System and method for spatial analysis of robot

Abstract

The present invention relates to a system and method for the spatial analysis of a robot. The system for the spatial analysis of a robot according to the present invention comprises: a LiDAR mounted on the robot; a camera mounted on the robot; and a spatial analysis unit that creates a spatial map by using LiDAR data generated from the LiDAR and estimates the location of the robot in space. The spatial analysis unit uses camera data, generated by the camera, to filter out LiDAR data generated from detecting smoke from the LiDAR data.

Description

Robot spatial analysis system and method

[0001] The present invention relates to a space analysis system and method for a robot, and more particularly, to a space analysis system and method for a robot capable of accurately identifying the space in which a robot is located in an invisible environment and the robot's location within that space.

[0002] There are ongoing efforts to use robots to work in environments that are difficult for humans to access.

[0003] Because there is often no information about the space (environment) in which the robot's work takes place, the robot can work in an unknown space by utilizing, for example, Simultaneous Localization and Mapping (SLAM) technology.

[0004] However, when visibility is difficult due to smoke or fog, it is not easy for the robot to obtain information about the space and its location within the space.

[0005] A method of combining sensors that use different wavelengths for measurement can be used to obtain information about space in an invisible environment.

[0006] For example, combining radar and lidar sensors can provide spatial information in non-visible environments. Specifically, by comparing radar and lidar measurements, if there are discrepancies, the lidar can be determined to have detected smoke or fog. The radar measurements can then be used as data for spatial analysis.

[0007] However, because the resolution of the radar sensor is lower than that of the lidar sensor, it can only partially determine whether the lidar has detected smoke, etc., and as a result, it is not easy to accurately obtain information about the space.

[0008] [Prior Art Literature]

[0009] [Patent Document]

[0010] (Patent Document 1) KR10-2069094 B1

[0011] Accordingly, the purpose of the present invention is to solve such conventional problems, and to provide a space analysis system and method for a robot that can accurately grasp information about a space even in an invisible environment due to smoke, etc.

[0012] The problems to be solved by the present invention are not limited to the problems mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below.

[0013] The above object is achieved by a space analysis system for a robot, according to the present invention, comprising: a lidar mounted on a robot; a camera mounted on the robot; and a space analysis unit that creates a space map using lidar data generated from the lidar and estimates the position of the robot within the space; wherein the space analysis unit filters lidar data that detects smoke among the lidar data using camera data generated from the camera.

[0014] The above spatial analysis unit may use the lidar data of the corresponding location to create a spatial map when it is confirmed that there is no smoke at a specific location in the camera data, and may not use the lidar data of the corresponding location to create a spatial map when it is confirmed that there is smoke at a specific location in the camera data.

[0015] The above spatial analysis unit can determine the presence or absence of smoke by comparing the clarity of a specific location with another location in the camera data.

[0016] The above other location may be a location further from the robot than the above specific location.

[0017] The above spatial analysis unit can determine the presence or absence of smoke by learning images without smoke and images with smoke.

[0018] The space analysis system of a robot according to the present invention further includes a 3D scanner mounted on the robot, and the space analysis unit can create a space map using filtered lidar data, the camera data, and scanner data generated from the 3D scanner.

[0019] The above spatial analysis unit may include a first learning unit that generates a densified model by inputting filtered lidar data and the camera data and learning using the scanner data as correct data, a first model storage unit in which the densified model is stored, and a first estimation unit that inputs the filtered lidar data and the camera data into the densified model to generate estimated scanner data.

[0020] The above spatial analysis unit may be equipped with an estimated lidar data generation unit that generates estimated lidar data of a blank point where the lidar data value is not generated.

[0021] The above-mentioned estimated lidar data generation unit can generate the estimated lidar data using lidar data generated at a time before and / or after the generation of the blank point.

[0022] The above-mentioned estimated lidar data generation unit may include a second learning unit that generates an estimated model for generating the estimated lidar data by learning the T-1 to Tn (n is a natural number)-th lidar data and / or the T+1 to T+m (m is a natural number)-th lidar data using the T-th lidar data as correct data; a second model storage unit in which the estimated model is stored; and a second estimation unit that inputs the t-1 to tn (n is a natural number)-th lidar data and / or the t+1 to t+m (m is a natural number)-th lidar data into the estimation model when the blank point is included in the t-th data generated from the lidar, thereby generating the estimated lidar data.

