in the previous positioning result belongs, the positioning device 100 can determine, as candidates, target objects 30 observable from the current position (strictly speaking, the most recent position) of the moving object 10, and determine the directions in which the respective observable target objects 30 are observable. By performing image recognition or the like and comparing the target object 30 that is in the image 60 with the determined candidates, the positioning device 100 can obtain an identification result of the target object 30 (i.e., 30a) that is in the image 60 and can detect the target object 30 (i.e., 30a) in the image 60. Further, the positioning device 100 can determine the in-space position of the identified target object 30 from the target object position data 134. The positioning device 100 may further refer to the orientation of the moving object 10 in the previous positioning result, thereby determining a direction in which the image 60 was captured, which further enhances the identification accuracy of the target object 30.

The positioning device 100 detects the mountaintop 30a in the image 60 shown in FIG. 3, namely, identifies the target object 30 in the image 60 as the mountaintop 30a, and determines the in-space position thereof as “(X21, Y21, Z21)”.

The method for identifying/detecting the target object 30 in the image 60 is not limited to the method described above.

For example, instead of using the area R, the positioning device 100 may determine, as candidates, target objects 30 that could be in the image 60 directly from the position (X, Y, Z coordinates) of the moving object 10 in the previous positioning result. For example, it is possible to create a 3D model of the space 500 in advance, and by simulation based on the 3D model, determine, as candidates, target objects 30 visible from the current position (most recent position) of the moving object 10. By comparing the image 60 with the candidates determined by the simulation, the positioning device 100 can obtain an identification result of each target object 30 that is in the image 60. Instead of the above 3D model, a map of the lunar surface including height information may be used to determine target objects 30 that could be in the image 60 as candidates.

Alternatively, each target object 30 that is actually in the image 60 May be identified by inputting the image 60 into a machine learning model. That is, it is possible to prepare a machine learning model that has learned positions, shapes, sizes and so forth of target objects 30 observable from a plurality of points on the lunar surface, and input the image 60 and the previous positioning result (position of the moving object 10 and, if necessary, orientation thereof too) into the machine learning model, thereby determining whether a target object(s) 30 as a candidate(s) is actually in the image 60, and if so, obtaining an identification result(s) of the target object(s) 30 as an output(s). The machine learning model may be stored in the storage 13, or may be stored in an external device. If the machine learning model is stored in an external device, data of the image 60 or the like is input to the machine learning model stored in the external device by a not-shown communicator that the moving object 10 has, and the result output from the external device is received thereby.

In either case, the in-space position of the identified target object 30 can be obtained from table data (which may be the target object position data 134) in which target objects 30 are associated with their respective in-space positions.

When detecting the light source 20a and the mountaintop 30a as position indicators in the image 60, the positioning device 100 derives the position and orientation of the moving object 10 in the space 500 on the basis of the in-image positions of the light source 20a and the mountaintop 30a in the image 60 and the in-space positions of the light source 20a and the mountaintop 30a in the space 500.

FIG. 6 is an illustration to explain a method for deriving the position (and orientation) of the moving object 10 in the space 500.

In this embodiment, the in-image position of the light source 20a in the image 60 is a point P1 (Px, Py), and the in-image position of the mountaintop 30a in the image 60 is a point Q1 (Qx, Qy). The positioning device 100 derives the in-space position (X01, Y01, Z01, wherein Z01 is fixed) of the moving object 10 (camera 14) in accordance with the principles of triangulation on the basis of the in-image position (point P1) of the light source 20a and the in-image position (point Q1) of the mountaintop 30a in the image 60 and the in-space position (X11, Y11, Z11) of the light source 20a and the in-space position (X21, Y21, Z21) of the mountaintop 30a in the space 500. The positioning device 100 further derives the orientation of the moving object 10 (camera 14) on the basis of the in-space position (X01, Y01, Z01) of the moving object 10 and the position(s) of the light source 20a and/or the mountaintop 30a in the image 60.

