WO2022044470A1 - Mobile body control method, mobile body control device, and mobile body - Google Patents

Abstract

A mobile body control method of the present disclosure includes a step (S21) for controlling movement of a mobile body by model predictive control using a predictive model for predicting a behavior of the mobile body, and a step (S33) for estimating a parameter, on the basis of the behavior of the mobile body when the mobile body was moved in a movement pattern for estimating a parameter used in the predictive model.

Description

Mobile control method, mobile control device, and mobile

The present disclosure relates to a moving body control method for controlling a moving body, a moving body control device, and a moving body.

Self-propelled vacuum cleaners are gaining popularity. The self-propelled vacuum cleaner generates a traveling route for cleaning a predetermined cleaning range, and performs cleaning while traveling along the traveling route. In order to clean the entire cleaning area, it is necessary to run a self-propelled vacuum cleaner so that it does not deviate from the specified travel route.

Japanese Unexamined Patent Publication No. 2015-28804

However, when the self-propelled vacuum cleaner travels on a carpet or the like, the present inventors may deviate from the traveling path due to the wheels being taken by the fluff of the carpet or the like. I recognized that as an issue. Then, in order to allow the self-propelled vacuum cleaner to travel along the traveling route even in such a case, it was examined to control the traveling of the self-propelled vacuum cleaner by model predictive control (MPC). Model predictive control is a control method that optimizes control while predicting future responses at each time using a predictive model (see, for example, Patent Document 1).

In order to improve the accuracy and efficiency when controlling a self-propelled vacuum cleaner by model predictive control, it is necessary to further improve the accuracy of the predictive model.

The present disclosure provides a technique for improving the accuracy of controlling a moving body.

The moving body control method in the present disclosure includes a step of controlling the movement of the moving body by model prediction control using a prediction model for predicting the behavior of the moving body, and a movement for estimating the parameters used in the prediction model. It comprises a step of estimating parameters based on the behavior of the moving body when the moving body is moved by a pattern.

The moving body control method in the present disclosure includes a step of controlling the movement of the moving body by model prediction control using a prediction model for predicting the behavior of the moving body. In the prediction model, the behavior of the moving body is based on the lateral slip speed, which indicates the speed at which the moving direction of the moving body shifts to the left or right, or the dead time, which represents the delay time until the control instruction issued to the moving body is reflected in the turning of the moving body. Predict.

In the moving body control method in the present disclosure, a step of controlling the movement of the moving body by model prediction control using a prediction model for predicting the behavior of the moving body and a step of controlling the movement of the moving body by model prediction control are controlled. Sometimes it is necessary to determine the accuracy of the prediction by the prediction model or the control by the model prediction control, the step to determine the necessity of estimating the parameters used in the prediction model based on the accuracy, and the estimation of the parameters. When it is determined that, a step of estimating a parameter is provided.

The moving body control device in the present disclosure is a control unit that controls the movement of the moving body by model prediction control using a prediction model for predicting the behavior of the moving body, and a parameter for estimating the parameters used in the prediction model. It is provided with a parameter estimation unit that estimates parameters based on the behavior of the moving body when the moving body is moved in a moving pattern.

The moving body in the present disclosure is a moving body, and includes a moving unit driven to move and a control device for controlling the moving unit. The control device moves the moving body with a control unit that controls the moving part by model prediction control using a prediction model for predicting the behavior of the moving body and a movement pattern for estimating the parameters used in the prediction model. It is provided with a parameter estimation unit that estimates parameters based on the behavior of the moving body when it is made to move.

According to the moving body control technology of the present disclosure, the accuracy of controlling the moving body can be improved.

It is a block diagram which shows the structure of the self-propelled vacuum cleaner which is an example of the moving body of this disclosure. It is a control flow diagram which shows the procedure of the moving body control method which concerns on Embodiment 1. FIG. It is a control flow diagram which shows the detailed procedure of step S21 of FIG. It is a figure which shows the test running pattern 1 in the moving body control method which concerns on Embodiment 1. FIG. It is a figure which shows the test running pattern 2 in the moving body control method which concerns on Embodiment 1. FIG. It is a figure which shows the time change of the turning angle when the self-propelled vacuum cleaner is run according to the test run pattern 2. It is a figure which shows the test running pattern 3 in the moving body control method which concerns on Embodiment 1. FIG. It is a figure which shows the time change of the turning angle when the self-propelled vacuum cleaner is run according to the test run pattern 3. It is a figure which shows the example of the traveling path of the self-propelled vacuum cleaner which concerns on Embodiment 1. FIG. It is a figure which shows the example of the test run in the performance deterioration area. It is a figure which shows the example of the traveling path of the self-propelled vacuum cleaner after re-estimating the parameter. 14 (a) and 14 (b) are diagrams for explaining the effect of the test running pattern 1. 14 (a) and 14 (b) are diagrams for explaining the effect of the test running pattern 2. 14 (a) and 14 (b) are diagrams for explaining the effect of the test running pattern 3. It is a figure which shows the contribution degree of 5 kinds of parameter estimation accuracy to the follow-up performance to a target path.

Hereinafter, embodiments will be described in detail with reference to the drawings. However, more detailed explanation than necessary may be omitted. For example, detailed explanations of already well-known matters or duplicate explanations for substantially the same configuration may be omitted.

It should be noted that the accompanying drawings and the following description are provided for those skilled in the art to fully understand the present disclosure, and are not intended to limit the subject matter described in the claims.

(Embodiment 1)
Hereinafter, the first embodiment will be described with reference to FIGS. 1 to 15. In the moving body control method according to the first embodiment, in order to move a moving body such as a self-propelled vacuum cleaner accurately and efficiently along a target path, the movement of the moving body is controlled by model prediction control. At the control timing, the first control instruction in the time series of future control inputs obtained by solving the optimization problem is input to the moving body. At the next control timing, the state of the actual moving object is acquired from the sensor, and by solving the optimization problem again, a control instruction for correcting the deviation between the target route and the actual traveling route is input to the moving object. .. In this way, in the model predictive control, the moving object is feedback-controlled while calculating the optimization problem at high speed online. As a result, the moving body can be efficiently moved while suppressing deviation from the target path and unnecessary movement.

