Self-position estimation method, information processing device, and program

The self-position estimation method employs a particle filter to combine sensor data from multiple directions and adjust sampling periods, enhancing accuracy and reducing positional deviations in featureless environments by prioritizing corrections based on sensor reliability.

WO2024232249A1PCT designated stage expired Publication Date: 2024-11-14SONY GROUP CORP

Patent Information

Application Number
PCT/JP2024/015672
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-09
Filing Date
2024-04-22
Publication Date
2024-11-14

AI Technical Summary

Technical Problem

Conventional self-position estimation methods for mobile objects in environments with few changes in feature values tend to deviate from the actual position, leading to accumulated errors and increased risk of collision, especially in featureless paths.

Method used

A self-position estimation method using a particle filter that combines sensor data from a first ranging sensor sensing the front and a second ranging sensor sensing the side of the mobile body, with a weighting coefficient set based on the variance in the traveling direction and orthogonal direction, to determine the reliability of sensor data and adjust the sampling period of the sensors.

Benefits of technology

This approach improves the accuracy of self-position estimation by maintaining the priority of corrections based on the reliability of sensor data, reducing the accumulation of errors and preventing deviations from the actual position, even in environments with minimal feature changes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2024015672_14112024_PF_FP_ABST
    Figure JP2024015672_14112024_PF_FP_ABST
Patent Text Reader

Abstract

The present disclosure relates to a self-position estimation method, an information processing device, and a program that make it possible to achieve an increase in accuracy of self-position estimation. The present disclosure includes: executing self-position estimation of a mobile body using a particle filter on the basis of sensor data from a first distance measuring sensor for sensing a distance in the forward direction of the mobile body and a second distance measuring sensor for sensing a distance in the side direction of the mobile body; setting a weight coefficient related to the reliability of the sensor data from the first distance measuring sensor and the second distance measuring sensor on the basis of variance of the traveling direction of the mobile body in the distribution in presence of the estimated self-position of the mobile body and variance of an orthogonal direction orthogonal to the traveling direction; and setting a sampling period of the first distance measuring sensor and the second distance measuring sensor according to the set weight coefficient. The present disclosure is applicable to a mobile body that moves autonomously.
Need to check novelty before this filing date? Find Prior Art

Description

Self-position estimation method, information processing device, and program

[0001] The present disclosure relates to a self-location estimation method, an information processing device, and a program, and more particularly to a self-location estimation method, an information processing device, and a program that enable improvement in the accuracy of estimating a self-location.

[0002] 2. Description of the Related Art There is known a technique for a mobile object moving within a predetermined moving area to estimate its own position within the moving area.

[0003] For example, Patent Document 1 discloses a self-position estimation device that, when estimating the final self-position of a moving body moving within a specified movement area for each step, adds one step's worth of movement information obtained by odometry to the final self-position in the previous step.

[0004] JP 2016-224680 A

[0005] Conventionally, when a mobile object moves in an environment where there is little change in feature quantities, the self-position recognized by the mobile object may deviate from the actual position.

[0006] The present disclosure has been made in consideration of such circumstances, and aims to improve the accuracy of estimating a self-position.

[0007] The self-location estimation method disclosed herein is a self-location estimation method that performs self-location estimation of a moving body using a particle filter based on sensor data from a first ranging sensor that senses the front of the moving body and a second ranging sensor that senses the side of the moving body, sets a weighting coefficient related to the reliability of the sensor data from the first ranging sensor and the second ranging sensor based on the variance in the direction of travel of the moving body and the variance in an orthogonal direction perpendicular to the direction of travel in the existence distribution of the estimated self-location of the moving body, and sets the sampling period of the first ranging sensor and the second ranging sensor according to the set weighting coefficient.

[0008] The information processing device of the present disclosure is an information processing device that includes a self-position estimation unit that performs self-position estimation of the moving body using a particle filter based on sensor data from a first ranging sensor that senses the front of the moving body and a second ranging sensor that senses the side of the moving body, a weight coefficient setting unit that sets a weight coefficient related to the reliability of the sensor data from the first ranging sensor and the second ranging sensor based on the variance in the moving body's direction of travel and the variance in an orthogonal direction perpendicular to the moving direction in the existence distribution of the estimated self-position of the moving body, and a sampling period setting unit that sets the sampling period of the first ranging sensor and the second ranging sensor in accordance with the set weight coefficient.

[0009] The program disclosed herein is a program for causing a computer to perform a process of estimating the self-position of a moving body using a particle filter based on sensor data from a first ranging sensor that senses the front of the moving body and a second ranging sensor that senses the side of the moving body, setting a weighting coefficient related to the reliability of the sensor data from the first ranging sensor and the second ranging sensor based on the variance in the direction of travel of the moving body and the variance in an orthogonal direction perpendicular to the direction of travel in the existence distribution of the estimated self-position of the moving body, and setting the sampling period of the first ranging sensor and the second ranging sensor according to the set weighting coefficient.

[0010] In the present disclosure, a particle filter is used to estimate the self-position of a moving body based on sensor data from a first ranging sensor that senses the front of the moving body and a second ranging sensor that senses the side of the moving body, and a weighting coefficient related to the reliability of the sensor data from the first ranging sensor and the second ranging sensor is set based on the variance in the direction of travel of the moving body and the variance in an orthogonal direction perpendicular to the direction of travel in the existence distribution of the estimated self-position of the moving body, and the sampling period of the first ranging sensor and the second ranging sensor is set according to the set weighting coefficient.

