JP7629787B2 - Self-propelled robot - Google Patents

Description

The present invention relates to a self-propelled robot, and is suitable for use in a self-propelled robot equipped with a non-destructive component estimation device that non-destructively measures the sugar content of fruit without plucking it from fruit trees.

Conventionally, to measure the sugar content of fruits such as tomatoes, a non-destructive measurement method using near-infrared spectroscopy has been put into practical use, in which the fruit is irradiated with near-infrared light, the absorbance is measured, and the sugar content is estimated from the absorbance.

However, typical non-destructive measurement methods are often performed after the fruit has been picked from the trees and collected, and fruit that is determined to have a sugar content that does not meet the required level is likely to be discarded as having no commercial value.

To solve this problem, a handheld saccharometer has been developed that can directly measure the sugar content of fruit while it is still on the tree (see Patent Document 1). This handheld saccharometer is designed to measure light that is irradiated from a light source and reflected by the surface of fruit or vegetable using a light receiving sensor, and calculates the sugar content based on the reflected light.

JP 2016-114544 A

However, the handheld sugar content meter shown in Patent Document 1 has the problem that the operator must measure the large number of fruits on the trees while holding each one by hand, which is very time-consuming for the operator. In addition, fruit trees are greatly affected by sunlight in the field, so a variable neutral density filter is provided in the light amount adjustment means, but this makes the configuration complicated.

In fruit trees exposed to sunlight, it is desirable to monitor the growth process of the large number of fruits that grow on the trees and to measure and manage the internal components, including sugar content, of each fruit with high accuracy, but the above-mentioned Patent Document 1 leaves problems that make it insufficient for practical use.

The present invention was made in consideration of the above points, and aims to propose a self-propelled robot equipped with a non-destructive component estimation device that can estimate with high accuracy the state of internal components during the fruit growth process.

In order to solve such problems, the present invention provides a self-propelled robot that is equipped with a non-destructive component estimation device that estimates sugar content by irradiating light on fruit and that travels on the ground autonomously or in response to external operation, the robot comprises a travel drive unit that drives the robot body to travel on the ground in a desired direction at a desired speed, an arm support unit that supports the base of an arm unit having a multi-joint mechanism with multiple degrees of freedom and is connected to the upper part of the travel drive unit integrally with the arm unit so as to be freely rotatable relative to the travel drive unit, and a travel control unit that drives and controls the travel drive unit and the arm support unit. The non-destructive component estimation device includes a light detection unit that detects scattered light generated from inside the fruit through the skin surface of the fruit when light is irradiated onto the fruit, a contact detection unit that is provided adjacent to the light detection unit and detects a state of contact with the fruit, an adsorption light shielding unit whose tip is made of an elastic material that can be adsorbed to the skin surface of the fruit and that covers the periphery of the light detection unit and the contact detection unit when adsorbed onto the fruit to block external light, a light source that irradiates light so as to irradiate the vicinity of the adsorption site of the fruit that is adsorbed onto the adsorption light shielding unit, and an image capturing unit that captures an image of the adsorption target by the adsorption light shielding unit and simultaneously captures the adsorption target. The device is equipped with an imaging distance measurement unit that measures the distance to the fruit, a spectroscopic measurement unit that disperses the scattered light detected by the light detection unit and measures a spectrum that represents the light intensity distribution for each wavelength, a library that stores the spectra of scattered light corresponding to a plurality of fruits of the same kind that are different from the fruit, an estimation model construction unit that constructs an estimation model that has been learned to estimate the state of the internal components of the fruit by performing multivariate analysis of the correlation between the spectrum measured by the spectroscopic measurement unit and the plurality of spectra stored in the library, and an internal state estimation unit that estimates the state of the internal components during the growth process of the fruit based on the estimation model constructed by the estimation model construction unit, and the travel control unit rotates the arm support unit and bends or extends each joint mechanism of the arm unit based on the image content of the target to be attracted by the adsorption light shielding unit captured by the imaging distance measurement unit and the measured distance to the target to be attracted, and then adsorbs the adsorption light shielding unit to the outer surface of the fruit while detecting the contact state of the fruit with the contact detection unit.