In order to improve the accuracy and efficiency of model prediction control, it is necessary to improve the accuracy of the prediction model for predicting the behavior of moving objects. In the present embodiment, the prediction model at least represents the lateral slip speed indicating the speed at which the moving direction of the moving body shifts to the left or right, and the delay time until the control instruction issued to the moving body is reflected in the turning of the moving body. One of the time, the constraint condition of the turning speed of the moving body, the natural angular frequency and the attenuation coefficient, which are the secondary delay factors in the turning control of the moving body, is used as a parameter. As a result, the accuracy can be improved while suppressing the calculation load of the prediction model from becoming excessive.

The values of these parameters depend on the material of the floor, the shape of the unevenness of the floor, the presence or absence of rugs such as carpets and tiles laid on the floor, the type and material, and the condition of the surface of the floor and rug. It can be different. In the present embodiment, the parameters are estimated based on the behavior of the moving body when the moving body is moved by the movement pattern for estimating the parameters used in these prediction models. As a result, the value of the parameter in the moving range of the moving body can be estimated and set with high accuracy, so that the accuracy of the prediction model can be further improved, and by extension, the moving body is moved more accurately and efficiently. be able to.

The values of these parameters are not always constant in the moving range of the moving object and may vary depending on the position. In addition, the parameter values may change over time due to changes in the state of the rug laid on the floor. In the present embodiment, when the movement of the moving body is controlled, the accuracy of prediction by the prediction model or the quality of the tracking performance of the moving body to the target path is determined, and based on the prediction accuracy or the quality of the tracking performance. , Determine the necessity of estimating the parameters used in the prediction model. When it is determined that the parameter needs to be estimated, the moving object is moved by the movement pattern for estimating the parameter to estimate the parameter. As a result, the accuracy of the prediction model can be maintained well, so that the moving body can be moved more accurately and efficiently.

[1-1. Constitution]
FIG. 1 is a block diagram showing a configuration of a self-propelled vacuum cleaner 50, which is an example of the moving body of the present disclosure. The self-propelled vacuum cleaner 50 includes a traveling unit 3, a position angle sensor 4, a model estimation unit 5, an overall control unit 6, a performance quality determination unit 9, a control target generation unit 10, and a control unit 11.

These configurations can be realized by the CPU, memory, and other LSIs of any computer in terms of hardware, and can be realized by programs loaded in memory in terms of software, but here, they are linked. It depicts the functional blocks that will be realized. Therefore, it is understood by those skilled in the art that these functional blocks can be realized in various forms only by hardware and by a combination of hardware and software.

The overall control unit 6 controls functions and operations such as cleaning by the self-propelled vacuum cleaner 50. The traveling unit 3 includes a configuration such as a wheel for moving the main body of the self-propelled vacuum cleaner 50, a motor for driving the wheel, and a steering device for changing the direction of the wheel. The position / angle sensor 4 detects the position and turning angle of the self-propelled vacuum cleaner 50. The position / angle sensor 4 may include any sensor capable of detecting a position or an angle, such as a LiDAR, an acceleration sensor, a gyro sensor, or a GPS.

The control target generation unit 10 includes a target route generation unit 1 and a test running pattern generation unit 7. The control unit 11 includes a model prediction control unit 2 and a test drive control unit 8. When the self-propelled vacuum cleaner 50 performs cleaning, the overall control unit 6 sets the traveling mode to the predictive control traveling mode. In the predictive control travel mode, the target route generation unit 1 of the control target generation unit 10 generates the target route, and the model predictive control unit 2 of the control unit 11 travels the self-propelled vacuum cleaner 50 along the target route. Let me. When the self-propelled vacuum cleaner 50 estimates the parameters, the overall control unit 6 sets the traveling mode to the test traveling mode. In the test driving mode, the test driving pattern generation unit 7 of the control target generation unit 10 generates a test driving pattern for estimating the parameters, and the test driving control unit 8 of the control unit 11 self-propells according to the test driving pattern. The running vacuum cleaner 50 is run.

The target route generation unit 1 generates a target route when the self-propelled vacuum cleaner 50 performs cleaning. The target route generation unit 1 generates a travel route for efficiently cleaning the entire cleaning range set by the overall control unit 6, and sets it in the model prediction control unit 2 as the target route of the self-propelled vacuum cleaner 50. do.

The model prediction control unit 2 controls the traveling unit 3 so as to move the self-propelled vacuum cleaner 50 along the target route generated by the target route generating unit 1. The model prediction control unit 2 holds a prediction model for predicting the behavior of the self-propelled vacuum cleaner 50. As described above, this prediction model at least represents the lateral slip speed, which indicates the speed at which the moving direction of the moving body shifts to the left and right, and the delay time until the control instruction issued to the moving body is reflected in the turning of the moving body. One of the time, the constraint condition of the turning speed of the moving body, the natural angular frequency and the attenuation coefficient, which are the secondary delay factors in the turning control of the moving body, is used as a parameter. The model prediction control unit 2 is a target path, the current position and direction of the self-propelled vacuum cleaner 50 detected by the position angle sensor 4, the traveling speed and turning angle by the traveling unit 3, and the future self-propelled type predicted by the prediction model. Based on the behavior of the vacuum cleaner 50 and the like, the control instructions to be input to the motor and the steering device of the traveling unit 3 are determined.

The test running pattern generation unit 7 generates a test running pattern for estimating the above parameters and outputs the test running pattern to the test running control unit 8. The test running pattern generation unit 7 may further generate a movement route to the target region for which the parameter is estimated. The target area may be the entire cleaning range of the self-propelled vacuum cleaner 50, or may be a part of the cleaning range. As will be described later, a region in which the estimation accuracy of the prediction model is not good may be a target region for estimating parameters. The test running pattern generation unit 7 generates a test running pattern for estimating the parameters in the target area. The details of the test running pattern will be described later. The test running pattern may be stored in advance in the storage device, or may be automatically generated by the test running pattern generation unit 7 according to the type and accuracy of the parameter to be estimated.