[0011] FIG. 1 is a block diagram showing an example of the configuration of a moving body using conventional technology; FIG. 2 is a flowchart explaining the flow of self-position estimation processing; FIG. 3 is a diagram showing an example of a moving body moving in an environment with little change in feature amount; FIG. 4 is a diagram explaining accumulated self-position error; FIG. 4 is a block diagram showing an example of the configuration of a moving body to which the technology according to the present disclosure is applied; FIG. 5 is a flowchart explaining the flow of self-position estimation processing; FIG. 5 is a block diagram showing another example of the configuration of a moving body; FIG. 6 is a block diagram showing yet another example of the configuration of a moving body; FIG. 7 is a block diagram showing an example of the configuration of computer hardware.

[0012] Modes for carrying out the present disclosure (hereinafter referred to as embodiments) will be described below in the following order.

[0013] 1. Configuration and operation of a mobile body using conventional technology 2. Problems with conventional technology 3. Configuration and operation of a mobile body to which the technology according to the present disclosure is applied 4. Other configurations of a mobile body 5. Example configurations of computer hardware

[0014] 1. Configuration and Operation of a Mobile Body Using Conventional Technology (Configuration of Mobile Body) FIG. 1 is a block diagram showing an example of the configuration of a mobile body 1 using conventional technology.

[0015] The mobile body 1 is configured to enable autonomous movement by dead reckoning (also called autonomous navigation), which estimates its own position from the route traveled, the distance traveled, etc. In dead reckoning, errors accumulate with each repetition of processing, so the mobile body 1 can maintain the accuracy of its own position estimation by periodically correcting the errors accumulated by dead reckoning through localization using feature amounts obtained by sensing the surrounding environment.

[0016] The mobile body 1 may be a transport robot or a serving robot that can move autonomously, an inspection robot that inspects the inside of pipes, or a vehicle that can travel autonomously. However, the mobile body 1 is not limited to these, and may be a drone that flies autonomously within a space.

[0017] 1, the mobile object 1 includes a wheel encoder 10, an IMU (Inertial Measurement Unit) 20, a distance measurement sensor 30, and a processor 50. Although not shown, the mobile object 1 also includes a motor, wheels, propellers, and the like as a movement mechanism for movement.

[0018] The wheel encoder 10 is one of the internal sensors that senses the internal state of the mobile object 1, and detects the rotation direction and rotation angle of the wheels of the mobile object 1 and supplies the detected rotation direction and rotation angle to the processor 50. If the mobile object 1 does not have wheels, the wheel encoder 10 does not need to be provided.

[0019] The IMU 20 is one of the internal sensors that senses the internal state of the moving body 1, and by detecting the inertial motion of the moving body 1, supplies the angular velocity (Yaw, Pitch, Roll) around the three axes of the aircraft (moving body 1) and the acceleration in the three axial directions (Ax, Ay, Az) to the processor 50.

[0020] The ranging sensor 30 senses the surroundings (particularly the front) of the moving body 1, generates sensor data (point cloud data) representing the distance to surrounding objects and their shapes, and supplies the sensor data to the processor 50. The ranging sensor 30 may be configured by a LiDAR (Light Detection And Ranging) sensor or a stereo camera.

[0021] The processor 50 performs self-position estimation of the moving body 1 based on the sensor data from the wheel encoders 10 , the IMU 20 , and the distance measurement sensor 30 .

[0022] The processor 50 is configured as an information processing device that operates by executing a predetermined program, for example. The processor 50 realizes, as functional blocks, a sensor data acquisition unit 51, a self-position estimation unit 52, a weighting coefficient setting unit 53, and a sampling period setting unit 54.

[0023] The sensor data acquisition unit 51 acquires sensor data from each of the wheel encoders 10, the IMU 20, and the distance measurement sensor 30, and supplies the data to the self-position estimation unit 52. The sensor data acquisition unit 51 converts the sensor data from each sensor into a format usable by the self-position estimation unit 52, as necessary. For example, the sensor data acquisition unit 51 converts the sensor data (point cloud data) from the distance measurement sensor 30 into angular velocities (Yaw, Pitch, Roll) about three axes of the mobile object 1 and accelerations (Ax, Ay, Az) in three axial directions, and supplies the converted data to the self-position estimation unit 52. The sensor data conversion may be performed by the distance measurement sensor 30.

[0024] The self-position estimation unit 52 estimates the self-position of the moving object 1 based on the sensor data from each sensor supplied from the sensor data acquisition unit 51 .

[0025] Specifically, the self-position estimator 52 performs self-position estimation of the moving body 1 using a Kalman filter based on the rotation direction and rotation angle of the wheel detected by the wheel encoder 10, thereby determining the self-position (x, y, z) of the moving body 1. The self-position estimator 52 also performs self-position estimation of the moving body 1 using a Kalman filter based on the angular velocity (Yaw, Pitch, Roll) about three axes and the acceleration (Ax, Ay, Az) in three axial directions of the moving body 1 detected by the IMU 20, thereby determining the self-position (x, y, z) of the moving body 1. Here, because the motion of the moving body 1 is nonlinear, an extended Kalman filter (EKF), which is applicable to nonlinear models, is used as the Kalman filter. In this manner, the self-position estimator 52 performs self-position estimation by dead reckoning using the EKF.

