustrates the gating process. [Figure 5] This diagram illustrates the general principles of machine learning for signal strength using an autoencoder. [Figure 6] This diagram shows an example of a situation in which the object tracking unit determines that an object has been lost. [Modes for carrying out the invention]

[0017] Hereinafter, with reference to the drawings, embodiments of a control device for a moving body, a control method for a moving body, and a program will be described. The moving body may move in both a roadway and a predetermined area different from the roadway. The moving body may be referred to as micromobility. An electric kick scooter is a type of micromobility. Also, the moving body may be a vehicle on which a passenger can board, or may be an autonomous moving body capable of autonomous driving without a driver. The latter autonomous moving body is used, for example, for transporting luggage and the like. The predetermined area is, for example, an indoor space. Also, the predetermined area may be part or all of a roadside strip, a bicycle lane, an open space, etc., or may include all of a sidewalk, a roadside strip, a bicycle lane, an open space, etc., The control device may remotely control the operation of the moving body by transmitting an instruction by communication from a position separated from the moving body.

[0018] FIG. 1 is a diagram showing an example of the configuration of a moving body 1 and a control device 100 according to an embodiment. In the moving body 1, for example, an external detection device 10, a moving body sensor 12, an operator 14, an internal camera 16, a positioning device 18, an acceleration sensor 20, a moving mechanism 30, a drive device 40, an external notification device 50, a storage device 70, and a control device 100 are mounted. Note that some of these configurations that are not essential for realizing the functions of the present invention may be omitted.

[0019] The external detection device 10 is various devices having the traveling direction of the moving body 1 as a detection range. The external detection device 10 includes an external camera, a radar device, LIDAR (Light Detection and Ranging), a sensor fusion device, and the like. The external detection device 10 outputs information (image, position of an object, etc.) indicating the detection result to the control device 100. The external detection device 10 is not limited to a specific one as long as it can detect the position of an object existing in the detection range, but in the present embodiment, as an example, the external detection device 10 outputs information on waves reflected by the detection target as an observation value. For example, the external detection device 10 is a 3D sonar that detects an object by sound waves.

[0020] The mobile sensor 12 includes, for example, a speed sensor, a yaw rate (angular velocity) sensor, a direction sensor, and an operation amount detection sensor attached to the operator 14. The operator 14 includes, for example, an operator for instructing acceleration and deceleration (such as an accelerator pedal or a brake pedal) and an operator for instructing steering (such as a steering wheel). In this case, the mobile sensor 12 may include an accelerator opening sensor, a brake depression amount sensor, a steering torque sensor, etc. The mobile body 1 may include an operator in a form other than the above as the operator 14 (for example, a non-annular rotary operator, a joystick, a button, etc.).

[0021] The internal camera 16 captures at least the head of the passenger of the mobile body 1 from the front. The internal camera 16 is a digital camera using an image pickup device such as a CCD (Charge Coupled Device) or a CMOS (Complementary Metal Oxide Semiconductor). The internal camera 16 outputs the captured image to the control device 100.

[0022] The positioning device 18 is a device for positioning the position of the mobile body 1. The positioning device 18 is, for example, a GNSS (Global Navigation Satellite System) receiver, and based on the signal received from the GNSS satellite, it identifies the position of the mobile body 1 and outputs it as position information. Note that the position information of the mobile body 1 may be estimated from the position of a Wi-Fi (registered trademark, the same applies hereinafter) base station to which the communication device described later is connected.

[0023] The acceleration sensor 20 detects the acceleration of the mobile body 1 and outputs a signal corresponding to the detected acceleration to the control device 100. The acceleration sensor 20 detects the acceleration acting in the vertical direction (height direction) in addition to the horizontal direction of the mobile body 1.

[0024] The moving mechanism 30 is a mechanism for moving the mobile body 1. The moving mechanism 30 is, for example, a wheel group including a steering wheel and drive wheels. Also, the moving mechanism 30 may be legs for multi-legged walking.