The test travel control unit 8 controls the travel unit 3 according to the test travel pattern generated by the test travel pattern generation unit 7. The test running control unit 8 does not perform feedback control such as model prediction control, and outputs control instructions according to the test running pattern to the running unit 3. When the movement route to the target area is set, the test travel control unit 8 moves the self-propelled vacuum cleaner 50 to the target area along the movement route, and then self-propells in the target area according to the test travel pattern. The running vacuum cleaner 50 is run. While controlling the test run of the self-propelled vacuum cleaner 50, the test run control unit 8 stores the position and turning angle of the self-propelled vacuum cleaner 50 detected by the position angle sensor 4 in association with the time. Record on the device. Recording of the position and the turning angle may be performed by the model estimation unit 5.

The model estimation unit 5 estimates the target in the target area based on the time-series data of the position and turning angle of the self-propelled vacuum cleaner 50 recorded in the storage device while the self-propelled vacuum cleaner 50 is in a test run. The parameters are estimated, and the prediction model held in the model prediction control unit 2 is updated. Details of the parameter estimation method will be described later.

The performance quality determination unit 9 was acquired from the target route generated by the target route generation unit 1 and the position / angle sensor 4 when the model prediction control unit 2 controls the traveling of the self-propelled vacuum cleaner 50. Compare with the actual driving route and judge whether the performance to follow the target route meets the standard. The performance quality determination unit 9 predicts performance in a certain region when statistical values such as an integrated value, an average value, and a maximum value of the magnitude of deviation between the target route and the actual traveling route exceed a predetermined value. May be judged not to be good. The performance quality determination unit 9 may determine the quality of the prediction performance or the follow-up performance in a certain region based on the control instruction value calculated by the model prediction control unit 2. For example, the difference between the control instruction value calculated by the model prediction control unit 2 and the control instruction value when the model prediction control is not performed, and the statistical values such as the integrated value, the average value, and the maximum value of the difference are predetermined values. If it exceeds, it may be judged that the prediction performance or the follow-up performance in the region is not good. The performance quality determination unit 9 records the determination result in the storage device in association with the position and the turning angle of the self-propelled vacuum cleaner 50.

In addition to the configuration shown in FIG. 1, the self-propelled vacuum cleaner 50 has a suction motor for sucking dust, a dust box for storing the sucked dust, a brush for sweeping the floor surface, and the like. However, it is omitted in FIG.

[1-2. motion]
FIG. 2 is a control flow diagram showing a procedure of the mobile body control method according to the first embodiment. FIG. 3 is a control flow diagram showing a detailed procedure of step S21 of FIG.

When the self-propelled vacuum cleaner 50 cleans a new cleaning range, a first test run is performed to estimate the parameters in the cleaning range prior to cleaning (S11). The test running pattern generation unit 7 generates a test running pattern for estimating all parameters in the new cleaning range, and the test running control unit 8 uses the self-propelled vacuum cleaner 50 according to the generated test running pattern. Let it run. The model estimation unit 5 estimates parameters based on the time-series data of the control instruction value obtained in step S11, the position of the self-propelled vacuum cleaner 50, and the turning angle, and sets the parameters in the prediction model of the model prediction control unit 2. (S12). Note that this initial parameter estimation may be omitted. In that case, typical values may be preset in the prediction model of the model prediction control unit 2 according to the material, type, surface condition, etc. of the floor or rug in the cleaning range of the self-propelled vacuum cleaner 50. ..

When the self-propelled vacuum cleaner 50 cleans the cleaning range, the target route generation unit 1 generates a target route for cleaning the cleaning range, and the model prediction control unit 2 generates a target route along the target route by model prediction control. The self-propelled vacuum cleaner 50 is run (S21).

Specifically, as shown in FIG. 3, the model prediction control unit 2 acquires the position and turning angle information of the self-propelled vacuum cleaner 50 from the position angle sensor 4 (S211). The model prediction control unit 2 calculates an output plan of the control instruction value so that the distance between the target route and the position of the self-propelled vacuum cleaner 50 becomes the shortest in the future using the prediction model (S212). The model prediction control unit 2 outputs the first output of the calculated output plan to the traveling unit 3 as a control instruction value (S213). Steps S211 to S213 are repeated (N in S214) until the traveling control of the self-propelled vacuum cleaner 50 is completed (Y in S214).

While the self-propelled vacuum cleaner 50 is running, the performance quality determination unit 9 determines whether the prediction performance or the follow-up performance is good or bad, and if there is a place where the prediction performance or the follow-up performance is not good, the position is recorded. (S22). If there is no place where the prediction performance or the follow-up performance is not good (N in S30), steps S21 and S22 are repeated at the next cleaning timing.

If there is a region where the prediction performance or tracking performance has deteriorated (Y in S30), the parameters are re-estimated in that region. Parameter re-estimation is performed for the number of performance deterioration regions output by the overall control unit 6. When the overall control unit 6 notifies the test travel pattern generation unit 7 of the performance deterioration region and the estimation target parameter, the test travel pattern generation unit 7 sets the movement route to the test travel start point in the performance deterioration region and the performance deterioration region. Generate a test run pattern for estimating the parameters to be estimated. The test travel control unit 8 moves the self-propelled vacuum cleaner 50 to the test travel start point in the performance deterioration region according to the generated movement route (S31). The test run control unit 8 performs the test run (S32) and the parameter estimation (S33) in the same manner as in steps S11 and S12. If there is a performance deterioration region for which the parameter should be estimated (N in S41), the process returns to step S31, moves to the test run start point of the next performance deterioration region, and performs test run and parameter estimation. When the estimation of the parameters in all the performance deterioration regions is completed (Y in S41), the test run is completed, and steps S21 and S22 are executed at the next cleaning timing.

In the test run and parameter estimation performed in steps S11 and S12, or steps S32 and S33, five types of parameters are estimated by three types of test run patterns. Model parameter estimation using these test driving patterns will be described with reference to FIGS. 4 to 8.

FIG. 4 shows a test running pattern 1 in the moving body control method according to the first embodiment. The test running pattern 1 is a test running pattern for estimating the lateral slip speed of the self-propelled vacuum cleaner 50. As shown by L11 in FIG. 4, the test running pattern 1 includes a straight path longer than a predetermined length along the traveling direction of the target path. If there is no influence of the lateral slip speed or the noise component contained in the self-propelled vacuum cleaner 50, the self-propelled vacuum cleaner 50 should go straight according to the test running pattern 1. However, due to the influence of the lateral slip speed and noise, the self-propelled vacuum cleaner 50 advances while shifting in the direction in which the lateral slip speed is generated, as shown by L12 in FIG. The noise component includes rattling of the main body and the carpet, irregular vibration caused by the turning angle control inside the self-propelled vacuum cleaner 50, and the like.