In this way, the self-propelled robot can drive to the location of a desired fruit tree, and then position the non-destructive component estimation device mounted on the arm for each of the multiple fruits on that tree, and estimate the state of the internal components of each fruit as it grows with a relatively high degree of accuracy. This is extremely effective when there are a large number of fruits on a tree, as it improves work efficiency significantly compared to individual manual work.

In this case, the non-destructive component estimation device can estimate the state of the internal components during the fruit's development with a relatively high degree of accuracy. In particular, by recording the changes in the internal components during the fruit's development in chronological order, it is possible to contribute to research into improving the varieties of each fruit that grows on fruit trees, based on the records. Furthermore, the non-destructive component estimation device can estimate the state of the internal components during the fruit's development with a relatively high degree of accuracy, without being affected by external light disturbances.

The present invention also includes an environmental imaging unit that captures images of the environment in which the robot body is traveling, an object detection unit that detects specific fruit trees in the traveling environment based on the images captured by the environmental imaging unit, and an environmental map creation unit that estimates the robot's position relative to the traveling environment and creates a two-dimensional or three-dimensional environmental map including the ground at the same time. The traveling control unit controls the traveling drive unit to move to a specific fruit tree detected by the object detection unit in the environmental map created by the environmental map creation unit, and then drives the arm unit to bring the suction light blocking unit into contact with each of the multiple fruits on the fruit tree, and sequentially assigns addresses to the fruits. Data representing the state of the internal components during the growth process of the fruit estimated by the internal state estimation unit of the non-destructive component estimation device is managed on an address-by-address basis.

As a result, the self-propelled robot can automatically move to the desired planting location for the fruit tree within the cultivation area, and by managing the individual addresses of multiple fruits growing on a specific fruit tree, it is possible to chronologically store the state of the internal components of each fruit as it grows.

Furthermore, in the present invention, the estimation model construction unit uses a supervised machine learning method as the estimation model, and the internal state estimation unit evaluates the accuracy of the learning results obtained by the supervised machine learning method through cross-validation. As a result, in the non-destructive component estimation device, the accuracy of the accuracy evaluation by the internal state estimation unit improves as the number of learning rounds increases, making it possible to estimate the state of the internal components during the fruit growth process with even higher accuracy.

Furthermore, in the present invention, the fruit is left on the tree, and the internal components of the fruit consist of at least one of the following: sugar content, acidity, γ-aminobutyric acid, hardness, density, and color. As a result, the non-destructive component estimation device can specifically confirm the changes in the internal components during the development of the fruit, either individually or in combination, and also the correlations between them.

The present invention makes it possible to realize a self-propelled robot equipped with a non-destructive component estimation device that can estimate the state of internal components during the growth process of a fruit with relatively high accuracy.

1 is an external view showing the overall configuration of a nondestructive component estimation device according to an embodiment of the present invention. 2 is a block diagram showing an internal configuration of the nondestructive component estimation device shown in FIG. 1. 1 is a graph for searching for an optimum wavelength region for estimating sugar content. 13 is a graph showing the results of sugar content estimation. 1 is a graph showing the relationship between the average sugar content and the error. 1 is an external view showing the overall configuration of a self-propelled robot according to an embodiment of the present invention; 7 is a schematic diagram showing the configuration of the bottom side of the self-propelled robot shown in FIG. 6. FIG. 2 is a block diagram showing the configuration of a control system of the self-propelled robot. FIG. 2 is a schematic diagram illustrating the autonomous movement of a self-propelled robot. 11 is a schematic diagram illustrating the sugar content estimation operation of a tomato by the nondestructive component estimating device;

One embodiment of the present invention will be described in detail below with reference to the drawings.

(1) Configuration of the Nondestructive Component Estimation Apparatus According to the Present Embodiment FIGS. 1(A) and 1(B) show a nondestructive component estimation apparatus 1 according to the present embodiment, which has a main body 2 that contains a control board and an electrical system for controlling the entire apparatus, an optical detection unit 3 provided at the front end of the main body 2, and an imaging unit 4 provided on the upper side of the main body.