[0026] Furthermore, the self-location estimation unit 52 performs self-location estimation of the moving object 1 using a particle filter based on the sensor data generated by the ranging sensor 30, thereby determining the self-location (x, y, z) of the moving object 1 as well as the dispersion of particles in the distribution of the self-location. Here, adaptive Monte Carlo Localization (AMCL), which can dynamically adjust the number of particles, is adopted as the self-location estimation using the particle filter. In this way, the self-location estimation unit 52 performs self-location estimation using AMCL.

[0027] The self-location estimation unit 52 then integrates the self-location estimation results of the moving object 1 obtained by self-location estimation based on the sensor data from each sensor, and determines the final self-location of the moving object 1. Specifically, the self-location estimation unit 52 corrects the self-location obtained by dead reckoning based on the self-location obtained by AMCL. The self-location estimation unit 52 supplies the finally determined self-location and the variance obtained by AMCL to the weighting coefficient setting unit 53.

[0028] The weighting factor setting unit 53 sets a weighting factor relating to the reliability of the sensor data from the distance measurement sensor 30 based on the variance from the self-position estimation unit 52, supplies the weighting factor to the sampling period setting unit 54, and also feeds it back to the self-position estimation unit 52. The self-position estimation unit 52, to which the weighting factor has been fed back, performs self-position estimation of the moving object 1 based on the sensor data for which the weighting factor has been set.

[0029] The weighting factor set here can be regarded as the reliability of the estimation result by AMCL (self-location estimation using a particle filter). Therefore, the self-location estimation unit 52 weights the self-location calculated by AMCL in response to the feedback of the weighting factor set by the weighting factor setting unit 53.

[0030] The sampling period setting unit 54 sets the sampling period of the distance measurement sensor 30 based on the weighting coefficient from the weighting coefficient setting unit 53, and supplies information indicating the set sampling period to the distance measurement sensor 30 and the self-position estimation unit 52. As a result, the distance measurement sensor 30 generates sensor data at the set sampling period, and the self-position estimation unit 52 executes AMCL at a timing synchronized with the sampling period.

[0031] (Operation of Mobile Body) Next, the flow of the self-position estimation process by the mobile body 1 of Fig. 1 will be described with reference to the flowchart of Fig. 2. In the mobile body 1, self-position estimation process by dead reckoning is executed in parallel with the process of Fig. 2.

[0032] In step S1, the weighting coefficient setting unit 53 reads from a storage area (not shown) a threshold value V_th for the dispersion of particles in the existence distribution of the self-position obtained by AMCL.

[0033] In step S2 , the sensor data acquisition unit 51 acquires sensor data from the distance measurement sensor 30 .

[0034] In step S3, the self-position estimation unit 52 performs self-position estimation based on the sensor data from the distance measurement sensor 30, thereby determining the self-position of the moving body 1 and the variance V of particles in the existence distribution of the self-position.

[0035] In step S4 , the weighting factor setting unit 53 acquires the variance V calculated by the self-position estimating unit 52 .

[0036] In step S5, the weighting coefficient setting unit 53 determines whether the variance V is greater than a threshold value V_th.

[0037] If it is determined in step S5 that the variance V is greater than the threshold value V_th, the process proceeds to step S6, where the weighting factor setting unit 53 sets the weighting factor of AMCL (sensor data from the distance measurement sensor 30) to a low value.

[0038] In step S7, the sampling period setting unit 54 sets a longer sampling period for the distance measurement sensor 30 based on the set weighting coefficient. Thereafter, the process returns to step S2, and the subsequent processes are repeated.

[0039] In other words, if the variance V is greater than the threshold value V_th, the reliability of the estimation result by AMCL is considered to be low, and when correcting the self-position obtained by dead reckoning, the priority of the self-position obtained by AMCL will be low and the frequency of such correction will also be low.

[0040] On the other hand, if it is determined in step S5 that the variance V is not greater than the threshold value V_th, the process proceeds to step S8, where the weighting factor setting unit 53 sets the weighting factor of AMCL (sensor data from the distance measurement sensor 30) high.

[0041] In step S9, the sampling period setting unit 54 shortens the sampling period of the distance measurement sensor 30 based on the set weighting coefficient. Then, the process returns to step S2, and the subsequent processes are repeated.

[0042] In other words, if the variance V is smaller than the threshold V_th, the reliability of AMCL is considered high, and in correcting the self-position determined by dead reckoning, the self-position determined by AMCL has a higher priority, and the frequency of correction also increases.

[0043] According to the above processing, the smaller the variance obtained by AMCL, the higher the priority of the self-position obtained by AMCL, and the larger the variance obtained by AMCL, the higher the priority of the self-position obtained by dead reckoning.

[0044] 2. Problems with the Related Art However, when the self-location estimation process by the moving body 1 described above is performed in a corridor, a warehouse, a tunnel, the inside of a train, or the like where there is little change in the feature amount, the estimated self-location will differ from the actual location.