The overall control unit 6 instructs the test travel pattern generation unit 7, the test travel control unit 8, and the model estimation unit 5 to estimate the strike-slip speed in the target range for estimating the parameters. The test running pattern generation unit 7 generates a test running pattern 1 for estimating the movement route to the test start point of the target range and the lateral slip speed. The test travel control unit 8 inputs a control instruction to the travel unit 3 according to the generated test travel pattern 1, and causes the self-propelled vacuum cleaner 50 to travel. While the self-propelled vacuum cleaner 50 is running, the test running control unit 8 records the position and turning angle acquired by the position angle sensor 4 in the storage device. When the running of the test running pattern 1 is completed, the model estimation unit 5 acquires the forward distance D11 and the lateral slip distance D12 from the path L12 recorded when the self-propelled vacuum cleaner 50 runs the test running pattern 1. Calculate the lateral slip speed with respect to the forward speed. The length of the straight path included in the test running pattern 1 may be a value such that the estimation accuracy of the strike-slip speed is higher than the required level.

FIG. 5 shows a test running pattern 2 in the moving body control method according to the first embodiment. The test running pattern 2 is a test running pattern for estimating the constraint conditions of the dead time and the turning speed of the self-propelled vacuum cleaner 50. First, as shown by L20 in FIG. 5, a route that advances by the distance D11 previously acquired and travels straight so as to shift laterally by the distance D12 is used as a reference route. As shown by L21 in FIG. 5, the test running pattern 2 includes a route along the reference route L20 halfway and turning right or left from the middle. The angle of turning to the right or left is set to an angle larger than a predetermined angle that touches the constraint condition of the turning speed of the self-propelled vacuum cleaner 50. The self-propelled vacuum cleaner 50 should proceed according to the test driving pattern 2 if it is not affected by the dead time, the constraint condition of the turning speed, the natural angular frequency, the attenuation coefficient, and the noise component of the self-propelled vacuum cleaner 50. Is. However, due to the influence of dead time, turning speed constraints, natural angular frequency, attenuation coefficient, and noise, the self-propelled vacuum cleaner 50 draws a forward bulging path as shown by L22 in FIG. ..

FIG. 6 is a diagram showing the time change of the turning angle when the self-propelled vacuum cleaner 50 is run according to the test running pattern 2. In FIG. 6, the alternate long and short dash line indicates the control instruction value of the turning angle when the self-propelled vacuum cleaner 50 is driven along the path L20. Further, the broken line indicates the control instruction value of the turning angle when the self-propelled vacuum cleaner 50 is driven according to the path L21, that is, the test running pattern 2. In the test running pattern 2, 0 ° is output to the running section 3 as a control instruction value of the turning angle until the middle of the route, but at a point where the turning angle is largely turned to the right or left, the angle is larger than the predetermined angle in a stepped manner. The control instruction value of the turning angle is output to the traveling unit 3. This turning angle is a target value in the traveling direction of the self-propelled vacuum cleaner 50, but since it is larger than the turning angle that touches the constraint condition of the turning speed of the traveling unit 3, the self-propelled vacuum cleaner 50 immediately follows. Can not. Therefore, in the traveling path L22 of the actual self-propelled vacuum cleaner 50, the turning angle increases or decreases with the inclination defined by the constraint condition of the turning speed, and eventually converges to the turning angle of the control instruction value. The constraint condition of the turning speed can be estimated from the slope of the time change of the turning angle. Further, at a point where the turning angle is largely turned to the right or left, the dead time is estimated from the delay from the output of the turning angle control instruction value to the traveling unit 3 until the turning angle of the self-propelled vacuum cleaner 50 starts to increase or decrease. can do.

The overall control unit 6 instructs the test travel pattern generation unit 7, the test travel control unit 8, and the model estimation unit 5 to estimate the constraint conditions of the dead time and the turning speed in the target range for estimating the parameters. The test running pattern generation unit 7 generates a test running pattern 2 for estimating the movement path from the test end point of the test run pattern 1 to the test start point of the test run pattern 2 and the constraint conditions of the dead time and the turning speed. do. The test travel control unit 8 moves the self-propelled vacuum cleaner 50 to the test start point, then inputs a control instruction to the travel unit 3 according to the generated test travel pattern 2, and travels the self-propelled vacuum cleaner 50. Let me. While the self-propelled vacuum cleaner 50 is running, the test running control unit 8 records the position and turning angle acquired by the position angle sensor 4 in the storage device. When the running of the test running pattern 2 is completed, the model estimation unit 5 outputs the control instruction value of the turning angle to the running unit 3 at the point where the test running pattern 2 makes a large right or left turn, and then the self-propelled vacuum cleaner. The dead time is estimated from the delay until the turning angle of 50 begins to increase or decrease. Further, the model estimation unit 5 estimates the constraint condition of the turning speed from the slope of the time change of the turning angle when the test running pattern 2 makes a large turn to the right or left. Further, the model estimation unit 5 estimates the natural angular frequency and the damping coefficient from the change in the turning angle before and after touching the constraint condition of the turning speed and holds them as temporary values for the generation of the test running pattern 3 described later. Use for.

FIG. 7 shows a test running pattern 3 in the moving body control method according to the first embodiment. The test running pattern 3 is a test running pattern for estimating the natural angular frequency and the attenuation coefficient of the self-propelled vacuum cleaner 50. In the test running pattern 3, as shown by L31 in FIG. 7, the bending angle of the path L21 of the test running pattern 2 is made smaller than a predetermined angle that does not touch the constraint condition of the turning speed. The turning angle that does not touch the constraint condition of the turning angle is obtained by the constraint condition of the turning speed obtained above and the tentatively estimated natural angular frequency and attenuation coefficient. The self-propelled vacuum cleaner 50 should proceed according to the test driving pattern 3 if it is not affected by the dead time, the constraint condition of the turning speed, the natural angular frequency, the attenuation coefficient, and the noise component of the self-propelled vacuum cleaner 50. Is. However, due to the influence of dead time, turning speed constraints, natural angular frequency, attenuation coefficient, and noise, the self-propelled vacuum cleaner 50 draws a forward bulging path as shown by L32 in FIG. ..