[0025] The drive unit 40 outputs force to the moving mechanism 30 to move the moving body 1. For example, the drive unit 40 includes a motor that drives the drive wheels, a battery that stores the power supplied to the motor, and a steering device that adjusts the steering angle of the steering wheels. The drive unit 40 may also be equipped with an internal combustion engine or a fuel cell as a means of outputting driving force or generating power. Furthermore, the drive unit 40 may also be equipped with a braking device that uses frictional force or air resistance.

[0026] The external notification device 50 is, for example, a lamp, display device, speaker, etc., provided on the outer panel of the mobile body 1 to notify information to the outside of the mobile body 1. For example, the external notification device 50 may perform different external notification operations depending on the state of the mobile body 1 while it is moving.

[0027] Figure 2 is a perspective view of the mobile body 1 from above. In the figure, FW is the steering wheel, RW is the drive wheel, SD is the steering mechanism, MT is the motor, and BT is the battery. The steering mechanism SD, motor MT, and battery BT are included in the drive system 40. AP is the accelerator pedal, BP is the brake pedal, WH is the steering wheel, SP is the speaker, and MC is the microphone. The mobile body 1 shown is a single-seater, and the occupant P is seated in the driver's seat DS wearing a seat belt SB. Arrow D1 is the direction of travel (velocity vector) of the mobile body 1. The external environment detection device 10 is located near the front end of the mobile body 1, and the internal camera 16 is located in front of the occupant P, in a position that allows it to capture images of the occupant P's head. An external notification device 50, which serves as a display device, is also located near the front end of the mobile body 1.

[0028] Returning to Figure 1, the storage device 70 is a non-transient storage device such as an HDD (Hard Disk Drive), flash memory, or RAM (Random Access Memory). The storage device 70 stores map information 72, a program 74 executed by the control device 100, and other data. In the figure, the storage device 70 is shown outside the frame of the control device 100, but the storage device 70 may be included within the control device 100.

[0029] [Control device] The control device 100 includes, for example, an object recognition unit 120, an object tracking unit 130, and a control unit 140. For example, it is realized by a hardware processor such as a CPU (Central Processing Unit) executing a program (software) 74. Some or all of these components may be realized by hardware (including circuitry) such as an LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), or GPU (Graphics Processing Unit), or by the cooperation of software and hardware. The program may be stored in the storage device 70 in advance, or it may be stored on a removable storage medium (non-transient storage medium) such as a DVD or CD-ROM and installed in the storage device 70 when the storage medium is mounted on a drive device.

[0030] The object recognition unit 120 recognizes objects present around the moving body 1 based on the output of the external environment detection device 10. Objects include some or all of the following: moving objects such as vehicles, bicycles, and pedestrians; road boundaries such as road markings, steps, guardrails, shoulders, and median strips; structures installed on the road such as road signs and billboards; and obstacles such as fallen objects present (lying on) the road. Furthermore, the objects recognized by the object recognition unit 120 include both static objects that do not move from their location and dynamic objects such as people and other moving objects. In addition to recognizing the presence of an object, the object recognition unit 120 may also recognize various states of the object (such as its position, speed, and direction of movement).

[0031] The object recognition unit 120 recognizes objects (such as other moving objects or obstacles) around the moving object 1 (mainly in the direction of travel) by inputting observed sound wave values ​​acquired by the 3D sonar of the external detection device 10 into the object recognition model. Here, the object recognition model is a trained model, for example, trained using a machine learning algorithm, which has been trained to take observed sound wave values ​​acquired by the 3D sonar as input and output information such as the presence, position, and type of other moving objects. The type of other moving object may be estimated based on the size in the image or the intensity of reflected waves received by the radar device of the external detection device 10, or it may be estimated based on an image captured by the external camera of the external detection device 10. The object recognition unit 120 may also perform object recognition by combining these estimation results. Furthermore, the object recognition unit 120 may acquire the speed of other moving objects detected by the radar device using, for example, Doppler shift.