The optical detection unit 3 is composed of a pair of light sources 5A, 5B that irradiate the fruit with light, and a light detection section 6 that detects the scattered light generated from inside the fruit through the surface of the fruit's skin. Each of the light sources 5A, 5B is made up of a halogen lamp with a wavelength range of 350 [nm] to 3,500 [nm]. The light detection section 6 has a light detection capability of 598 [nm] to 1,047 [nm], and detects the scattered light obtained from inside the fruit.

The optical detection unit 3 has a contact detection section 10 that detects the contact state with the fruit, located next to the light detection section 6, and an adhesive light-shielding section 11 made of an elastic material is provided to surround the light detection section 6 and the contact detection section 10.

The contact detection unit 10 is composed of a photoconductive element such as a cadmium sulfide (CdS) cell, and detects the state of contact with the fruit. The adsorption and light shielding unit 11 has a structure (such as a suction cup structure) at the tip that allows it to be adsorbed to the skin surface of the fruit, and when it is adsorbed to the fruit, it covers the periphery of the light detection unit 6 and the contact detection unit 10 to block external light.

In the optical detection unit 3, a pair of light sources 5A, 5B are provided at positions sandwiching the adsorption light-shielding part 11 (including the light detection part 6 and the contact detection part 10) from above and below, so that light can be irradiated to the vicinity of the adsorption part of the fruit that is adsorbed to the adsorption light-shielding part 11.

The pair of light sources 5A and 5B can irradiate light that is reflected by an internal mirror provided at the irradiation port of the housing of the optical detection unit 3 as a collective light, and are designed to make the optical detection unit 3 itself as small as possible.

The imaging unit 4 is also composed of a target imaging section 20 that images the target to be attracted by the suction and light shielding section 11, and a distance measuring section 21 that measures the distance to the target imaged by the target imaging section 20. The target imaging section 20 is made of an imaging element such as a CMOS image sensor, and the distance measuring section 21 is made of an optical displacement sensor. Note that the distance measuring section 21 may be of various types other than optical, such as eddy current, ultrasonic, or laser focus, as long as it is a non-contact distance measuring sensor.

In this nondestructive component estimation device 1, as shown in FIG. 2, the main body 2 is provided with a device control unit 30 and a memory unit 31 as a control board that controls the entire device. The device control unit 30 has the functions of a spectroscopic measurement unit 32, an estimation model construction unit 33, and an internal state estimation unit 34.

In this embodiment, we used tomatoes as the fruit and conducted experiments to estimate the sugar content as an internal component of the fruit.

The spectroscopic measurement unit 32 separates the scattered light detected by the light detection unit 6 and measures a spectrum that represents the light intensity distribution for each wavelength in the wavelength band of 598 nm to 1,047 nm. The memory unit 31 stores scattered light spectra corresponding to multiple fruits of the same type but different from the fruit as a library.

The estimation model construction unit 33 constructs an estimation model that is trained to estimate the state of the internal components of the fruit by performing multivariate analysis of the correlation between the spectrum measured by the spectroscopic measurement unit 32 and multiple spectra stored in the library of the storage unit 31. This estimation model is based on a supervised machine learning method, and various methods such as a neural network, a support vector machine, or a hidden Markovnikov may be applied.

For example, as shown in Figure 3, experiments have shown that among the combinations of scattered light in the wavelength range of 598 nm to 1,047 nm, the wavelength band with the highest accuracy in estimating sugar content is the 794 nm to 961 nm wavelength band (the range included in frame F in the figure).

The internal state estimation unit 34 estimates the state of the internal components during the fruit growth process based on the estimation model constructed by the estimation model construction unit 33. That is, the internal state estimation unit 34 evaluates the accuracy of the learning results obtained by the supervised machine learning method through cross-validation.

Specifically, the internal state estimation unit 34 calculates a calibration curve by multiple regression analysis, which is a type of multivariate analysis, using the wavelength band 794 [nm] to 961 [nm] that has the highest estimation accuracy obtained in FIG. 3 described above. Note that in addition to multiple regression analysis, various other multivariate analysis methods such as principal component analysis, factor analysis, and cluster analysis may also be applied.

Figure 4(A) shows the results of estimating the sugar content of 100 tomatoes of the same variety (Frutica: product name) using an estimation model (using a wavelength range of 794 nm to 961 nm) created based on the scattered light spectra of 336 tomatoes of the same variety. Figure 4(B) shows the relationship between the average sugar content and the error based on the results of Figure 4(A). As Figure 4(B) shows, the sugar content estimation accuracy is ±0.60 degrees.