[0045] 3 shows the mobile unit 1 moving through a long, featureless passageway. In FIG. 3, the direction in which the passageway extends is the Y-axis direction, and the direction of the passageway wall is the X-axis direction.

[0046] When the mobile unit 1 moves in the Y-axis direction, the error in the self-position is larger in the direction of travel (Y-axis direction) than in the orthogonal direction (X-axis direction) perpendicular to the direction of travel of the mobile unit 1. This is because there is less change in the scenery (i.e., change in feature quantities) in the direction of travel than in the orthogonal direction, making it difficult to recognize the distance traveled.

[0047] For example, immediately after the mobile unit 1 starts moving along a long, featureless passage, as shown in Figure 4A, the variance V calculated by AMCL is small, and the deviation between the estimated self-position and the actual position is also small.

[0048] On the other hand, after the moving body 1 has been moving down a long, featureless passage for some time, the variance V calculated by AMCL becomes large, particularly in the direction of travel. In this case, the priority of the self-position calculated by dead reckoning becomes higher, but in dead reckoning, errors accumulate with each repetition of the process. Therefore, for example, the self-position relative to a wall at the end of a long, featureless passage will differ by the accumulated error between the internal map held inside the moving body 1 and the actual measurement value based on dead reckoning. In this case, the risk of the moving body 1 colliding with the wall increases.

[0049] Therefore, in the technology disclosed herein, even if the variance obtained by AMCL is large, the priority of the self-position obtained by AMCL is not lowered, thereby realizing correction of the self-position obtained by dead reckoning.

[0050] 3. Configuration and Operation of a Moving Body to Which the Technology According to the Present Disclosure is Applied> (Configuration of Moving Body) FIG. 5 is a block diagram showing an example configuration of a moving body 1 to which the technology according to the present disclosure is applied.

[0051] The moving body 1 shown in FIG. 5 differs from the moving body 1 shown in FIG. 1 in that it includes distance measurement sensors 30t and 30n and a processor 150 instead of the distance measurement sensor 30 and the processor 50.

[0052] The ranging sensor 30t (first ranging sensor) senses the area ahead (traveling direction) of the moving body 1, thereby generating sensor data (point cloud data) representing the distance to surrounding objects and their shapes, and supplies the sensor data to the processor 150. The ranging sensor 30t may be configured by a LiDAR or a stereo camera.

[0053] The ranging sensor 30n (second ranging sensor) senses the side of the moving body 1 (the direction perpendicular to the traveling direction) to generate sensor data (point cloud data) representing the distance to surrounding objects and their shapes, and supplies the sensor data to the processor 150. The ranging sensor 30 may also be configured by LiDAR or a stereo camera.

[0054] The processor 150 performs self-position estimation of the moving body 1 based on the sensor data from the wheel encoders 10, the IMU 20, and the distance measurement sensors 30t and 30n.

[0055] The processor 150 is configured as an information processing device that operates by executing a predetermined program, for example. The processor 150 realizes, as functional blocks, a sensor data acquisition unit 151, a self-position estimation unit 152, a weighting coefficient setting unit 153, and a sampling period setting unit 154.

[0056] The sensor data acquisition unit 151, the self-position estimation unit 152, the weighting coefficient setting unit 153, and the sampling period setting unit 154 provided in the processor 150 basically have the same functions as the sensor data acquisition unit 51, the self-position estimation unit 52, the weighting coefficient setting unit 53, and the sampling period setting unit 54 provided in the processor 50 in Figure 1, so detailed explanations will be omitted.

[0057] However, the self-position estimation unit 152 performs self-position estimation of the moving object 1 using a particle filter based on the sensor data generated by the distance measurement sensors 30t and 30n.

[0058] Specifically, the self-position estimation unit 152 performs self-position estimation using AMCL based on the sensor data generated by the ranging sensor 30t, thereby determining the self-position in the direction of travel of the moving body 1 (Y-axis direction) as well as the variance in the direction of travel of particles in the distribution of the self-position existence.

[0059] In addition, the self-position estimation unit 152 performs self-position estimation using AMCL based on the sensor data generated by the ranging sensor 30n, thereby determining the self-position in the orthogonal direction (X-axis direction) of the moving body 1, as well as the variance in the orthogonal direction of the particles in the distribution of the self-position existence.

[0060] In this way, the self-position estimation unit 152 can determine the direction perpendicular to the traveling direction of the moving body 1 by using the sensor data from the distance measurement sensor 30t and the distance measurement sensor 30n.

[0061] The weighting coefficient setting unit 153 sets a weighting coefficient relating to the reliability of the sensor data from each of the distance measuring sensors 30t, 30n based on the variance in the traveling direction and the variance in the orthogonal direction calculated by the self-position estimation unit 152, and supplies the weighting coefficient to the sampling period setting unit 154 and also feeds it back to the self-position estimation unit 152. The self-position estimation unit 152, to which the weighting coefficients have been fed back, performs self-position estimation of the moving object 1 based on the sensor data for which the weighting coefficients have been set.

[0062] The sampling period setting unit 154 sets the sampling period of each of the distance measuring sensors 30t, 30n based on the weighting coefficient set by the weighting coefficient setting unit 153, and supplies information indicating the set sampling period to each of the distance measuring sensors 30t, 30n and the self-position estimation unit 152. As a result, each of the distance measuring sensors 30t, 30n generates sensor data at the set sampling period, and the self-position estimation unit 152 executes AMCL at timing synchronized with each sampling period.