FIG. 8 is a diagram showing the time change of the turning angle when the self-propelled vacuum cleaner 50 is run according to the test running pattern 3. In FIG. 8, the alternate long and short dash line indicates the control instruction value of the turning angle when the self-propelled vacuum cleaner 50 is driven according to the path L21, that is, the test running pattern 2. The broken line indicates the control instruction value of the turning angle when the self-propelled vacuum cleaner 50 is driven according to the path L31, that is, the test running pattern 3. In the test running pattern 3, a control instruction value having a turning angle smaller than that of the test running pattern 2 is output to the running unit 3. The turning angle of the actual self-propelled vacuum cleaner 50 in the traveling path L32 draws a curve and converges to the turning angle of the control instruction value while having vibration due to noise. By obtaining a quadratic lag curve that fits this curve, it is possible to estimate the natural angular frequency and the attenuation coefficient in the change in the turning speed of the self-propelled vacuum cleaner 50.

The overall control unit 6 instructs the test drive pattern generation unit 7, the test drive control unit 8, and the model estimation unit 5 to estimate the natural angular frequency and the attenuation coefficient in the target range for estimating the parameters. The test running pattern generation unit 7 generates a moving path from the test ending point of the test running pattern 2 to the test starting point of the test running pattern 3, and a test running pattern 3 for estimating the natural angular frequency and the attenuation coefficient. The test travel control unit 8 moves the self-propelled vacuum cleaner 50 to the test start point, then inputs a control instruction to the travel unit 3 according to the generated test travel pattern 3, and travels the self-propelled vacuum cleaner 50. Let me. While the self-propelled vacuum cleaner 50 is running, the test running control unit 8 records the position and turning angle acquired by the position angle sensor 4 in the storage device. When the running of the test running pattern 3 is completed, the model estimation unit 5 obtains a quadratic lag curve that matches the curve of the turning angle change after the point where the test running pattern 3 makes a large right or left turn, thereby determining the intrinsic angle. Estimate frequency and attenuation coefficient.

As described above, the traveling paths L12, L22, and L32 of the actual self-propelled vacuum cleaner 50 all include irregular vibration due to the influence of noise and the like, but the influence of noise is caused by using the test traveling pattern 1. The suppressed lateral slip speed can be estimated, and by using the test running pattern 2, it is possible to estimate the constraint conditions of the dead time and the turning speed in which the influence of noise is suppressed, and the test running pattern 3 is used. Therefore, it is possible to estimate the intrinsic angular frequency and the attenuation coefficient in which the influence of the turning speed constraint condition and noise is suppressed.

In steps S31, S32, and S33, a test run is performed in the performance deterioration region detected in step S21, and the parameters are estimated and used again in addition to the parameters estimated in step S12 only in that region. Test running and parameter estimation in such a performance deterioration region will be described with reference to FIGS. 9 to 11.

FIG. 9 shows an example of a traveling route of the self-propelled vacuum cleaner 50 according to the first embodiment. In steps S11 and S12, when the initial test run and the initial parameter estimation are performed at the point 20 near the start point of the target route, the prediction model of the model prediction control unit 2 and the actual one are as long as the characteristics of the point 20 and the floor are the same. Since the behavior of the self-propelled vacuum cleaner 50 is the same, the model prediction control unit 2 can accurately predict the behavior of the self-propelled vacuum cleaner 50 and accurately follow the target path L41. However, when there is an area 30 in which the characteristics of the floor are different, for example, an area 30 in which the length of the hair of the laid carpet is different, in the area 30, the prediction model of the model prediction control unit 2 and the actual self-propelled vacuum cleaner are used. Since the behaviors of the 50s do not match, the self-propelled vacuum cleaner 50 deviates from the target path L41. In step S21, the performance quality determination unit 9 starts from the target path L41 based on the target path L41 generated by the target path generation unit 1 and the path L42 of the self-propelled vacuum cleaner 50 acquired by the position angle sensor 4. The deviation and the position and turning angle of the self-propelled vacuum cleaner 50 when the deviation occurs are recorded. The overall control unit 6 determines a test travel target point for which the parameter should be re-estimated based on the information of the position and the turning angle at which the follow-up performance to the target route is below the reference. For example, the performance deterioration region 30 may be divided according to a position, a traveling direction, or the like, and a test running target point may be set in the divided region.

FIG. 10 shows an example of a test run in a performance deterioration region. In the example of FIG. 10, the performance deterioration region 30 recorded in the example of FIG. 9 is divided into three by the overall control unit 6, and test travel target points T1 to T3 are set in each region. The test running pattern generation unit 7 generates a movement route from the test running start point 21 to the test running end point 22 through three types of test running patterns at each of the test running target points T1 to T3. In step S31, the test run control unit 8 moves the self-propelled vacuum cleaner 50 to the start point of the test run pattern at the test run target point T1. In step S32, the test run control unit 8 carries out a test run according to three types of test run patterns. In step S33, the model estimation unit 5 estimates the parameters at the test travel target point T1 based on the data of the position and the turning angle of the self-propelled vacuum cleaner 50 obtained in step S32. The model estimation unit 5 records the estimated area-specific parameters in the storage device in association with the area, and reflects them in the prediction model of the model prediction control unit 2. The above steps are carried out for all test driving target points T1 to T3.

FIG. 11 shows an example of a traveling path of the self-propelled vacuum cleaner 50 after re-estimating the parameters. As shown in FIG. 10, when different parameters are set for each region, the model prediction control unit 2 predicts based on the position and turning angle of the self-propelled vacuum cleaner 50 detected by the position angle sensor 4. Switch the parameters used in the model for each region. As a result, even if there are areas 30 with different floor characteristics within the cleaning range, the behavior of the prediction model and the actual self-propelled vacuum cleaner 50 can be matched, and the behavior of the self-propelled vacuum cleaner 50 can be matched in any area based on accurate prediction. Control instructions can be output. Therefore, as shown in FIG. 11, the follow-up performance to the target path can be maintained above the standard in all regions.