Inside the main body 2, there is provided an internal battery 35 consisting of a secondary battery, and at the rear end of the main body 2, there is provided an arm mount (not shown) for detachably mounting the tip of the arm 44 of the self-propelled robot 40 (Figure 5).

This arm mount is provided with a power supply terminal capable of conducting electricity with the built-in battery 35, and as necessary, power can be supplied from the driving battery 76 (FIG. 7) on the self-propelled robot 40 side via the power supply terminal to charge the built-in battery 35. On the other hand, when the non-destructive component estimation device 1 is attached to the self-propelled robot 40, it is also possible to supply the power directly from the driving battery 76 on the self-propelled robot 40 side as operating power for the entire non-destructive component estimation device 1.

(2) Configuration of Self-Propelled Robot According to this Embodiment Fig. 5 shows a self-propelled robot 40 according to this embodiment as a whole. Self-propelled robot 40 is a two-wheel drive mobile body that travels on a floor surface autonomously or in response to external operation, and includes a traveling base unit (traveling drive unit) 43 that drives and rotates a pair of drive wheels 41A, 41B simultaneously or independently to travel robot main body 42 in a desired direction, and an arm support unit 45 that is connected to the upper section of the traveling base unit 43 and supports the end of an arm unit 44 that has at least one joint mechanism.

As shown in the bottom view of FIG. 6, the traveling base unit 43 is provided with a pair of omni-wheels 46A, 46B as the front wheels, and an auxiliary wheel 47 is provided at the center of the rear end between the pair of drive wheels 41A, 41B, which are the rear wheels. This allows the self-propelled robot 40 to travel in the forward and backward directions by the traveling base unit 43 simultaneously rotating and driving the pair of drive wheels 41A, 41B, but also to travel in either the left or right direction by independently rotating and moving the pair of drive wheels 41A, 41B, with each omni-wheel 46A, 46B following suit.

That is, the pair of drive wheels 41A, 41B are rotated independently by drive motors (e.g., in-wheel motors) 48A, 48B, respectively, and the self-propelled robot 40 moves forward and backward by the forward or backward rotation of the drive wheels 41A, 41B, and moves to the right or left while moving forward by giving a difference in the forward rotation angle of the drive wheels 41A, 41B. Also, by rotating the drive wheels 41A, 41B in opposite directions, the self-propelled robot 40 spins, i.e., changes direction at that position.

The arm support section 45 is supported by a base body section 50 connected to the upper section of the running base section 43, and the end of the arm section 45, which has at least one joint mechanism, is rotatably supported with the vertical direction as the center of rotation. A tray 51 is fixed to the upper section of the running base section 43 at a position close to the connection section of the arm support section 45.

A laser range sensor 52 and an RGB-D sensor 53 capable of three-dimensional scanning are provided on the front side (travel direction side) of the traveling base unit 43, and multiple 3D distance image sensors 54 are arranged at predetermined intervals around the upper side in all directions, to detect obstacles diagonally forward and to the left and right.

Specifically, the laser range sensor 52 shines light onto an object (obstacle) as seen from the installation position, receives the reflected light, and calculates the distance. By measuring the distance at regular angular intervals, it is possible to obtain fan-shaped distance information on a flat surface within a range of up to 30 m and an angle of 240 degrees.

In addition to its RGB color camera function, the RGB-D sensor 53 also has a depth sensor that can measure the distance to an object (obstacle) as seen by the camera, allowing it to perform 3D scanning of the object. This depth sensor is made up of an infrared sensor, and captures an image of the object while projecting a single pattern of structured light onto it, and uses the parameters to calculate the depth of each point on the image by triangulation.

For example, if Kinect (a registered trademark of Microsoft Corporation) is used as the RGB-D sensor 53, it is possible to capture images with a horizontal field of view of 57 degrees, a vertical field of view of 43 degrees, and a sensor range of 1.2 m to 3.5 m, and RGB images can be captured at 640 x 480 pixels, and depth images at 320 x 240 pixels, both at 30 frames per second.

The RGB-D sensor 53 is installed in the center of the top of the running base 43 because a height of 0.6m to 1.8m from the floor is required to ensure a vertical field of view.