[0063] (Operation of Mobile Body) Next, the flow of the self-position estimation process by the mobile body 1 in Fig. 5 will be described with reference to the flowchart in Fig. 6. In the mobile body 1, self-position estimation process by dead reckoning is executed in parallel with the process in Fig. 6.

[0064] In step S101, the self-position estimation unit 152 reads the self-position in the direction perpendicular to the direction of travel that has been obtained and corrected by dead reckoning, and the weighting coefficient setting unit 153 reads the threshold V_th for the dispersion of particles in the distribution of the existence of the self-position from a memory area not shown.

[0065] In step S102, the sensor data acquisition unit 151 acquires sensor data from each of the distance measurement sensors 30t and 30n.

[0066] In step S103, the self-position estimation unit 152 performs self-position estimation based on sensor data from the ranging sensors 30t and 30n, thereby determining the self-position in the direction of travel of the moving body 1 and the self-position in the perpendicular direction, as well as the variance Vt in the direction of travel of the particles and the variance Vn in the perpendicular direction in the distribution of the self-position.

[0067] In step S104 , the weighting factor setting unit 153 acquires the variance Vt in the traveling direction and the variance Vn in the orthogonal direction calculated by the self-position estimating unit 152 .

[0068] In step S105, the weighting coefficient setting unit 153 determines whether the variance Vt in the traveling direction is greater than a threshold V_th or whether the variance Vn in the orthogonal direction is greater than a threshold V_th (Vt>V_th or Vn>V_th).

[0069] If it is determined in step S105 that Vt>V_th or Vn>V_th, the process proceeds to step S106, where the weighting coefficient setting unit 153 determines whether the variance Vt in the traveling direction is greater than the variance Vn in the perpendicular direction (Vt>Vn).

[0070] If it is determined in step S106 that Vt > Vn, that is, for example, if there is a feature on the wall of the passageway through which the moving body 1 is moving, the process proceeds to step S107, and the weighting coefficient setting unit 153 maintains the weighting coefficient of the AMCL (sensor data from the distance measurement sensor 30t) for the front distance measurement sensor 30t.

[0071] Then, in step S108, the sampling period setting unit 154 also maintains (does not change) the sampling period of the distance measurement sensor 30t based on the maintained weighting coefficient. Thereafter, the process returns to step S102, and the subsequent processes are repeated.

[0072] In other words, even if there is little change in the scenery in the direction of travel and the variance Vt in the direction of travel is greater than the threshold V_th, the reliability of the estimation result by AMCL in the direction of travel is maintained, and correction of the self-position determined by dead reckoning continues based on the self-position in the direction of travel determined by AMCL.

[0073] On the other hand, if it is determined in step S106 that Vt is not greater than Vn, that is, for example, if there are no features on the wall of the passageway through which the moving body 1 moves, but there is a feature on the wall in front, the process proceeds to step S109, and the weighting coefficient setting unit 153 maintains the weighting coefficient of the AMCL (sensor data from the distance measurement sensor 30t) for the forward distance measurement sensor 30t.

[0074] Then, in step S110, the sampling period setting unit 154 also maintains (does not change) the sampling period of the distance measurement sensor 30t based on the maintained weighting coefficient. Thereafter, the process returns to step S102, and the subsequent processes are repeated.

[0075] In other words, even if there is little change in the scenery in the orthogonal direction and the variance Vt in the orthogonal direction is greater than the threshold V_th, if there is a change in the scenery in the direction of travel, the reliability of the estimation result by AMCL in the direction of travel is maintained, and correction of the self-position determined by dead reckoning is continued based on the self-position in the direction of travel determined by AMCL.

[0076] Now, if it is determined in step S105 that Vt>V_th or Vn>V_th is not satisfied, the process proceeds to step S111.

[0077] In step S111, the weighting coefficient setting unit 153 determines whether the variance Vt in the traveling direction is greater than a threshold V_th and the variance Vn in the orthogonal direction is greater than a threshold V_th (Vt>V_th and Vn>V_th).

[0078] If it is determined in step S111 that Vt > V_th and Vn > V_th, that is, for example, if the moving body 1 is moving in a room surrounded by featureless walls, the process proceeds to step S112, and the weighting coefficient setting unit 153 sets a low weighting coefficient for the AMCL (sensor data from the distance measurement sensor 30t) for the front distance measurement sensor 30t.

[0079] Then, in step S113, the sampling period setting unit 154 sets the sampling period of the front distance measuring sensor 30t to be long based on the set weighting coefficient. Thereafter, the process returns to step S102, and the subsequent processes are repeated.

[0080] In other words, if both the variance Vt in the direction of travel and the variance Vn in the perpendicular direction are greater than the threshold value V_th, the reliability of the estimation result by AMCL is deemed to be low, and in correcting the self-position determined by dead reckoning, the priority of the self-position determined by AMCL will be low and the frequency of such correction will also be low.