The test run according to the test run patterns 1 to 3 may be executed separately from the cleaning route for the self-propelled vacuum cleaner 50 to clean the cleaning range. For example, if it is necessary to estimate the parameters after the self-propelled vacuum cleaner 50 finishes cleaning and returns to the charging station or the like, a test run is executed at a predetermined timing until the next cleaning is executed. May be good. In this case, the overall control unit 6 may drive a suction motor, a brush, or the like to perform a cleaning operation while the test travel control unit 8 is executing the test travel according to the test travel patterns 1 to 3. As a result, conditions such as vibration caused by driving the suction motor and the brush and the influence on the floor surface can be made the same as in the actual cleaning, so that the estimation accuracy of the parameters can be further improved. The overall control unit 6 may stop driving the suction motor, the brush, or the like while the test travel control unit 8 is executing the test travel according to the test travel patterns 1 to 3. As a result, even when the test run is executed at night or during breaks, the noise and vibration generated by the test run can be suppressed.

The test run according to the test run patterns 1 to 3 is incorporated into the cleaning path for the self-propelled vacuum cleaner 50 to clean the cleaning range, and the parameters are estimated when the self-propelled vacuum cleaner 50 cleans the cleaning range. May be good. In the target area where the parameters are estimated, only a test run may be performed, or both a test run and a run for cleaning may be performed. In the former case, when the self-propelled vacuum cleaner 50 is traveling along the cleaning route in the predictive control travel mode and reaches the test travel start point, the overall control unit 6 switches the travel mode to the test travel mode. Perform a test run. When the test run is completed, the overall control unit 6 switches the run mode to the predictive control run mode, and the self-propelled vacuum cleaner 50 runs along the cleaning route from the test run end point. The overall control unit 6 may execute the cleaning operation even during the test run. In the latter case, when the test run is completed, the self-propelled vacuum cleaner 50 returns to the test run start point. The overall control unit 6 switches the travel mode to the predictive control travel mode, and the self-propelled vacuum cleaner 50 travels along the cleaning route from the test travel start point. This makes it possible to efficiently estimate the parameters and update the prediction model.

In order to show that the estimation accuracy of the five parameters of lateral slip speed, dead time, turning speed constraint condition, natural angular frequency, and damping coefficient can be improved by using the test running patterns 1 to 3 to show that the estimation accuracy can be improved. The behavior when the running type vacuum cleaner 50 is running under the control of model prediction control and when running according to the test running patterns 1 to 3 is reproduced by simulation, and five types of behaviors are reproduced based on each behavior. Estimated parameters.

FIGS. 12 (a) and 12 (b) are diagrams for explaining the effect of the test running pattern 1. FIG. 12A shows the result of simulating the behavior of the self-propelled vacuum cleaner 50 when traveling under the control of model prediction control. The lateral slip speed of the self-propelled vacuum cleaner 50 is set to 0.05 [m / s]. As the strike-slip velocity parameter of the MPC, 0.10 [m / s] having an error of + 100% from the true value is set. A normal distribution random number with a standard deviation of 2 deg is given as noise in the turning direction, and a standard deviation of 0.01 m is given as noise in the position. Since it is controlled by model predictive control, although there is an error in the lateral slip velocity parameter of the predictive model, it is controlled so that the deviation from the target path is small although it is incomplete, and as a result, the effect of lateral slip Is hard to appear. Therefore, when the strike-slip speed was estimated based on this behavior, a value of −23.2% of the true value was calculated. FIG. 12B shows the result of simulating the behavior of the self-propelled vacuum cleaner 50 during the test run according to the test run pattern 1. Since the test travel control unit 8 outputs a control instruction to the travel unit 3 so as to go straight forward regardless of the deviation from the target route, the lateral deviation of the actual travel route is directly reflected in the self-propelled vacuum cleaner in the target area. Reflects 50 lateral shifts. Therefore, when the strike-slip speed was estimated based on this behavior, a value of + 0.2% of the true value was calculated. That is, by estimating the strike-slip speed using the test running pattern 1, the estimation accuracy can be significantly improved.

13 (a) and 13 (b) are diagrams for explaining the effect of the test running pattern 2. FIG. 13A shows the result of simulating the behavior of the self-propelled vacuum cleaner 50 when traveling under the control of model prediction control. The dead time of the self-propelled vacuum cleaner 50 is set to 0.5 [s], and the upper limit value of 12.5 [deg / s] is set as a constraint condition of the turning speed. Since the control instruction is output in small steps to follow the target path, the influence of the constraints of the dead time and the turning speed is unlikely to appear. Therefore, when the constraints of dead time and turning speed were estimated based on this behavior, + 10% and -25% of the true values were calculated, respectively. FIG. 13B shows the result of simulating the behavior of the self-propelled vacuum cleaner 50 during the test run according to the test run pattern 2. Since the test travel control unit 8 outputs the control instruction value of the stepped turning angle to the travel unit 3 regardless of the deviation from the target path, the influence of the constraint conditions of the dead time and the turning speed appears remarkably. Therefore, when the constraints of dead time and turning speed were estimated based on this behavior, + 5% and -2.5% of the true values were calculated, respectively. That is, by estimating the constraint conditions of the dead time and the turning speed using the test running pattern 2, the estimation accuracy can be significantly improved.

14 (a) and 14 (b) are diagrams for explaining the effect of the test running pattern 3. FIG. 14A shows the result of simulating the behavior of the self-propelled vacuum cleaner 50 when traveling under the control of model prediction control. The natural angular frequency of the self-propelled vacuum cleaner 50 is set to 3.0 [rad / s], and the attenuation coefficient is set to 0.90. Since the control instruction is output in small steps to follow the target path, the influence of the natural angular frequency and the attenuation coefficient is unlikely to appear. Therefore, when the natural angular frequency and the attenuation coefficient were estimated based on this behavior, + 50% and + 25% of the true values were calculated, respectively. FIG. 14B shows the result of simulating the behavior of the self-propelled vacuum cleaner 50 during the test run according to the test run pattern 3. Since the test travel control unit 8 outputs a control instruction value of a stepped turning angle smaller than that of the test travel pattern 2 to the travel unit 3 regardless of the deviation from the target path, the influence of the natural angular frequency and the attenuation coefficient is affected. Appears prominently. Therefore, when the natural angular frequency and the attenuation coefficient were estimated based on this behavior, the true values of −2.5% and −5% were calculated, respectively. That is, by estimating the natural angular frequency and the attenuation coefficient using the test running pattern 3, the estimation accuracy can be significantly improved.