The 3D distance image sensor 54 emits an LED pulse, measures the arrival time of the reflected light from the object in pixel units, and simultaneously calculates distance information to the object in pixel units by superimposing the acquired image information.

This 3D distance image sensor 54 is necessary as a complementary sensor for outdoor use because it has a more accurate detection capability than the above-mentioned RGB-D sensor 53 and a wider viewing angle than the laser range sensor 52. When Pixel Soleil (a product name of Nippon Signal Co., Ltd.) is used as the 3D distance image sensor 54, for example, it is possible to capture an image with a horizontal field of view of 72 degrees, a vertical field of view of 72 degrees, and a sensor range of 0.3 m to 4.0 m.

The self-propelled robot 40 of the present invention uses a laser range sensor 52, an RGB-D sensor 53, and a 3D distance image sensor 54 to realize SLAM (Simultaneous Localization and Mapping) technology that estimates its own position relative to the external environment and creates an environmental map at the same time.

The self-propelled robot 40 using this SLAM technology can determine its own movement path and move autonomously within the environment by dynamically generating an environmental map that represents the three-dimensional positions of objects that exist in the real space while estimating its own position with high accuracy.

The base body 50 of the arm support 45 is also provided with an imaging camera 55 equipped with multiple ultra-wide-angle lenses, allowing it to capture images of the space in all directions (360 degrees) of the self-propelled robot 40. The base body 50 of the arm support 45 is also equipped with a sound-collecting microphone and speaker (neither shown), which collects sounds from the surrounding environment and can emit speech or warning sounds as necessary.

(3) Configuration of the arm section in this embodiment In the arm support section 45, the arm section 44 is constituted by a root main body section 50 connected to the running base section 43 as a base, a pivoting support section 60 connected to the root main body section 50 so as to be freely rotatable, a first upper arm section 61 connected to the pivoting support section 60 so as to be rotatable in a direction perpendicular to the pivoting support section 60, a second upper arm section 62 connected to be rotatable in the same direction as the first upper arm section 61, a forearm section 63 connected to be rotatable in a direction perpendicular to the pivoting direction of the second upper arm section 62, and a wrist section 64 connected to be rotatable in a direction perpendicular to the pivoting direction of the forearm section 63.

That is, the arm support section 45 is configured such that the arm section 44 (rotating support section 60, first upper arm section 61, second upper arm section 62, forearm section 63, wrist section 64) of a multi-joint structure with five degrees of freedom is connected to the base main body section 50 so as to be rotatable about each axis (Axis A to Axis E). The nondestructive component estimation device 1 (Figure 1) can be detachably attached to the tip of the wrist section 64 constituting the arm section 44 as an end effector.

Specifically, the base body 50 and the pivoting support 60 are connected to be rotatable around the A axis, the pivoting support 60 and the first upper arm 61 are connected to be rotatable around the B axis, the first upper arm 61 and the second upper arm 62 are connected to be rotatable around the C axis, the second upper arm 62 and the forearm 63 are connected to be rotatable around the D axis, and the forearm 63 and the wrist 64 are connected to be rotatable around the E axis.

The joint parts between the base body 50 and the pivoting support 60, the joint part between the pivoting support 60 and the first upper arm 61, the joint part between the first upper arm 61 and the second joint part 62, the joint part between the second joint part 62 and the forearm 63, and the joint part between the forearm 63 and the wrist 64 are each provided with an actuator, e.g. a DC servo motor, and are rotated via a transmission mechanism (not shown).

A connector (not shown) is formed at the tip of the wrist 64 that constitutes the arm 44, and is adapted to be electrically connected to an arm mount (not shown) formed on the main body 2 of the nondestructive component estimation device 1. When the nondestructive component estimation device 1 is connected to the arm 44, the self-propelled robot 40 is adapted to drive and control the nondestructive component estimation device 1 as an end effector in conjunction with the operation of the arm 44.

(4) Internal Configuration of Self-Propelled Robot Fig. 7 shows the internal circuit configuration of self-propelled robot 40. In self-propelled robot 40, which travels on a floor surface autonomously or in response to external operations, travel control unit 70, which is responsible for controlling the entire robot, is mainly composed of a microcomputer.