[0081] On the other hand, if it is determined in step S111 that Vt > V_th and Vn > V_th are not true, that is, if there are changes in the scenery in both the direction of travel and the perpendicular direction, the process proceeds to step S114, and the weighting coefficient setting unit 153 sets a high weighting coefficient for the AMCL (sensor data from the ranging sensors 30t, 30n) for each of the ranging sensors 30t, 30n.

[0082] Then, in step S115, the sampling period setting unit 154 sets the sampling period of each of the distance measurement sensors 30t, 30n to be shorter based on the set weighting coefficients. Thereafter, the process returns to step S102, and the subsequent processes are repeated.

[0083] That is, if both the variance Vt in the traveling direction and the variance Vn in the orthogonal direction are smaller than the threshold V_th, the reliability of the estimation result by AMCL is deemed to be high, and in correcting the self-position determined by dead reckoning, the self-position determined by AMCL has a higher priority and is corrected more frequently. In other words, since the self-position determined by dead reckoning is corrected frequently based on the self-position determined by AMCL, errors due to dead reckoning do not accumulate.

[0084] According to the above process, even when the variance of the traveling direction determined by AMCL is large, the priority of the self-position of the traveling direction determined by AMCL is maintained, thereby making it possible to continue correcting the self-position determined by dead reckoning. Therefore, even when the mobile object moves through an environment with little change in feature quantities, such as a corridor, a warehouse, a tunnel, or the inside of a train, it is possible to suppress the accumulation of errors due to dead reckoning and prevent the self-position recognized by the mobile object from deviating from its actual position. As a result, it is possible to improve the accuracy of estimating the self-position.

[0085] For example, if a food delivery robot can move without its recognized position deviating from its actual position, it can give a sense of security to people around it. Also, if a transport robot can move without its recognized position deviating from its actual position, it can prevent unintended abnormal stops and improve transport throughput.

[0086] 4. Other Configurations of the Moving Body The moving body 1 described above can also have other configurations.

[0087] (Configuration Using Environmental Map) FIG. 7 is a block diagram showing another configuration example of the moving body 1. In FIG.

[0088] The mobile body 1 shown in FIG. 7 differs from the mobile body 1 shown in FIG. 5 in that it further includes an environmental map DB (Database) 210 .

[0089] The environmental map DB (Database) 210 stores an environmental map of the environment in which the mobile object 1 moves. The self-position estimation unit 152 can determine the traveling direction and the perpendicular direction of the mobile object 1 by using the environmental map stored in the environmental map DB 210 in addition to the sensor data from the distance measurement sensors 30t and 30n.

[0090] (Configuration without Internal Sensor) FIG. 8 is a block diagram showing yet another configuration example of the moving body 1. In FIG.

[0091] The moving body 1 shown in FIG. 8 differs from the moving body 1 shown in FIG. 5 in that it does not include the wheel encoders 10 and the IMU 20, which are internal sensors.

[0092] That is, in the moving body 1 shown in FIG. 8, the self-position estimating unit 152 does not perform self-position estimation by dead reckoning, but performs self-position estimation only by AMCL.

[0093] Even with this configuration, by executing the self-position estimation process described with reference to Figure 6, it is possible to improve the accuracy of estimating the self-position even when the mobile body 1 moves through an environment where there is little change in the features, such as a corridor, a warehouse, a tunnel, or the inside of a train.

[0094] In addition, the functional blocks realized by the processor 150 of the mobile body 1 may be realized by a device external to the mobile body 1, such as a computer capable of communicating with the mobile body 1 via a network, or a controller or terminal device for controlling the mobile body 1.

[0095] 5. Example of Computer Hardware Configuration The above-described series of processes can be executed by hardware or software. When the series of processes is executed by software, the program constituting the software is installed from a program recording medium into a computer incorporated in dedicated hardware, a general-purpose personal computer, or the like.

[0096] 9 is a block diagram showing an example of the hardware configuration of a computer that executes the above-described series of processes using a program. The processor 150 that constitutes the mobile object 1 may be configured, for example, by a computer having a configuration similar to that shown in FIG.

[0097] A CPU (Central Processing Unit) 301 , a ROM (Read Only Memory) 302 , and a RAM (Random Access Memory) 303 are interconnected by a bus 304 .

[0098] An input / output interface 305 is also connected to the bus 304. An input unit 306 including a keyboard, a mouse, etc., and an output unit 307 including a display, a speaker, etc. are connected to the input / output interface 305. Also connected to the input / output interface 305 are a storage unit 308 including a hard disk, a nonvolatile memory, etc., a communication unit 309 including a network interface, etc., and a drive 310 that drives removable media 311.

[0099] In a computer configured as described above, the CPU 301 performs the above-described series of processes by, for example, loading a program stored in the storage unit 308 into the RAM 303 via the input / output interface 305 and the bus 304 and executing the program.

[0100] The program executed by the CPU 301 is provided, for example, by being recorded on a removable medium 311 or via a wired or wireless transmission medium such as a local area network, the Internet, or digital broadcasting, and is installed in the storage unit 308 .

[0101] The program executed by the computer may be a program that processes in chronological order according to the order described in this specification, or may be a program that processes in parallel or at the required timing, such as when called.

[0102] In this specification, a system refers to a collection of multiple components (devices, modules (components), etc.), regardless of whether all of the components are contained in the same housing. Therefore, multiple devices housed in separate housings and connected via a network, and a single device housed in a single housing with multiple modules, are both systems.