[0032] The object tracking unit 130 performs tracking processing on dynamic objects among the objects recognized by the object recognition unit 120 using a Kalman filter. By using a Kalman filter, the object tracking unit 130 can estimate the position of a time-changing dynamic object from the observed values ​​(assuming errors are included) of the external detection device 10. More specifically, the object tracking unit 130 estimates the current state of the tracking dynamic object based on the result of predicting the current state from past states and the current observed values.

[0033] The control unit 140 controls the drive unit 40 according to the set driving mode, for example. The driving mode is set to one of several modes corresponding to the entity performing steering and acceleration / deceleration. For example, the driving modes include a first driving assistance mode in which the occupant performs steering operations and acceleration / deceleration control is performed automatically, a second driving assistance mode in which the occupant performs acceleration / deceleration operations and steering control is performed automatically, a manual driving mode in which the occupant performs steering operations and acceleration / deceleration operations, and rudderThis includes an automatic driving mode in which control and acceleration / deceleration control are performed automatically. The driving mode can be changed by the occupant or by the control unit 140, and the mobile unit 1 may be equipped with a mechanical switch or a GUI (Graphical User Interface) switch set on a touch panel to accept operation to switch the driving mode.

[0034] For example, the control unit 140 sets a movement path for the mobile body 1 that avoids collisions with objects (static and dynamic objects) recognized by the object recognition unit 120, and controls the movement mechanism 30 and drive unit 40 of the mobile body 1 to move along the set movement path. The movement path may be determined as appropriate depending on the driving mode, the set destination, etc.

[0035] Figure 3 is a flowchart illustrating an example of the processing flow in which the control device 100 tracks dynamic objects around the moving body 1. Here, we will describe the processing for one cycle of the process that is repeatedly executed in the tracking process. In other words, in the actual tracking process, the processing flow in Figure 3 is repeatedly executed. First, in the control device 100, the object tracking unit 130 determines whether or not the Kalman filter initialization flag is set (S101). As described above, the processing flow in Figure 3 is repeatedly executed, so if it is the first time that S101 is executed, the object tracking unit 130 will determine that the initialization flag is at its initial value (for example, a value indicating that the flag is not set), and thereafter, it will determine the result of setting the initialization flag according to the previous processing flow.

[0036] If the initialization flag is determined to be set, the object tracking unit 130 initializes the Kalman filter (S102), obtains the current observed value from the external detection device 10 (S103), and performs prediction processing using the Kalman filter with the obtained observed value (S104). On the other hand, if the initialization flag is determined not to be set in S101, the object tracking unit 130 skips the initialization process in S102 and proceeds to S103.

[0037] Initializing the Kalman filter in S102 means returning the Kalman gain to its initial value. The initial value of the Kalman gain may be set appropriately based on the results of tests conducted in advance. The prediction process of the Kalman filter in S104 is a process that predicts the future state of the tracked object from the current state estimated for the tracked object. For example, the prediction process of the Kalman filter incorporates a motion model of the tracked object, and the object tracking unit 130 can obtain a predicted value of the tracked object's state in the future (next cycle) by inputting the estimation result of the current state into the Kalman filter.

[0038] Next, the object tracking unit 130 calculates the variance of the observed values ​​of the object being tracked and determines whether the value of the variance is greater than a predetermined threshold Vth1 (S105). If it determines that the variance is less than or equal to the threshold Vth1, the object tracking unit 130 sets an initialization flag (S106) and terminates the series of processing flows. In this case, the observation of the object being tracked is considered stable, and the reliability of the state prediction result by the Kalman filter is considered high. By terminating after setting the initialization flag, the Kalman filter can be initialized at the start of the next periodic processing. The variance threshold Vth1 here is an example of a "third threshold".

[0039] On the other hand, if in S105 the variance value is determined to be greater than the threshold Vth1, the object tracking unit 130 performs a gating process on the observed values ​​(S107) and determines whether or not there are any observed values ​​that have passed the gating process (S108). The gating process is a process of extracting observed values ​​to be tracked from the acquired group of observed values. Figure 4 is a diagram illustrating the overview of the gating process. In the example in Figure 4, the group of observed values ​​P11 to P15 is plotted on an xy plane that provides an overhead view of the moving object 1 and its surroundings. "Observed values ​​that have passed the gating process" means observed values ​​that have been extracted as tracked values ​​in the gating process.