[0023] The second estimation unit may input the LIDAR data including the generated estimated LIDAR data into the estimation model to generate the estimated LIDAR data again, if the blank point is included in each of the data continuously generated from the LIDAR.

[0024] The above-mentioned estimated lidar data generation unit can generate the estimated lidar data using the lidar data and the camera data generated before and / or after the generation of the blank point, and the camera data at the time of the generation of the blank point.

[0025] According to another embodiment of the present invention, a method for analyzing space of a robot is provided, including: a data generation step of generating lidar data and camera data from a lidar and a camera mounted on a robot; a data filtering step of filtering lidar data by detecting smoke among the lidar data using the camera data; and a spatial analysis step of creating a spatial map and estimating the position of the robot within the space using the lidar data filtered in the data filtering step.

[0026] The spatial analysis method of a robot according to the present invention may further include a blank point confirmation step for confirming whether a blank point, which is a point where a lidar data value is not generated, is included in the lidar data, which is performed between the data generation step and the data filtering step; and a lidar data estimation step for generating estimated lidar data of the blank point using lidar data generated at a time before and / or after the generation of the blank point.

[0027] According to the space analysis system of the robot of the present invention, it is possible to create a space map accurately and in detail even in an environment where smoke or the like exists.

[0028] Accordingly, the robot can smoothly perform its given task even in an invisible environment.

[0029] When 3D scanner data, in addition to camera and lidar data, is learned through artificial intelligence for spatial analysis, it is possible to create more accurate and dense spatial maps in real time.

[0030] And even when some lidar data values ​​are not generated, the accuracy of spatial analysis can be improved by estimating lidar data values ​​using artificial intelligence.

[0031] Figure 1 is a schematic diagram of a spatial analysis system for a robot according to the present invention.

[0032] Figure 2 is a schematic diagram of a spatial analysis system according to another embodiment of the present invention;

[0033] Figures 3 and 4 are explanatory diagrams of the operation of the space analysis unit when the space analysis system of the robot according to the present invention further includes a 3D scanner.

[0034] Figures 5 to 8 are explanatory diagrams of the operation of the estimated lidar data generation unit constituting the space analysis system of the robot according to the present invention.

[0035] Figure 9 is a flowchart of a space analysis method of a robot according to the present invention.

[0036] Hereinafter, specific embodiments of the present invention will be described in detail with reference to the drawings.

[0037]

[0038] Figure 1 shows a schematic configuration diagram of a space analysis system (1) for a robot according to the present invention.

[0039] The space analysis system (1) of a robot according to the present invention is largely composed of a lidar (10), a camera (20), and a space analysis unit (30).

[0040] Among the components of the spatial analysis system (1) for a robot according to the present invention, at least a lidar (10) and a camera (20) are mounted on the robot, so that the spatial analysis system (1) for a robot according to the present invention can operate in real time for a moving robot. The robots to which the present invention is applied may include all robots capable of movement, such as bipedal walking robots, quadruped walking robots, and autonomous vehicles.

[0041] LiDAR (10) emits laser pulses and measures the distance to objects by receiving the light reflected from surrounding objects, thereby determining the space in which the robot is located. LiDAR (10) can be mounted on top of the robot, for example.

[0042] The camera (20) captures images of the robot's surroundings to determine the space in which the robot is located. Like the lidar (10), the camera (20) can be mounted on top of the robot and configured to capture images of all sides of the robot.

[0043] The spatial analysis unit (30) may be equipped with a sensing data storage unit (31), a filtering unit (32), and an analysis unit (33).

[0044] The sensing data storage unit (31) stores lidar data generated from the lidar (10) and camera data generated from the camera (20). The lidar data and the camera data can be stored while being temporally and spatially synchronized. Since the lidar data and the camera data have different resolutions, when spatially synchronized, the lidar data representing a point can be matched to a portion (having an area) that includes the point.

[0045] The filtering unit (32) filters the lidar data by detecting smoke and other objects using camera data. While lidar (10) detects smoke and fog, it cannot distinguish between smoke and objects (including building structures). Therefore, if lidar data is used as is in an environment with smoke and other objects, it is impossible to accurately create a spatial map. Therefore, the lidar data is filtered using camera data that can distinguish between smoke and other objects.