[0103] The effects described in this specification are merely examples and are not limiting, and other effects may also be present.

[0104] The embodiments of the present disclosure are not limited to the above-described embodiments, and various modifications are possible within the scope of the gist of the present disclosure.

[0105] For example, the embodiment of the present disclosure can be configured as a cloud computing system in which a single function is shared and processed collaboratively by multiple devices via a network.

[0106] Furthermore, each step described in the above flowchart can be executed by one device, or can be shared and executed by a plurality of devices.

[0107] Furthermore, when one step includes multiple processes, the multiple processes included in that one step can be executed by one device or can be shared and executed by multiple devices.

[0108] The effects described in this specification are merely examples and are not limiting, and other effects may also be present.

[0109] Furthermore, the technology disclosed herein may have the following configurations: (1) A self-location estimation method that performs self-location estimation of a moving body using a particle filter based on sensor data from a first ranging sensor that senses a direction ahead of the moving body and a second ranging sensor that senses a direction to the side of the moving body, sets weighting factors related to the reliability of the sensor data from the first ranging sensor and the second ranging sensor based on a variance in the moving body's direction of travel and a variance in an orthogonal direction orthogonal to the moving direction in a presence distribution of the estimated self-location of the moving body, and sets sampling periods of the first ranging sensor and the second ranging sensor according to the set weighting factors. (2) The self-location estimation method described in (1), in which the self-location estimation of the moving body is performed based on the sensor data for which the weighting factors have been set. (3) The self-location estimation method described in (2), in which, when either the variance in the moving direction or the variance in the orthogonal direction is greater than a predetermined threshold, at least maintains the weighting factor of the sensor data from the first ranging sensor and does not change the sampling period of the first ranging sensor. (4) The self-localization method according to (3), wherein, if both the variance in the traveling direction and the variance in the orthogonal direction are greater than a predetermined threshold, the weighting coefficient of the sensor data from the first ranging sensor is set low and the sampling period of the first ranging sensor is set long. (5) The self-localization method according to (4), wherein, if both the variance in the traveling direction and the variance in the orthogonal direction are smaller than a predetermined threshold, the weighting coefficient of the sensor data from the first ranging sensor and the second ranging sensor is set high and the sampling period of the first ranging sensor and the second ranging sensor is set short. (6) The self-localization method according to any of (1) to (5), wherein self-localization is performed by adaptive Monte Carlo localization (AMCL) as the self-localization using the particle filter. (7) The self-localization method according to any of (1) to (6), wherein the first ranging sensor and the second ranging sensor are configured by LiDAR (Light Detection And Ranging).(8) The self-localization method according to any one of (1) to (6), wherein the first ranging sensor and the second ranging sensor are configured by a stereo camera. (9) The self-localization method according to any one of (1) to (8), wherein the traveling direction and the orthogonal direction are determined using the sensor data from the first ranging sensor and the second ranging sensor. (10) The self-localization method according to (9), wherein the traveling direction and the orthogonal direction are determined further using an environmental map of an environment in which the mobile object moves. (11) The self-localization method according to any one of (1) to (10), wherein the self-localization of the mobile object is estimated using a Kalman filter based on the sensor data from an internal sensor that senses an internal state of the mobile object, and the self-localization result using the particle filter and the self-localization result using the Kalman filter are integrated. (12) The self-localization method according to (11), wherein the internal sensor includes a wheel encoder that detects the rotation direction and the rotation angle of a wheel provided on the mobile object. (13) The self-location estimation method according to (11), wherein the internal sensor includes an IMU (Inertial Measurement Unit) that detects inertial motion of the moving body. (14) The self-location estimation method according to any one of (1) to (13), wherein the moving body is a transport robot or a food delivery robot capable of autonomous movement. (15) The self-location estimation method according to any one of (1) to (13), wherein the moving body is a vehicle capable of autonomous driving. (16) The self-location estimation method according to any one of (1) to (13), wherein the moving body is a drone capable of autonomous flight.(17) An information processing device comprising: a self-position estimation unit that performs self-position estimation of a moving body using a particle filter based on sensor data from a first ranging sensor that senses the front of the moving body and a second ranging sensor that senses the side of the moving body; a weight coefficient setting unit that sets a weight coefficient related to the reliability of the sensor data from the first ranging sensor and the second ranging sensor based on the variance in the direction of travel of the moving body and the variance in an orthogonal direction perpendicular to the direction of travel in the existence distribution of the estimated self-position of the moving body; and a sampling period setting unit that sets the sampling period of the first ranging sensor and the second ranging sensor in accordance with the set weight coefficient. (18) A program for causing a computer to execute the following process: estimating the self-position of a moving body using a particle filter based on sensor data from a first ranging sensor that senses the front of the moving body and a second ranging sensor that senses the side of the moving body; setting a weighting coefficient related to the reliability of the sensor data from the first ranging sensor and the second ranging sensor based on the variance in the direction of travel of the moving body and the variance in an orthogonal direction perpendicular to the direction of travel in the existence distribution of the estimated self-position of the moving body; and setting the sampling period of the first ranging sensor and the second ranging sensor according to the set weighting coefficient.