Under the control of the travel control unit 70, the travel drive unit 71 controls the left and right motor drivers 65A, 65B to control the rotation of the drive motors 48A, 48B, thereby driving the robot body 42 to travel at a desired speed in a desired direction on the floor surface.

The environmental map creation unit 72 estimates the vehicle's own position relative to the driving environment based on driving route information from the target driving route memory unit 73, which stores preset driving route information, and the detection signals from the laser range sensor 52, RGB-D sensor 53, and 3D distance image sensor 54. At the same time, it creates a two-dimensional or three-dimensional environmental map that includes the floor surface, while determining whether the driving route is appropriate or needs to be changed, and whether there are any driving obstacles.

For example, when the self-propelled robot 40 travels along a taught travel path in a vinyl greenhouse (cultivation area), it determines whether or not it will come into contact with an obstacle such as a wall or stone steps, and stops just before coming into contact and changes its travel path in a direction along that travel path.

In fact, the self-propelled robot 40 uses the above-mentioned SLAM technology to automatically create an environmental map of the target area within the greenhouse (cultivation area) in which the robot body 42 can move. Specifically, the self-propelled robot 40 creates an environmental map that represents the entire desired target area by setting local maps on a two-dimensional grid as areas that represent the movement environment, based on distance and angle information from the target object obtained from the laser range sensor 52 and 3D distance image sensor 54.

At the same time, the robot calculates the distance traveled by the robot itself based on the rotation angle read from an encoder (not shown) corresponding to a pair of drive wheels 41A, 41B of the self-propelled robot 40, and estimates its own position based on the distance traveled by the robot itself and the matching between the next location map and the environmental map created up to the current point in time. An environmental map M1 that actually represents the entire target area is shown in Figure 8 (A).

In addition, in the self-propelled robot 40, the driving environment of the robot body 42 is captured by an imaging camera (environment detection unit) 55, and the target detection unit 74 detects a specific target in the driving environment based on the image captured by the imaging camera 55. Figure 8 (B) shows a state in which a specific target S1 in the driving environment is superimposed on the environmental map M1 shown in Figure 8 (A) described above.

The travel control unit 70 also controls the driving of the travel base unit 43 and the arm support unit 45, and also controls the nondestructive component estimation device 1 attached to the arm unit 44 as an end effector. In this self-propelled robot 40, when the nondestructive component estimation device 1 is connected to the arm unit 44, the nondestructive component estimation device 1 is controlled by the travel control unit 70 to perform its functions.

The self-propelled robot 40 also includes a communication unit 75 that wirelessly communicates with an external information input device (not shown), and transmits data on the environmental map according to the control of the driving control unit 70, and receives data indicating the content of operational instructions from the information input device.

The self-propelled robot 40 also has a relatively large capacity built-in driving battery 76 made of a secondary battery or a capacitor, and by conductively connecting the charging terminal of the driving battery 76 to a power supply terminal provided on an external power supply stand (not shown), it is possible to supply power from a commercial power source to the driving battery 76 via the power supply terminal to charge it.

(5) Operation of the Self-Propelled Robot According to this Embodiment In this embodiment, the travel control unit 70 of the self-propelled robot 40 drives and controls the travel base unit 43 and the arm support unit 45 so that the self-propelled robot 40 moves to a position in front of a specific tomato tree that is to be measured among multiple tomato trees planted in a vinyl greenhouse (cultivation area), for example.

That is, the travel control unit 70 detects specific fruit trees in the travel environment based on the captured image of the travel environment of the robot body 42, while estimating its own position relative to the travel environment, and at the same time creates a two-dimensional or three-dimensional environmental map that includes the ground, and controls the travel drive unit 71 to move to a specific fruit tree on the environmental map. As a result, the self-propelled robot 40 can automatically move to the planting position of the desired tomato tree within the cultivation area.

Next, the travel control unit 70 controls the imaging unit 4 in the non-destructive component estimation device 1 and at the same time drives and controls the arm unit 44 to bring the adsorption and light shielding portion 11 of the optical detection unit 3 into contact with each of the multiple fruits on a specific tomato tree (Figure 9).