[0046] The analysis unit (33) creates a spatial map using the filtered lidar data and estimates the position of the robot within the space, i.e., performs SLAM (Simultaneous Localization and Mapping).

[0047]

[0048] According to the space analysis system (1) of the robot of the present invention, it is possible to create a space map accurately and in detail even in an environment where smoke, etc. exists by using lidar data that can create a space map in relatively detail and filtering out data unnecessary for creating a space map using camera data.

[0049] Accordingly, the robot can smoothly perform its given task even in an invisible environment.

[0050]

[0051] The filtering unit (32) of the spatial analysis unit (30) operates more specifically as follows.

[0052] That is, if the camera data confirms that there is no smoke (including fog) at a specific location, the lidar data from the corresponding location is used to create a spatial map. If the smoke is confirmed at a specific location, the lidar data from the corresponding location is not used to create a spatial map. Note that the specific location refers to the location to be confirmed for smoke.

[0053] As mentioned above, since lidar data and camera data can be spatially synchronized, if smoke is identified in the camera data, it is possible to accurately exclude the lidar data caused by the smoke from the spatial map.

[0054]

[0055] The filtering unit (32) can determine the presence or absence of smoke by comparing the clarity of a specific location with another location in camera data, for example.

[0056] Smoke is usually white or gray in color and appears hazy as it disperses and forms in the air. As a result, the overall clarity of the image captured by the camera (20) is reduced in the areas containing smoke. Accordingly, the clarity of the areas containing smoke is lower than that of the areas without smoke.

[0057] Consequently, if the clarity at a particular location is lower than at other locations, it can be determined that there is smoke at that particular location.

[0058]

[0059] Other locations may be further away from the robot than a specific location.

[0060] The smoke will be located in the space between the robot and the wall (or object), and the distance between the robot and the wall will usually be longer than the distance between the robot and the smoke.

[0061] Therefore, when determining the presence or absence of smoke at a specific location, the clarity of another location that is located further away from the specific location in the lidar data and thus has a relatively higher probability of being a wall than the specific location is compared with the clarity of the specific location.

[0062] The farthest location in the lidar data can be assumed to be a wall.

[0063]

[0064] The filtering unit (32) can also determine the presence or absence of smoke through artificial intelligence learning.

[0065] That is, after training a filtering model with a number of camera images without smoke and a number of camera images with smoke, the presence or absence of smoke can be confirmed by inputting camera data captured by the camera (20) in real time into the filtering model.

[0066]

[0067] The space analysis system (1) of a robot according to the present invention may further include a 3D scanner (40). An explanatory diagram for such a case is illustrated in Fig. 2.

[0068] The 3D scanner (40) is mounted on the robot, similar to the lidar (10) and camera (20), and serves to create a high-density spatial map. The camera (20), which generates camera data, may form part of the 3D scanner (40). That is, the 3D scanner (40) can create a high-density spatial map by applying the camera data to a NeRF (Neural Radiance Field) deep learning model.

[0069] When the space analysis system (1) of a robot according to the present invention further includes a 3D scanner (40), it is possible to create a high-density, accurate space map in real time using scanner data, filtered lidar data, and camera data generated from the 3D scanner (40). While scanner data from the 3D scanner (40) may have high density, its accuracy and real-time performance may be low. However, by using filtered lidar data and camera data in addition to the scanner data, a high-density, accurate space map can be created in real time.

[0070] Specifically, the spatial analysis unit (30) may further include a first learning unit (34), a first model storage unit (35), and a first estimation unit (36).

[0071] The first learning unit (34), as illustrated in FIG. 3, generates a high-density model by inputting filtered lidar data and camera data and learning using scanner data as correct data. For reference, in FIG. 3, the space represented by the scanner data is expressed differently from the space represented by the filtered lidar data and camera data, but it is obvious that the filtered lidar data, camera data, and scanner data should all be for the same space. Learning can proceed until the data inferred using the filtered lidar data and camera data has a similarity level higher than a certain level with the scanner data. The first learning unit (34) can proceed with learning using deep learning technology.

[0072] The first model storage unit (35) stores the high-density model created in the first learning unit (34).

[0073] The first estimation unit (36) inputs filtered lidar data and camera data generated in real time into a high-density model, as illustrated in FIG. 4, to generate estimated scanner data. Although the estimated scanner data is not scanner data directly generated by the 3D scanner (40), it has high-density characteristics like the scanner data, and like the lidar data and camera data, it has high accuracy and real-time formation characteristics.