[0040] In this case, the object tracking unit 130 performs a gating process to extract observed values ​​P11 to P13 from the group of observed values ​​P11 to P15 that are within the field of view from the viewpoint position of the moving object 1 (the viewpoint position of the external detection device 10) and that are included in a predetermined detection range R1. The example in Figure 4 shows a case where the detection range R1 is defined as a rectangle with the forward direction of the moving object 1 as its longitudinal direction, but this is just one example, and the detection range R1 may be arbitrarily set according to the desired detection sensitivity, detection directionality, etc. By narrowing down the observed values ​​to be processed through such a gating process, it is possible to suppress the deterioration of real-time performance due to a large amount of processing.

[0041] Returning to Figure 3, the object tracking unit 130 terminates the processing flow if it determines in S108 that there are no observed values ​​that have passed the gating process. This is because the absence of observed values ​​that have passed the gating process means that there are no observed values ​​for the tracking target (processing target). On the other hand, if it determines in S108 that there are observed values ​​that have passed the gating process, the object tracking unit 130 performs the Kalman filter update process (S109).

[0042] As described above, the Kalman filter estimates the true state of the tracked object in the current period based on the observed state and Kalman gain of the tracked object in the current period, and the predicted state of the tracked object in the current period based on the state estimation results of the previous period. More specifically, the Kalman filter takes observed values ​​and observation errors as input for each period, updates the Kalman gain based on the observation error and the estimation error for the current period predicted in the filter calculation of the previous period (the uncertainty of the estimation, hereinafter referred to as the "predicted estimation error"), and estimates the true value of the state value in the current period based on the updated Kalman gain and the state value for the current period predicted in the filter calculation of the previous period (hereinafter referred to as the "predicted state value"). The Kalman filter also calculates the estimation error of the true value and outputs the estimated value and estimation error as the estimation result for the current period. The update process in S109 includes this Kalman gain update process and the process of estimating the state of the tracked object using the updated Kalman gain.

[0043] The Kalman filter calculates the state value (predicted state value) and estimation error (prediction estimation error) for the next period based on the state estimation results for the current period, and uses the prediction result as input for the update process in the next period.

[0044] Next, the object tracking unit 130 learns the signal strength using machine learning based on the observed values ​​of the current period (S110). Here, we will explain assuming an autoencoder as the machine learning algorithm, but the algorithm used is not limited to this. An autoencoder is one of the unsupervised machine learning methods that uses a neural network and performs unsupervised learning with the aim of outputting data that matches the input data.

[0045] Figure 5 illustrates the general outline of machine learning for signal intensity using an autoencoder. The autoencoder is configured to compress the input data to reduce its dimensionality, then restore the dimensionality to its original value before outputting. This allows for the extraction of important information for data reconstruction and the formation of a network that efficiently generates the original data using the extracted results. Furthermore, as mentioned above, the autoencoder performs unsupervised learning, so it can learn signal intensity online by cumulatively inputting the signal intensity of the observed data for each period.

[0046] Alternatively, instead of using an autoencoder, a method that directly compares the signal strength to a threshold to detect a decrease in signal strength reliability is also possible. However, in this case, the threshold must be set in advance using a rule-based system, and it is difficult and costly to set rules that take into account all possible situations regarding signal strength beforehand. In contrast, when using an autoencoder, the restoration error between the autoencoder's input and output is compared to a threshold. If the restoration error is greater than the threshold, it is considered abnormal, and if it is less than or equal to the threshold, it is considered normal. This allows for easier detection of a decrease in signal strength reliability in a way that can handle a wider range of situations.