That is, the travel control unit 70 captures an image of the fruit using the target imaging unit 20 in the imaging unit 4 of the nondestructive component estimation device 1, while simultaneously measuring the distance to the fruit using the distance measurement unit 21, and controls the drive of the arm unit 44 to align the adsorbing and shading unit 11 of the optical detection unit 3 with the fruit, and then adsorbs the adsorbing and shading unit 11 to the surface of the fruit while detecting the contact state with the fruit using the contact detection unit 10 of the optical detection unit 3 (Figure 10).

In this way, the self-propelled robot 40 is extremely effective when there are many fruits on the fruit trees, as it improves work efficiency significantly compared to individual manual work.

Then, the travel control unit 70 controls the device control unit 30 of the non-destructive component estimation device 1 to irradiate light from a pair of light sources 5A and 5B onto the fruit, and causes the light detection unit 6 to detect scattered light generated from inside the fruit through the skin surface of the fruit.

The device control unit 30 of the non-destructive component estimation device 1 disperses the scattered light detected by the light detection unit 6, measures a spectrum representing the light intensity distribution for each wavelength, and constructs an estimation model that is trained to estimate the state of the internal components of the fruit by performing multivariate analysis of the correlation between the measured spectrum and multiple spectra stored in the library (storage unit 31), and estimates the state of the internal components during the growth process of the fruit based on the constructed estimation model.

In this way, the non-destructive component estimation device 1 can estimate the state of the internal components of a fruit as it grows with a relatively high degree of accuracy. In particular, by recording the changes in the internal components of the fruit as it grows over time, it is possible to contribute to research into improving the varieties of each fruit that grows on fruit trees, based on the records.

Furthermore, for a specific fruit tree detected by the target detection unit 74, the self-propelled robot 40 drives the arm unit 44 to bring the suction light shielding unit 11 of the optical detection unit 3 in the non-destructive component estimation device 1 into contact with each of the multiple fruits growing on that tree, and sequentially assigns addresses to each of the fruits. The self-propelled robot 40 manages, on an address-by-address basis, data representing the state of the internal components during the growth process of the fruit, which is estimated by the internal state estimation unit 34 (device control unit 30) of the non-destructive component estimation device 1.

As a result, the self-propelled robot 40 can manage the addresses of multiple fruits growing on a specific fruit tree individually, and can store the state of the internal components of each fruit in chronological order as it grows.

(6) Other embodiments As described above, in this embodiment, a two-wheel drive type mobile body that can run autonomously or in response to external operation is used as the self-propelled robot 40, but the present invention is not limited to this, and a four-wheel drive type may also be used, and a wide variety of drive methods and numbers of wheels may be applied as long as the robot can move while supporting the arm unit 44. Also, although the pair of front wheels are each omni-wheels, various other running mechanisms such as casters or caterpillar tracks may also be applied.

In the above embodiment, a multi-joint structure with five degrees of freedom is used as the arm unit 44, but the present invention is not limited to this, and various arm units may be used as long as they are made of a multi-joint mechanism with multiple degrees of freedom.

Furthermore, in the above embodiment, a tomato is used as the fruit, and sugar content is estimated as the internal component of the fruit, but the present invention is not limited to this, and various fruits other than tomatoes (apples, mandarins, etc.) may be used, and further, in addition to sugar content, at least one of acidity, gamma-aminobutyric acid, hardness, density, and color may be applied as the internal component of the fruit. As a result, the non-destructive component estimation device can confirm the correlation between the changes in the internal components during the growth process of the fruit, either individually or in combination.

1...Non-destructive component estimation device, 2...Main body, 3...Optical detection unit, 4...Imaging unit, 5A, 5B...Light source, 6...Light detection unit, 10...Contact detection unit, 11...Absorption and light shielding unit, 20...Target imaging unit, 21...Distance measurement unit, 30...Device control unit, 31...Memory unit, 32...Spectroscopic spectrum measurement unit, 33...Estimation model construction unit, 34...Internal state estimation unit, 35...Built-in battery, 40...Self-propelled robot, 41A, 41B...Driving wheels, 42...Robot body, 43...Running base unit, 44...Arm unit, 45...Arm support unit, 46A, 46 B...Omni wheel, 47...Auxiliary wheels, 48A, 48B...Drive motor, 50...Base main body portion, 51...Tray, 52...Laser range sensor, 53...RGB-D sensor, 54...3D distance image sensor, 55...Imaging camera, 60...Rotation support portion, 61...First upper arm portion, 62...Second upper arm portion, 63...Forearm portion, 64...Wrist portion, 65A, 65B...Motor driver, 70...Driving control portion, 71...Driving drive portion, 72...Environmental map creation portion, 73...Target driving route memory portion, 74...Object detection portion, 75...Communication portion, 76...Driving battery.