With this method, the positioning device 100 can derive the position and orientation (horizontal direction) of the moving object 10 (camera 14) on the assumption that the height and two degrees of freedom (pitch and roll) of the attitude angle of the moving object 10 (camera 14) are fixed. In the case where the attitude angle of the moving object 10 could vary due to, for example, a road surface being uneven, the positioning device 100 may derive the position and orientation of the moving object 10 that have been corrected on the basis of the detection result of the attitude angle by the attitude angle detector 17. For example, the positioning device 100 may derive the position of the moving object 10 after correcting the orientation or the like of the image 60 (positions of the light source 20a and the mountaintop 30a in the image 60) on the basis of the detection result of the attitude angle by the attitude angle detector 17.

The detection accuracy of the positions of the target objects 30 in the image 60 tends to be lower than the detection accuracy of the positions of the light sources 20 in the image 60. This is because, for example, the target objects 30 are not luminous objects and therefore difficult to distinguish from their surrounding objects in the image 60 and also because the accuracy of another process (e.g., image recognition) for identifying the target objects 30 affects the detection accuracy of the positions of the target objects 30 in the image 60. Therefore, in the case where a plurality of target objects 30 is detectable in the image 60 in addition to one light source 20, positioning may be performed using the one light source 20 and at least two target objects 30 among the plurality of target objects 30 as position indicators (i.e., using three or more position indicators), which enhances the positioning accuracy. For example, in the example shown in FIG. 3, in addition to the light source 20a and the mountaintop 30a as a target object 30, the lunar base 30c (e.g., a feature point of a corner thereof (in-image position: point Q2)) as a target object 30 may further be used for positioning.

FIG. 7 shows another example of the image 60.

The image 60 shown in FIG. 7 is an image obtained by photographing a portion of the space 500 in the daytime with one of the cameras 14, the portion being different from the portion shown in FIG. 3. In this image 60, the emission patterns of two light sources 20, i.e., 20a (in-image position: point P1) and 20d (in-image position: point P2), which are located in two shaded depressions 40, are observable, so that these light sources 20a, 20d are detectable in the image 60. With reference to the light source position data 133, the in-space positions of the light sources 20a, 20d are obtainable. The positioning device 100 derives the position and orientation of the moving object 10 on the basis of the in-image positions of the light sources 20a, 20d in the image 60 and the in-space positions of the light sources 20a, 20d in the space 500 with the same method described with reference to FIG. 6 except that the mountaintop 30a in FIG. 6 is replaced by the light source 20d, and the point Q1 in the image 60 in FIG. 6 is replaced by the point P2 corresponding to the light source 20d.

In the case where two light sources 20 are detectable in the image 60, a target object(s) 30, which is/are different from the light sources 20, may further be used as a position indicator(s). That is, positioning may be performed by using three or more position indicators including the two light sources 20 and a target object(s) 30. For example, in the image 60 shown in FIG. 7, the lunar base 30c (in-image position: point Q2) and the mountaintop 30b (in-image position: point Q3) as target objects 30 are detectable in addition to the light sources 20a, 20d. At least one of the lunar base 30c and the mountaintop 30b may further be used as a position indicator(s) in addition to the light sources 20a, 20d. This enhances the positioning accuracy as compared with the case of positioning using only two light sources 20.

By using three or more position indicators, positioning of more degrees of freedom is performable. That is, in the case where two position indicators are used, positioning of three degrees of freedom that are the horizontal position (X, Y coordinates) and the horizontal direction is performable, whereas in the case where three or more position indicators are used, positioning of four or more degrees of freedom that are the abovementioned three degrees of freedom and at least one of the coordinate in the vertical direction (Z coordinate) and two degrees of freedom (pitch and roll) of the attitude angle is performable. Ideally, six position indicators are used, so that positioning of six degrees of freedom that are the three-dimensional coordinates (horizontal position and height) and the three-axis attitude angle (yaw (horizontal direction), pitch and roll) is performable. If seven or more position indicators are detected, light sources 20 are preferentially used as position indicators, which further enhances the positioning accuracy.