FIG. 15 shows the contribution of five types of parameter estimation accuracy to the tracking performance of the target path. As the values of the five types of parameters, four types are fixed to the above true values, and the behavior of the self-propelled vacuum cleaner 50 that runs under the control of model prediction control is simulated while changing the remaining one type. , The average value of the deviation from the target path was plotted. It can be seen that in any of the parameters, the more the estimated value deviates from the true value, the lower the tracking performance to the target path. In particular, it can be seen that the contribution of the strike-slip speed is large, and the estimation error of the strike-slip speed is directly linked to the deviation from the target path. The dead time, proper angular frequency, and attenuation coefficient also have a large contribution, and especially when a value smaller than the true value is estimated, the deviation from the target path is large. As for the constraint condition of the turning speed, if the estimation error is large, the deviation from the target path becomes large.

As described above, according to the technique of the present embodiment, the model while predicting the behavior of the self-propelled vacuum cleaner 50 with a prediction model using five types of parameters estimated with high accuracy by three types of test running patterns. By controlling the self-propelled vacuum cleaner 50 by predictive control, the follow-up performance to the target route can be significantly improved.

[1-3. Effect, etc.]
As described above, in the present embodiment, the moving body control method is used in the step of controlling the movement of the moving body by the model prediction control using the prediction model for predicting the behavior of the moving body, and in the prediction model. It is provided with a step of estimating parameters based on the behavior of the moving body when the moving body is moved by a movement pattern for estimating the parameters. As a result, the accuracy of the prediction model can be improved, so that the accuracy of controlling the moving body can be improved.

Further, in the present embodiment, the moving body control method moves the moving body in a plurality of movement patterns in order to estimate each of the plurality of parameters used in the prediction model. As a result, the accuracy of the prediction model can be further improved, so that the accuracy of controlling the moving body can be further improved.

Further, in the present embodiment, the above parameter includes a lateral slip speed representing a speed at which the moving direction of the moving body shifts to the left or right, and the moving pattern includes a linear moving path longer than a predetermined length. As a result, the strike-slip speed can be estimated with high accuracy, so that the accuracy of the prediction model can be improved.

Further, in the present embodiment, the above parameters include a dead time representing a delay time until the control instruction issued to the moving body is reflected or a constraint condition of the turning speed of the moving body, and the movement pattern has a predetermined angle. Includes a travel path that swivels at a greater angle. As a result, the constraint condition of the dead time or the turning speed can be estimated with high accuracy, so that the accuracy of the prediction model can be improved.

Further, in the present embodiment, the above parameters include the natural angular frequency or the attenuation coefficient, which are secondary delay elements in the movement control of the moving body, and the movement pattern is a movement path that turns at an angle smaller than a predetermined angle. include. As a result, the natural angular frequency or the attenuation coefficient can be estimated with high accuracy, so that the accuracy of the prediction model can be improved.

Further, in the present embodiment, the above parameters are estimated for each of a plurality of regions included in the moving range of the moving body. As a result, the accuracy of the prediction model can be further improved, so that the accuracy of controlling the moving body can be further improved.

Further, in the present embodiment, the step of estimating the parameter is executed before or after the moving body moves in the area where the parameter is not estimated. As a result, the accuracy of the prediction model can be further improved, so that the accuracy of controlling the moving body can be further improved.

Further, in the present embodiment, the moving body control method includes a step of determining the accuracy of prediction by the prediction model or control by the model prediction control when the movement of the moving body is controlled by the model prediction control. It further includes a step of determining whether or not the parameter needs to be estimated based on the above, and a step of estimating the parameter when it is determined that the parameter needs to be estimated. As a result, the accuracy of the prediction model can be maintained high, so that the accuracy of controlling the moving body can be further improved.

Further, in the present embodiment, the necessity of parameter estimation is determined for each of a plurality of regions included in the moving range of the moving body, and the parameters are estimated in the region where it is determined that the parameter estimation is necessary. .. As a result, the accuracy of the prediction model can be further improved, so that the accuracy of controlling the moving body can be further improved.

Further, in the present embodiment, the moving body is a vacuum cleaner, and the movement of the moving body is controlled by model prediction control so that the moving body moves along a cleaning path for cleaning a predetermined cleaning range. .. This makes it possible to improve the cleaning efficiency of the vacuum cleaner.

Further, in the present embodiment, the step of estimating the parameter is executed in a movement pattern different from the cleaning route. As a result, the accuracy of the prediction model can be further improved, so that the accuracy of vacuum cleaner control and the cleaning efficiency can be further improved.

Further, in the present embodiment, the cleaning operation by the vacuum cleaner is also executed in the step of estimating the parameters. As a result, the accuracy of the prediction model can be further improved, so that the accuracy of vacuum cleaner control and the cleaning efficiency can be further improved.

Further, in the present embodiment, the cleaning path includes a movement pattern, and the parameters are estimated when the vacuum cleaner cleans a predetermined cleaning range. This can improve the efficiency of parameter estimation.

Further, in the present embodiment, the moving body control method includes a step of controlling the movement of the moving body by model prediction control using a prediction model for predicting the behavior of the moving body, and the prediction model is a moving body. The behavior of the moving body is predicted by using the lateral slip speed, which indicates the speed at which the moving direction shifts to the left or right, or the dead time, which represents the delay time until the control instruction issued to the moving body is reflected, as parameters. As a result, the accuracy of the prediction model can be improved, so that the accuracy of controlling the moving body can be improved.

Further, in the present embodiment, the moving body control method includes a step of controlling the movement of the moving body by model prediction control using a prediction model for predicting the behavior of the moving body, and a movement of the moving body by model prediction control. A step to determine the accuracy of prediction by the predictive model or control by model predictive control when is controlled, a step to determine the necessity of estimating the parameters used in the predictive model based on the accuracy, and the parameter It comprises a step of estimating parameters when it is determined that estimation is necessary. As a result, the accuracy of the prediction model can be improved, so that the accuracy of controlling the moving body can be improved.