Claims (4)

A self-propelled robot that is equipped with a non-destructive component estimation device that estimates sugar content by shining light on fruit and that moves on the ground autonomously or in response to external control.
a travel drive unit that drives the robot body to travel in a desired direction on the ground at a desired speed;
an arm support section that supports a base section of an arm section having a multi-joint mechanism with multiple degrees of freedom, and is connected to an upper section of the traveling drive section integrally with the arm section so as to be rotatable relative to the traveling drive section;
a travel control unit that drives and controls the travel drive unit and the arm support unit,
The non-destructive component estimation device comprises:
a light detection unit that detects scattered light generated from inside the fruit through a skin surface of the fruit when light is irradiated onto the fruit;
a contact detection unit provided adjacent to the light detection unit and configured to detect a contact state with the fruit;
an adsorption/light shielding portion having a tip made of an elastic material capable of being adsorbed to a surface of the skin of the fruit, the adsorption/light shielding portion covering the periphery of the light detection portion and the contact detection portion when the adsorption/light shielding portion is adsorbed to the fruit, thereby shielding the light detection portion from external light;
a light source that irradiates light so as to irradiate a vicinity of an adsorption portion of the fruit that is adsorbed to the adsorption light-shielding portion;
an image pickup distance measurement unit that captures an image of an object to be attracted by the attraction and light blocking unit and simultaneously measures a distance to the object;
a spectroscopic measurement unit that separates the scattered light detected by the light detection unit and measures a spectrum that represents a light intensity distribution for each wavelength;
a library in which spectra of the scattered light corresponding to a plurality of fruits of the same kind but different from the fruit are stored;
an estimation model constructing unit that constructs an estimation model that is trained to estimate a state of an internal component of the fruit by performing a multivariate analysis of correlation between the spectrum measured by the spectroscopic measurement unit and a plurality of the spectra stored in the library;
an internal state estimation unit that estimates a state of an internal component during a growth process of the fruit based on the estimation model constructed by the estimation model construction unit;
Equipped with
The traveling control unit is
Based on the image of the object to be attracted by the adsorbing and light-shielding part captured by the imaging distance measuring part and the measured distance to the object, the arm support part is rotated and each joint mechanism of the arm part is bent or extended to align the adsorbing and light-shielding part with the fruit to be attracted, and then the adsorbing and light-shielding part is caused to adhere to the outer surface of the fruit while detecting the contact state of the fruit by the contact detection part.
A self-propelled robot characterized by the above. An environment imaging unit that images an environment in which the robot body is traveling;
an object detection unit that detects a specific fruit tree in the driving environment based on an image captured by the environmental imaging unit;
an environmental map creation unit that estimates its own position with respect to the traveling environment and creates a planar or three-dimensional environmental map including the ground at the same time, wherein the traveling control unit controls the traveling drive unit to move to a specific fruit tree detected by the object detection unit in the environmental map created by the environmental map creation unit, and then drives the arm unit to bring the adsorbing and shading unit into contact with each of the multiple fruits growing on the fruit tree, and sequentially assigns addresses to the fruits,
The data representing the state of the internal components during the growth process of the fruit estimated by the internal state estimation unit is managed in units of the addresses.
2. The self-propelled robot according to claim 1 . the estimation model construction unit uses a supervised machine learning method as the estimation model,
The self-propelled robot according to claim 1 or 2, wherein the internal state estimation unit evaluates accuracy of the learning result obtained by the supervised machine learning method by cross-validation. The fruit is in a state where it is still attached to the fruit tree,
The self-propelled robot according to claim 1 or 2 , characterized in that the internal components of the fruit include at least one of sugar content, sourness, γ-aminobutyric acid, hardness, density, and color.

Images