Next, a positioning-and-moving process that is performed by the CPU 11 for moving the moving object 10 while performing the above-described positioning will be described.

FIG. 8 is a flowchart showing a control procedure of the positioning-and-moving process.

The positioning-and-moving process is started when the moving object 10 starts to operate on the lunar surface.

When the positioning-and-moving process is started, the CPU 11 obtains the image 60 captured by the camera 14 (Step S101). The CPU 11 determines whether it is nighttime (Step S102). In this embodiment, the CPU 11 determines whether it is nighttime on the basis of the current time being measured by the time measurer 15. Alternatively, the CPU 11 May determine whether it is nighttime on the basis of the brightness of the image 60. For example, if the average brightness of the pixels of the image 60 is equal to or less than a predetermined reference brightness, the CPU 11 determines that it is nighttime.

If the CPU 11 determines that it is nighttime

(Step S102; YES), the CPU 11 attempts to detect light sources 20 in the image 60 with the method described above, and determines whether the CPU 11 has detected a plurality of light sources 20 in the image 60 (Step S103). If the CPU 11 determines that the CPU 11 has detected a plurality of light sources 20 in the image 60 (Step S103; YES), the CPU 11 performs positioning, namely, derives the position and orientation of the moving object 10, on the basis of the in-image positions and the in-space positions of two or more light sources 20 among the plurality of light sources 20 detected in the image 60, and records the positioning result in the positioning history data 135 (Step S104).

If the CPU 11 determines that the CPU 11 has not detected a plurality of light sources 20 in the image 60 (Step S103; NO), the CPU 11 transmits a control signal to the movement driver 16 to cause the moving object 10 to stop moving (Step S105). This is because the moving object 10 moving around without knowing its own position and/or orientation may greatly deviate from a route to a destination or may cause an unforeseen accident. After Step S105, the CPU 11 returns the process to Step S101.

In Step S102, if the CPU 11 determines that it is not nighttime, namely, it is daytime (Step S102; NO), the CPU 11 attempts to detect light sources 20 in the image 60, and determines whether the CPU 11 has detected a plurality of light sources 20 in the image 60 (Step S106). If the CPU 11 determines that the CPU 11 has detected a plurality of sources 20 in the image 60 (Step S106; YES), the CPU 11 moves the process to Step S104 to perform positioning using two or more light sources 20.

If the CPU 11 determines that the CPU 11 has not detected a plurality of light sources 20 in the image 60

(Step S106; NO), the CPU 11 determines whether a previous positioning result is recorded (stored) in the positioning history data 135 (Step S107). If the CPU 11 determines that no previous positioning result is recorded therein (Step S107; NO), the CPU 11 moves the process to Step S105 to cause the moving object 10 to stop moving because the CPU 11 cannot obtain the current position of the moving object 10, which is needed to identify/detect target objects 30 in the image 60.

If the CPU 11 determines that a previous positioning result is recorded therein (Step S107; YES), the CPU 11 determines whether the CPU 11 has detected one light source 20 in the image 60 (Step S108). If the CPU 11 determines that the CPU 11 has detected one light source 20 in the image 60 (Step S108; YES), the CPU 11 determines and obtains, as candidates, target objects 30 photographable with the camera 14 on the basis of the previous positioning result and the current time being measured by the time measurer 15 (Step S109).

The CPU 11 attempts to detect, in the image 60, target objects 30 matching the candidates (obtain identification results of target objects 30), and determines whether the CPU 11 has detected one or more target objects 30 in the image 60 (Step S110). If the CPU 11 determines that the CPU 11 has detected no target object 30 in the image 60 (Step S110; NO), the CPU 11 moves the process to Step S105 to cause the moving object 10 to stop moving. If the CPU 11 determines that the CPU 11 has detected one or more target objects 30 in the image 60 (Step S110; YES), the CPU 11 performs positioning on the basis of the in-image positions and the in-space positions of the one light source 20 determined in Step S108 and the one or more target objects 30 determined in Step S110, and records the positioning result in the positioning history data 135 (Step S111).