[0074]

[0075] The spatial analysis unit (30) may be equipped with an estimated lidar data generation unit (37).

[0076] Since lidar data is generated by a laser pulse that is reflected and returned, lidar data values ​​may not be generated at some points if the laser pulse is not received by the lidar (10) depending on the reflection characteristics of the laser pulse.

[0077] The estimated lidar data generation unit (37) plays a role in generating estimated lidar data for a blank point, which is a point where no lidar data value is generated.

[0078] It is possible to create an accurate spatial map even if lidar data values ​​at some points are not generated by estimated lidar data.

[0079] The estimated lidar data generation unit (37) can generate estimated lidar data using lidar data generated before and / or after the generation of a blank point.

[0080] More specifically, the estimated lidar data generation unit (37) may include a second learning unit (37a), a second model storage unit (37b), and a second estimation unit (37c).

[0081] As illustrated in FIG. 5, the second learning unit (37a) learns the T-1 to Tn (n is a natural number)-th lidar data and / or the T+1 to T+m (m is a natural number)-th lidar data using the T-th lidar data as the correct data, thereby generating an estimation model for generating the estimated lidar data. The learning may proceed until the T-th lidar data and the inferred data have a similarity greater than a certain value. The second learning unit (37a) may proceed with the learning using deep learning technology. FIG. 5 illustrates an example in which the T-1 to Tn-th lidar data are learned to generate an estimation model.

[0082] The second model storage unit (37b) stores the estimated model generated in the second learning unit (37a).

[0083] The second estimation unit (37c) inputs the lidar data generated in real time into the estimation model to generate the estimated lidar data, as illustrated in FIG. 6. The second estimation unit (37c) can operate when the t-th generated data (currently generated data) from the lidar (10) includes a blank point, and can input the t-1 to tn-th lidar data and / or the t+1 to t+m-th lidar data into the estimation model to generate the estimated lidar data. When the second learning unit (37a) uses the T-1 to Tn-th lidar data to generate the estimation model, the second estimation unit (37c) also inputs the t-1 to tn-th lidar data into the estimation model to generate the estimated lidar data, and when the second learning unit (37a) uses the T+1 to T+m-th lidar data to generate the estimation model, the second estimation unit (37c) also inputs the t+1 to t+m-th lidar data into the estimation model to generate the estimated lidar data. And when the T-1 to Tn-th lidar data and the T+1 to T+m-th lidar data are used to generate the estimation model in the second learning unit (37a), the T-1 to Tn-th lidar data and the T+1 to T+m-th lidar data are also input into the estimation model in the second estimation unit (37c) to generate the estimation lidar data. In order to increase the real-time performance of the estimation lidar data generation, it is preferable to use the T-1 to Tn-th lidar data to generate the estimation model.

[0084] The second estimation unit (37c) can input the LIDAR data including the generated estimated LIDAR data into the estimation model to generate the estimated LIDAR data again when each of the data continuously generated from the LIDAR (10) includes a blank point. That is, for example, when the t-th and t+1-th LIDAR data include a blank point, the t-1 to tn-th LIDAR data are input into the estimation model to generate the t-th estimated LIDAR data, and again the t-th estimated LIDAR data and the t-1 to t-(n-1)-th LIDAR data are input into the estimation model to generate the t+1-th estimated LIDAR data.

[0085] Accordingly, it is possible to generate estimated lidar data even when blank points are included in continuously generated lidar data.

[0086]

[0087] The estimated lidar data generation unit (37) can generate estimated lidar data using camera data generated at a time before and / or after the creation of the blank point in addition to lidar data generated at a time before and / or after the creation of the blank point and camera data generated at a time before and / or after the creation of the blank point.

[0088] In this case, it is possible to infer the estimated lidar data more accurately, and as a result, the spatial analysis unit (30) can create a spatial map more accurately.

[0089] When further utilizing camera data in generating estimated lidar data, a dataset constructed in which lidar data and camera data are temporally and spatially synchronized can be used. Specifically, as illustrated in FIG. 7, the second learning unit (37a) uses the T-th dataset (lidar data and camera data) as the correct data, and generates an estimation model by learning the T-th camera data and the T-1 to Tn-th dataset and / or the T+1 to T+m-th dataset. In addition, as illustrated in FIG. 8, in the second estimation unit (37c), when the t-th generated data includes a blank point, the t-th estimated dataset can be generated by inputting the t-th camera data and the t-1 to tn-th dataset and / or the t+1 to t+m-th dataset into the estimation model. For reference, the t-th camera data may be data for a location corresponding to a blank point in the lidar data and data for the area around the location corresponding to the blank point in terms of space.