[0047] Returning to Figure 3, the object tracking unit 130 then inputs the signal intensity of the observed value for the current period into the autoencoder learned in S110, thereby obtaining the signal intensity as its output, which has been compressed internally by the autoencoder and then restored. The object tracking unit 130 calculates the restoration error of the output signal intensity relative to the input signal intensity (S111) and determines whether the restoration error is greater than a predetermined threshold Eth (S112). If it is determined that the restoration error is less than or equal to the threshold Eth, the object tracking unit 130 terminates the series of processing steps. This is based on the understanding that if noise is mistakenly detected as the tracking target, the variation in signal intensity increases, and the restoration error tends to increase. The threshold Eth of the restoration error here is an example of a "first threshold".

[0048] On the other hand, in S112, if it is determined that the reconstruction error is greater than the threshold Eth, the object tracking unit 130 calculates the variance of the observed values ​​acquired during the period from the present to a predetermined point in the past, and determines whether the variance is greater than a predetermined threshold Vth2 (S113). If it is determined that the variance is less than or equal to the threshold Vth2, the object tracking unit 130 terminates the series of processing flows. This is based on the understanding that even if the variation in signal intensity is large, if the variation in observed values ​​is small, it is highly likely that it is not noise. The variance threshold Vth2 here is an example of a "second threshold".

[0049] On the other hand, in S113, if the object tracking unit 130 determines that the variance of the observed values ​​is greater than the threshold Vth2, it determines that the object related to the observed value is noise (S114: Lost determination). In this case, the object tracking unit 130 sets an initialization flag (S115) and terminates the series of processing flows. By setting the initialization flag here, the Kalman filter is initialized at the beginning of the next cycle, and thereafter, when a tracking target is generated, the prediction process and update process are repeatedly executed, so that the Kalman filter is updated as needed.

[0050] Furthermore, in the processing flow described above, the object tracking unit 130 can track multiple objects (including noise). That is, the object tracking unit 130 can manage the Kalman filter and initialization flag for each object being tracked. The object tracking unit 130 can also learn the signal intensity of the observed values ​​for each object being tracked. In addition, the object tracking unit 130 can determine the reconstruction error and variance for each object being tracked.

[0051] Figure 6 shows an example of the situation in which the object tracking unit 130 performs lost object detection. In Figure 6, the upper graph represents the monitoring status of signal strength, and the lower graph represents the monitoring status of the variance of observed values. In this example, the object tracking unit 130 sequentially learns and monitors the signal strength of the observed values ​​input in a time series, and for the period from time t0 to t1, it recognizes that the signal strength of the observed values ​​is the signal strength when a person is detected. This is recognized by determining that the signal strength recovery error is below the threshold Eth, as described above, which means that the reliability of the observed values ​​is higher than a predetermined standard.

[0052] Next, at time t2, the object tracking unit 130 detects that the signal intensity of the observed value has reached the signal intensity that would occur if noise were detected. As described above, this is recognized by determining that the signal intensity restoration error is greater than the threshold Eth, which means that the reliability of the observed value is below a predetermined standard. From this change in signal intensity, the object tracking unit 130 determines that the reliability of the observed value may have decreased and starts monitoring the variance of the observed value from this point t2.

[0053] The example in Figure 6 illustrates a situation where the tracking target becomes noise around time t2, when the signal strength drops significantly, and the variance of the observed values ​​gradually increases. In this example, the object tracking unit 130 starts variance monitoring at time t2, and then at time t3 detects that the variance has become greater than the loss detection threshold Vth2, and performs a loss detection.

[0054] According to the embodiment described above, by monitoring changes in the signal intensity of the observed values, it is possible to accurately detect when the likelihood of noise being the target of tracking increases, and then perform a loss determination based on the variance of the observed values. Therefore, it is possible to quickly detect situations in which noise has been mistakenly identified as the target of tracking and to suppress the continuation of such situations.

[0055] Furthermore, according to this embodiment, distributed monitoring is started only after it is detected that there is a high probability that noise is being tracked. Therefore, if the processing load of distributed monitoring is high, the hardware resources required for distributed monitoring can be reduced, thereby lowering the processing load of the control device 100.