Further, in the present embodiment, the moving body control device is a control unit that controls the movement of the moving body by model prediction control using a prediction model for predicting the behavior of the moving body, and parameters used in the prediction model. It is provided with a parameter estimation unit that estimates parameters based on the behavior of the moving body when the moving body is moved by the movement pattern for estimating. As a result, the accuracy of the prediction model can be improved, so that the accuracy of controlling the moving body can be improved.

Further, in the present embodiment, the moving body includes a moving unit driven to move and a control device for controlling the moving unit, and the control device is a prediction model for predicting the behavior of the moving body. The parameters are estimated based on the behavior of the moving body when the moving body is moved by the control unit that controls the moving unit by the model prediction control using and the movement pattern for estimating the parameters used in the prediction model. It is equipped with a parameter estimation unit. As a result, the accuracy of the prediction model can be improved, so that the accuracy of controlling the moving body can be improved.

(Other embodiments)
As described above, the first embodiment has been described as an example of the technique disclosed in the present application. However, the technique in the present disclosure is not limited to this, and can be applied to embodiments in which changes, replacements, additions, omissions, etc. have been made. It is also possible to combine the components described in the first embodiment to form a new embodiment.

Therefore, other embodiments will be exemplified below.

In the first embodiment, the self-propelled vacuum cleaner 50 has been described as an example of the moving body. The moving body may be any moving body that can move autonomously, such as an autonomous vehicle, a carrier, and a robot.

Since the above-described embodiment is for exemplifying the technique in the present disclosure, various changes, replacements, additions, omissions, etc. can be made within the scope of claims or the equivalent thereof.

This disclosure can be used to control moving objects.

1 Target route generation unit 2 Model prediction control unit 3 Travel unit 4 Position angle sensor 5 Model estimation unit 6 Overall control unit 7 Test travel pattern generation unit 8 Test travel control unit 9 Performance quality judgment unit 10 Control target generation unit 11 Control unit 20 Point 21 Test run start point 22 Test run end point 30 Performance deterioration area 50 Self-propelled vacuum cleaner

Claims (17)

1. A step of controlling the movement of the moving body by model predictive control using a predictive model for predicting the behavior of the moving body, and
   A step of estimating the parameter based on the behavior of the moving body when the moving body is moved by a movement pattern for estimating the parameter used in the prediction model.
   A mobile control method comprising.
2. The moving body control method according to claim 1, wherein the moving body is moved by a plurality of movement patterns in order to estimate each of the plurality of parameters used in the prediction model.
3. The parameter includes a lateral slip speed representing a speed at which the moving direction of the moving body shifts to the left or right.
   The moving body control method according to claim 1 or 2, wherein the moving pattern includes a straight moving path longer than a predetermined length.
4. The parameter includes a dead time representing a delay time until the control instruction issued to the moving body is reflected, or a constraint condition of the turning speed of the moving body.
   The moving body control method according to any one of claims 1 to 3, wherein the moving pattern includes a moving path that turns at an angle larger than a predetermined angle.
5. The parameters include natural angular frequencies or attenuation coefficients that are secondary lag factors in the movement control of the moving object.
   The moving body control method according to any one of claims 1 to 4, wherein the moving pattern includes a moving path that turns at an angle smaller than a predetermined angle.
6. The moving body control method according to any one of claims 1 to 5, wherein the parameter is estimated for each of a plurality of regions included in the moving range of the moving body.
7. The moving body control method according to any one of claims 1 to 6, wherein the step of estimating the parameter is executed before or after the moving body moves in a region where the parameter is not estimated.
8. A step of determining the accuracy of prediction by the prediction model or control by the model prediction control when the movement of the moving object is controlled by the model prediction control.
   A step of determining the necessity of estimating the parameter based on the accuracy, and
   A step of estimating the parameter when it is determined that the parameter needs to be estimated, and a step of estimating the parameter.
   The mobile body control method according to any one of claims 1 to 7.
9. Whether or not the parameter is estimated is determined for each of a plurality of regions included in the moving range of the moving body.
   The mobile control method according to claim 8, wherein the parameter is estimated in a region where it is determined that the parameter needs to be estimated.
10. The moving body is a vacuum cleaner.
    The moving body according to any one of claims 1 to 9, wherein the movement of the moving body is controlled by the model predictive control so that the moving body moves along a cleaning path for cleaning a predetermined cleaning range. Control method.
11. The moving body control method according to claim 10, wherein the step of estimating the parameter is executed in the moving pattern different from the cleaning route.
12. The mobile control method according to claim 11, wherein the cleaning operation by the vacuum cleaner is executed even in the step of estimating the parameters.
13. The moving body control method according to claim 10, wherein the cleaning path includes the moving pattern, and the parameters are estimated when the vacuum cleaner cleans the predetermined cleaning range.
14. A step of controlling the movement of the moving body by model predictive control using a predictive model for predicting the behavior of the moving body is provided.
    In the prediction model, the behavior of the moving body is measured by using the lateral slip speed representing the speed at which the moving direction of the moving body shifts to the left or right or the dead time representing the delay time until the control instruction issued to the moving body is reflected as a parameter. Predictive mobile control method.
15. A step of controlling the movement of the moving body by model predictive control using a predictive model for predicting the behavior of the moving body, and
    A step of determining the accuracy of prediction by the prediction model or control by the model prediction control when the movement of the moving object is controlled by the model prediction control.
    A step of determining the necessity of estimating the parameters used in the prediction model based on the accuracy, and
    When it is determined that the parameter needs to be estimated, the step of estimating the parameter and the step of estimating the parameter
    A mobile control method comprising.
16. A control unit that controls the movement of the moving object by model predictive control using a predictive model for predicting the behavior of the moving object.
    A parameter estimation unit that estimates the parameters based on the behavior of the moving body when the moving body is moved by a movement pattern for estimating the parameters used in the prediction model.
    A mobile control device.
17. It ’s a mobile body,
    A moving part driven to move,
    A control device that controls the moving unit and
    Equipped with
    The control device is
    A control unit that controls the moving unit by model prediction control using a predictive model for predicting the behavior of the moving object.
    A parameter estimation unit that estimates the parameters based on the behavior of the moving body when the moving body is moved by a movement pattern for estimating the parameters used in the prediction model.
    A mobile body equipped with.

Images