[0090] In Fig. 7, an example is shown in which an estimation model is created by learning the T-th camera data and the T-1 to Tn-th data sets.

[0091]

[0092] Hereinafter, a spatial analysis method for a robot according to the present invention will be described. While describing the spatial analysis method for a robot according to the present invention, a detailed description of matters mentioned in the description of the spatial analysis system (1) for a robot according to the present invention may be omitted.

[0093] Figure 9 shows a flowchart of a space analysis method of a robot according to the present invention.

[0094] The space analysis method of a robot according to the present invention comprises a data generation step (S10), a data filtering step (S40), and a space analysis step (S50).

[0095] In the data generation step (S10), lidar data and camera data are generated from the lidar (10) and camera (20) mounted on the robot. The generated lidar data and camera data can be temporally and spatially synchronized and stored.

[0096] In the data filtering step (S40), camera data is used to filter out lidar data that detects smoke among the lidar data. That is, in an environment with smoke, lidar data can detect smoke that is unnecessary for creating a spatial map in addition to detecting objects (including building structures such as walls) necessary for creating a spatial map, so lidar data that detects smoke is filtered. Since smoke can be distinguished in images captured by a camera (20), lidar data that detects smoke can be filtered using camera data.

[0097] Filtering of lidar data can be done by comparing the clarity of a specific location in the camera data to another location to check for the presence of smoke, or by creating a filtering model by learning a number of camera images with and without smoke in advance and then inputting the camera data into the filtering model in real time.

[0098] In the spatial analysis step (S50), the filtered lidar data from the data filtering step (S40) is used to create a spatial map and estimate the robot's position within the space. Because only lidar data from objects is used for spatial map creation and other tasks, filtering the lidar data ensures accurate spatial map creation and other tasks.

[0099]

[0100] In the spatial analysis step (S50), spatial analysis can be performed using camera data in addition to LiDAR data. In this case, the robot further includes a 3D scanner (40), and performs artificial intelligence learning on scanner data, LiDAR data, and camera data generated from the 3D scanner (40) in advance, and generates a densified model capable of generating estimated scanner data by inputting LiDAR data and camera data. Then, the LiDAR data and camera data generated in real time are input into the densified model to generate estimated scanner data. The estimated scanner data is spatially accurate and densely formed, and can be generated quickly in real time.

[0101]

[0102] The space analysis method of a robot according to the present invention may further include a blank point confirmation step (S20) and a lidar data estimation step (S30).

[0103] The blank point confirmation step (S20) and the lidar data estimation step (S30) can be performed between the data generation step (S10) and the data filtering step (S40).

[0104] In the blank point verification step (S20), it is checked whether the lidar data includes a blank point, which is a point where no lidar data value is generated. The blank point can be verified, for example, by comparing the laser emitted by the lidar (10) with the laser reflected back.

[0105] In the lidar data estimation step (S30), estimated lidar data of the blank point is generated using lidar data generated before and / or after the creation of the blank point. To estimate the lidar data, the lidar data of a specific point in time is used as the correct data in advance, and an estimation model is generated by learning a predetermined number of lidar data before and / or after the specific point in time, and when the blank point is confirmed in real time, a predetermined number of lidar data before and / or after the creation of the blank point is input into the estimation model to generate estimated lidar data.

[0106] To generate estimated LiDAR data, in addition to LiDAR data generated before and / or after the blank point generation, camera data generated before and / or after the blank point generation and camera data at the time of blank point generation may be used. In this case, it is natural that LiDAR data and camera data should be trained together to generate an estimation model, and LiDAR data and camera data should be input together into the estimation model to generate real-time estimated LiDAR data.

[0107]

[0108] The scope of the present invention is not limited to the embodiments described above, but can be implemented in various forms within the scope of the appended claims. Any person skilled in the art, without departing from the spirit of the invention as claimed in the claims, may make various modifications to the invention, which are deemed to fall within the scope of the claims.