[0056] The embodiments described above can be expressed as follows. The control device for the moving object, which recognizes the surroundings of the moving object and moves accordingly, A storage medium that stores computer-readable instructions, A processor connected to the storage medium, The processor executing the computer-readable instructions to: Object recognition processing that recognizes objects present around the moving object based on observation values ​​from sensors provided by the moving object, An object tracking process that tracks the object recognized in the object recognition process based on the time-series data of the observed values, Execute, In the object tracking process, it is determined whether or not the object is the object to be tracked based on the signal intensity of the observed value and the variance of the observed value. A control device for mobile vehicles.

[0057] Although embodiments for carrying out the present invention have been described above using examples, the present invention is not limited in any way to these embodiments, and various modifications and substitutions can be made without departing from the spirit of the present invention. [Explanation of Symbols]

[0058] 1...Moving body, 10...External environment detection device, 12...Moving body sensor, 14...Operator, 16...Internal camera, 18...Positioning device, 20...Accelerometer, 30...Movement mechanism, 40...Drive device, 50...External notification device, 70...Storage device, 72...Map information, 74...Program, 100...Control device, 120...Object recognition unit, 130...Object tracking unit, 140...Control unit

Claims

1. A control device for a moving object that recognizes its surroundings and moves accordingly, An object recognition unit that recognizes objects present around the moving object based on observation values ​​from sensors provided by the moving object, An object tracking unit tracks the object recognized by the object recognition unit based on the time-series data of the observed values, Equipped with, The object tracking unit determines whether the object is a target for tracking based on the signal strength of the observed value and the variance of the observed value. When the reliability of the observed value based on the signal strength changes from a state higher than a predetermined standard to a state lower than the standard, it starts monitoring the variance of the observed value, and when the variance of the observed value becomes greater than the threshold, it determines that the object is no longer a target for tracking. A control device for mobile vehicles.

2. The object tracking unit determines that the reliability is recognized based on the error between the signal intensity of the observed value and the estimated result of the signal intensity of the observed value, and determines that the object is not a target for tracking if the error is greater than a first threshold and the variance is greater than a second threshold. A control device for a mobile body according to claim 1.

3. The object tracking unit learns the signal intensity of the observed value using an autoencoder that compresses and then restores the signal intensity of the observed value, and compares the restoration error between the signal intensity before compression by the autoencoder and the signal intensity after restoration with the first threshold. A control device for a mobile body according to claim 2.

4. The object tracking unit tracks the object by applying the observed values ​​to a Kalman filter, and initializes the Kalman filter when it determines that the object is not the object to be tracked. A control device for a mobile body according to claim 1.

5. The object tracking unit performs a gating process when the variance of the observed values ​​is greater than a third threshold, and performs an update process for the Kalman filter when there are observed values ​​that have passed the gating process. A control device for a mobile body according to claim 4.

6. A method for controlling a moving object by recognizing its surroundings, The control device of the mobile body Object recognition processing that recognizes objects present around the moving object based on observation values ​​from sensors provided by the moving object, An object tracking process that tracks the object recognized in the object recognition process based on the time-series data of the observed values, This is what does the following: In the object tracking process, it is determined whether or not the object is a target for tracking based on the signal strength of the observed value and the variance of the observed value. When the reliability of the observed value based on the signal strength changes from a state higher than a predetermined standard to a state lower than the standard, monitoring of the variance of the observed value is started, and when the variance of the observed value becomes greater than the threshold, it is determined that the object is no longer a target for tracking. A method for controlling a moving object.

7. The control device for the moving object, which recognizes its surroundings and moves accordingly, Object recognition processing that recognizes objects present around the moving object based on observation values ​​from sensors provided by the moving object, An object tracking process that tracks the object recognized in the object recognition process based on the time-series data of the observed values, This will cause it to execute, In the object tracking process, it is determined whether or not the object is a target for tracking based on the signal strength of the observed value and the variance of the observed value. When the reliability of the observed value based on the signal strength changes from a state higher than a predetermined standard to a state lower than the standard, monitoring of the variance of the observed value is started, and when the variance of the observed value becomes greater than the threshold, it is determined that the object is no longer a target for tracking. A program for that purpose.

Images