[0110] REFERENCE SIGNS LIST 1 Mobile body, 10 Wheel encoder, 20 IMU, 30t, 30n Distance measurement sensors, 150 Processor, 151 Sensor data acquisition unit, 152 Self-position estimation unit, 153 Weighting coefficient setting unit, 154 Sampling period setting unit, 210 Environmental map DB

Claims

1. A self-location estimation method comprising: performing self-location estimation of a moving body using a particle filter based on sensor data from a first ranging sensor that senses the front of the moving body and a second ranging sensor that senses the side of the moving body; setting a weighting coefficient related to the reliability of the sensor data from the first ranging sensor and the second ranging sensor based on the variance in the moving direction of the moving body and the variance in an orthogonal direction perpendicular to the moving direction in the presence distribution of the estimated self-location of the moving body; and setting a sampling period of the first ranging sensor and the second ranging sensor according to the set weighting coefficient.

2. The self-location estimation method according to claim 1, further comprising the step of: estimating the self-location of the moving object based on the sensor data to which the weighting coefficient has been set.

3. A self-position estimation method as described in claim 2, wherein, when either the variance in the travel direction or the variance in the perpendicular direction is greater than a predetermined threshold, at least the weighting coefficient of the sensor data from the first ranging sensor is maintained, and the sampling period of the first ranging sensor is not changed.

4. A self-position estimation method as described in claim 3, wherein, when both the variance in the travel direction and the variance in the perpendicular direction are greater than a predetermined threshold, the weighting coefficient of the sensor data from the first ranging sensor is set low, and the sampling period of the first ranging sensor is set long.

5. A self-position estimation method as described in claim 4, wherein, when both the variance in the travel direction and the variance in the perpendicular direction are smaller than a predetermined threshold, the weighting coefficient of the sensor data from the first ranging sensor and the second ranging sensor is set high, and the sampling period of the first ranging sensor and the second ranging sensor is set short.

6. The self-location estimation method according to claim 1, wherein self-location estimation using the particle filter is performed by AMCL (Adaptive Monte Carlo Localization).

7. The self-location estimation method according to claim 1, wherein the first ranging sensor and the second ranging sensor are configured using LiDAR (Light Detection And Ranging).

8. The self-location estimation method according to claim 1, wherein the first distance measuring sensor and the second distance measuring sensor are constituted by stereo cameras.

9. The self-location estimation method according to claim 1, wherein the traveling direction and the orthogonal direction are determined using the sensor data from the first distance measurement sensor and the second distance measurement sensor.

10. The self-location estimation method according to claim 9, further comprising: determining the traveling direction and the orthogonal direction by further using an environmental map of the environment in which the mobile object moves.

11. The self-location estimation method according to claim 1, further comprising: performing self-location estimation of the moving body using a Kalman filter based on the sensor data from an internal sensor that senses the internal state of the moving body; and integrating the self-location estimation result using the particle filter and the self-location estimation result using the Kalman filter.

12. The self-location estimation method according to claim 11, wherein the internal sensor includes a wheel encoder that detects a rotation direction and a rotation angle of a wheel equipped on the moving body.

13. The self-location estimation method according to claim 11, wherein the internal sensor includes an IMU (Inertial Measurement Unit) that detects inertial motion of the moving body.

14. The self-location estimation method according to claim 1, wherein the moving body is a transport robot or a food delivery robot capable of autonomous movement.

15. The self-location estimation method according to claim 1, wherein the moving body is a vehicle capable of autonomous driving.

16. The self-location estimation method according to claim 1, wherein the moving body is a drone capable of autonomous flight.

17. An information processing device comprising: a self-position estimation unit that performs self-position estimation of a moving body using a particle filter based on sensor data from a first ranging sensor that senses the front of a moving body and a second ranging sensor that senses the side of the moving body; a weighting coefficient setting unit that sets a weighting coefficient related to the reliability of the sensor data from the first ranging sensor and the second ranging sensor based on the variance in the moving direction of the moving body and the variance in an orthogonal direction perpendicular to the moving direction in the presence distribution of the estimated self-position of the moving body; and a sampling period setting unit that sets the sampling period of the first ranging sensor and the second ranging sensor in accordance with the weighting coefficient that has been set.

18. A program for causing a computer to execute the following processes: perform self-position estimation of a moving body using a particle filter based on sensor data from a first distance measurement sensor that senses in front of the moving body and a second distance measurement sensor that senses the side of the moving body; set a weighting coefficient related to the reliability of the sensor data from the first distance measurement sensor and the second distance measurement sensor based on the variance in the moving direction of the moving body and the variance in an orthogonal direction perpendicular to the moving direction in the presence distribution of the estimated self-position of the moving body; and set a sampling period of the first distance measurement sensor and the second distance measurement sensor in accordance with the set weighting coefficient.

Citation Information

Patent Citations

  • Unmanned trackless vehicle

    JP1994332531A

  • Traveling position detector

    JP2002063682A

  • Autonomous mobile robot and control method of the same

    JP2015170127A

  • Self position estimation device and mobile body provided with self position estimation device

    JP2016091202A

  • Self-position estimation device

    JP2021018638A

Cited By

  • Wall-climbing robot full-coverage path planning obstacle avoidance method based on laser radar

    CN120143824A