In Step S108, if the CPU 11 determines that the CPU 11 has detected no light source 20 in the image 60 (Step S108; NO), the CPU 11 determines and obtains, as candidates, target objects 30 photographable with the camera 14 (Step S112) as in Step S109.

The CPU 11 attempts to detect, in the image 60, target objects 30 matching the candidates, and determines whether the CPU 11 has detected two or more target objects 30 in the image 60 (Step S113). If the CPU 11 determines that the CPU 11 has not detected two or more target objects 30 in the image 60 (Step S113; NO), the CPU 11 moves the process to Step S105 to cause the moving object 10 to stop moving. If the CPU 11 determines that the CPU 11 has detected two or more target objects 30 in the image 60 (Step S113; YES), the CPU 11 performs positioning on the basis of the in-image positions and the in-space positions of the two or more target objects 30 determined in Step S113, and records the positioning result in the positioning history data 135 (Step S114).

After Step S114, the CPU 11 reduces the moving speed of the moving object 10 to be lower than a predetermined reference value (Step S115). The positioning accuracy of positioning using only target objects 30 as position indicators is lower than the positioning accuracy of positioning using two light sources 20 as position indicators. If the moving object 10 moves on the basis of the positioning result of positioning using only target objects 30 as position indicators, the moving object 10 is likely to greatly deviate from a planned route. For this reason, in Step S115, the CPU 11 reduces the moving speed of the moving object 10, thereby suppressing increase in the deviation. If the moving speed is already lower than the reference value when the CPU 11 is about to start Step 115, the CPU 11 skips Step S115.

After Step S111 too, the CPU 11 May move the process to Step S115 to reduce the moving speed. That is, the CPU 11 may reduce the moving speed of the moving object 10 whenever the CPU 11 performs positioning using one or more target objects 30 as position indicators. If the CPU 11 performs positioning using not only one or more target objects 30 but also two or more light sources 20 as position indicators, the CPU 11 May not reduce the moving speed of the moving object 10.

After Step S104, Step S111 or Step S115, the CPU 11 determines a route to a destination on the basis of the positioning result, and causes the moving object 10 to move along the route (Step S116). The route to the destination is set such that a plurality of light sources 20 is observable from any point on the route. The CPU 11 determines whether the moving object 10 has arrived at the destination (Step S117). If the CPU 11 determines that the moving object 10 has not arrived at the destination (Step S117; NO), the CPU 11 returns the process to Step S101. If the CPU 11 determines that the moving object 10 has arrived at the destination (Step S117; YES), the CPU 11 ends the positioning-and-moving process.

Instead of the flowchart shown in FIG. 8, the CPU 11 May follow the flowchart shown in FIG. 9. The flowchart shown in FIG. 9 is a flowchart with which the CPU 11 uses target objects 30 as position indicators if the CPU 11 detects these in the image 60 no matter whether the CPU 11 detects light sources 20 in the image 60. This enhances the positioning accuracy to the best possible positioning accuracy according to the detection state of light sources 20 and target objects 30 in the image 60.

Step S201 to Step S205 of the positioning-and-moving process in FIG. 9 are the same as Step S101 to Step S105 in FIG. 8, and hence their descriptions will be omitted.

In Step S202, if the CPU 11 determines that it is not nighttime (Step S202; NO), the CPU 11 determines whether a previous positioning result is recorded in the positioning history data 135 (Step S206). If the CPU 11 determines that no previous positioning result is recorded therein (Step S206; NO), the CPU 11 moves the process to Step S203. If the CPU 11 determines that a previous positioning result is recorded therein (Step S206; YES), the CPU 11 determines and obtains, as candidates, target objects 30 photographable with the camera 14 on the basis of the previous positioning result and the current time being measured by the time measurer 15 (Step S207).

The CPU 11 attempts to detect, in the image 60, target objects 30 matching the candidates, and determines whether the CPU 11 has detected one or more target objects