[0109]

[0110] [Explanation of symbols]

[0111] 1: Robot spatial analysis system

[0112] 10: Lidar

[0113] 20: Camera

[0114] 30: Spatial Analysis Department

[0115] 34: First Learning Department

[0116] 35: Model 1 storage unit

[0117] 36: First Estimate

[0118] 37: Estimated lidar data generation unit

[0119] 37a: Second Study Group

[0120] 37b: Second model storage unit

[0121] 37c: Second Estimate

[0122] 40: 3D scanner

Claims

1. Lidar mounted on the robot; Camera mounted on the robot; and A spatial analysis unit is included that creates a spatial map using lidar data generated from the lidar and estimates the location of the robot within the space; The above spatial analysis unit, A space analysis system for a robot characterized in that the lidar data is filtered by detecting smoke among the lidar data using camera data generated from the above camera.

2. In paragraph 1, The above spatial analysis unit, If it is confirmed that there is no smoke at a specific location from the above camera data, the lidar data of the corresponding location is used to create a spatial map. A spatial analysis system for a robot, characterized in that if smoke is identified at a specific location in the above camera data, the lidar data of the corresponding location is not used to create a spatial map.

3. In paragraph 2, The above spatial analysis unit, A space analysis system for a robot characterized in that it determines the presence or absence of smoke by comparing the clarity of a specific location with another location in the above camera data.

4. In paragraph 3, The other locations above are, A space analysis system for a robot, characterized in that the location is further away from the specific location based on the robot.

5. In paragraph 2, The above spatial analysis unit, A spatial analysis system for a robot characterized by determining the presence or absence of smoke through learning images with and without smoke.

6. In paragraph 1, It also includes a 3D scanner that is attached to the robot, The above spatial analysis unit, A spatial analysis system for a robot, characterized in that it creates a spatial map using filtered lidar data, the camera data, and scanner data generated from the 3D scanner.

7. In paragraph 6, The above spatial analysis unit, A first learning unit that creates a high-density model by inputting filtered lidar data and the camera data and learning using the scanner data as the correct data, A first model storage unit in which the above high-density model is stored, and A space analysis system for a robot, characterized in that it comprises a first estimation unit that inputs the filtered lidar data and the camera data into the high-density model to generate estimated scanner data.

8. In paragraph 1, The above spatial analysis unit, A space analysis system for a robot, characterized by having an estimated lidar data generation unit that generates estimated lidar data of a blank point where the above lidar data value is not generated.

9. In paragraph 7, The above estimated lidar data generation unit, A spatial analysis system for a robot characterized in that the estimated lidar data is generated using lidar data generated before and / or after the generation of the blank point.

10. In paragraph 9, The above estimated lidar data generation unit, A second learning unit that generates an estimation model for generating the estimated lidar data by learning the T-1 to Tn (n is a natural number) lidar data and / or the T+1 to T+m (m is a natural number) lidar data using the T-th lidar data as the correct data, A second model storage unit in which the above-mentioned estimated model is stored, and A space analysis system for a robot, characterized in that it comprises a second estimation unit that inputs the t-1 to tn (n is a natural number) th lidar data and / or the t+1 to t+m (m is a natural number) th lidar data into the estimation model when the blank point is included in the t th generated data from the lidar.

11. In paragraph 10, The second estimation part above is, A space analysis system for a robot, characterized in that when each of the data continuously generated from the lidar includes the blank point, the lidar data including the generated estimated lidar data is input into the estimation model to generate the estimated lidar data again.

12. In paragraph 9, The above estimated lidar data generation unit, A space analysis system for a robot characterized in that the estimated lidar data is generated using the lidar data and the camera data generated before and / or after the creation of the blank point, and the camera data at the time of creation of the blank point.

13. Data generation step for generating lidar data and camera data from lidar and camera mounted on the robot; A data filtering step for filtering lidar data by detecting smoke among the lidar data using the above camera data; and A method for spatial analysis of a robot, comprising a spatial analysis step of creating a spatial map using the filtered lidar data in the above data filtering step and estimating the position of the robot within the space.

14. In paragraph 13, As performed between the above data generation step and the above data filtering step, A blank point verification step for checking whether the above lidar data includes a blank point where no lidar data value is generated; and A method for spatial analysis of a robot, further comprising a lidar data estimation step of generating estimated lidar data of the blank point using lidar data generated before and / or after the generation of the blank point.

Images