Method for navigating a robot, navigation system and robot

Abstract

Computer-implemented method 100 for navigating a robot in an environment includes: Step 402, constructing a point cloud map representing the environment, based on 3D sensor data. Step 404, determining

Description

TECHNICAL FIELD

[0001] Various embodiments relate to methods for navigating a robot in an environment, navigation systems and robots. BACKGROUND

[0002] There is an increasing demand for autonomous mobile robots, which can be deployed to perform tasks such as delivery of goods. To meet the demand, there is a need for autonomous navigation technologies that can perform well in urban environments. Navigation technologies using two-dimensional (2D) floor maps have been able to accomplish most indoor navigation tasks. However, 2D navigation strategies are difficult to apply in outdoor environments due to factors such as lighting, road conditions, weather, and irregular terrain. In view of the above, there is a need for an improved navigation method that can address at least some of the abovementioned problems. SUMMARY

[0003] According to various embodiments, a computer-implemented method for navigating a robot in an environment may be provided. The method may include constructing a point cloud map representing the environment, based on 3D sensor data. The method may further include determining pose of the robot based on the 3D sensor data, generating an elevation map based on the point cloud map, generating a traversability map based on the elevation map, and planning a motion path for the robot based on the traversability map.

[0004] According to various embodiments, a computer program product may be provided. The computer program product may include instructions, which, when the computer program product is executed by a computer, cause the computer to carry out the abovementioned method for navigating a robot.

[0005] According to various embodiments, a data carrier signal may be provided. The data carrier signal may carry the abovementioned computer program product.

[0006] According to various embodiments, a navigation system for navigating a robot in an environment may be provided. The navigation system may include a mapping module configured to construct a point cloud map based on 3D sensor data, a pose module configured to determine pose of the robot based on the 3D sensor data, an elevation module configured to generate an elevation map based on the point cloud map, a traversability module configured to generate a traversability map based on the elevation map, and a path planner configured to plan a motion path for the robot based on the traversability map.

[0007] According to various embodiments, a robot may be provided. The robot may include the above-mentioned navigation system.

[0008] Additional features for advantageous embodiments are provided in the dependent claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In the drawings, like reference characters generally refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead generally being placed upon illustrating the principles of the invention. In the following description, various embodiments are described with reference to the following drawings, in which:

[0010] FIG. lisa simplified schematic diagram that illustrates a method for navigating a robot in an environment, according to various embodiments

[0011] FIG. 2 shows a block diagram of an implementation of the method for navigating a robot in an environment.

[0012] FIGS. 3Ato3C show maps used in the experiment to validate the method for navigating a robot in an environment.

[0013] FIG. 4 shows a flow diagram of the method for navigating a robot in an environment, according to various embodiments.

[0014] FIG. 5 shows a block diagram of a navigation system for navigating a robot in an environment according to various embodiments.

[0015] FIG. 6 shows a block diagram of a robot according to various embodiments. DESCRIPTION

[0016] Embodiments described below in context of the devices are analogously valid for the respective methods, and vice versa. Furthermore, it will be understood that the embodiments described below may be combined, for example, a part of one embodiment may be combined with a part of another embodiment.

[0017] It will be understood that any property described herein for a specific device may also hold for any device described herein. It will be understood that any property described herein for a specific method may also hold for any method described herein. Furthermore, it will be understood that for any device or method described herein, not necessarily all the components or steps described must be enclosed in the device or method, but only some (but not all) components or steps may be enclosed.

[0018] The term “coupled” (or “connected”) herein may be understood as electrically coupled or as mechanically coupled, for example attached or fixed, or just in contact without any fixation, and it will be understood that both direct coupling or indirect coupling (in other words: coupling without direct contact) may be provided.

[0019] In this context, the device as described in this description may include a memory which is for example used in the processing carried out in the device. A memory used in the embodiments may be a volatile memory, for example a DRAM (Dynamic Random Access Memory) or a non-volatile memory, for example a PROM (Programmable Read Only Memory), an EPROM (Erasable PROM), EEPROM (Electrically Erasable PROM), or a flash memory, e.g., a floating gate memory, a charge trapping memory, an MRAM (Magnetoresistive Random Access Memory) or a PCRAM (Phase Change Random Access Memory).

[0020] In order that the invention may be readily understood and put into practical effect, various embodiments will now be described by way of examples and not limitations, and with reference to the figures.

[0021] According to various embodiments, a computer-implemented method 100 for navigating a robot in an environment, is provided. The method 100 may include analysing three-dimensional (3D) point cloud data to determine passable versus impassable areas in the surroundings of the robot. The 3D point cloud data may be collected by a sensor onboard the robot. The analysis may enable the robot to navigate around its environment, and even access inclined surfaces such as a real-world handicapped ramp access. This may allow the robot to make use of urban passageways that were designed for humans. The method 100 may also be useful for navigating robots in carparks, bus stops and other urban settings where there may be multiple obstacles that the robot needs to avoid. In the context of this application, robots may include autonomous vehicles.

[0022] FIG. 1 is a simplified schematic diagram that illustrates the method 100 for navigating a robot in an environment, according to various embodiments. The method 100 may include performing perception 102. The process of perception 102 may include generating 3D data 110 about the environment around the robot, using a sensor. The sensor may be capable of perceiving the environment in 3D. Examples of the sensor may include a 3D LiDAR sensor, a 3D radar sensor, or a plurality of sensors that work together to generate 3D perception data. The method 100 may further include performing a simultaneous localizing and mapping (SLAM) process 104, using the 3D data 110. The sensor may continuously, or periodically, generate the 3D data 110. The 3D data 110 may then be provided to a SLAM module continuously or periodically, so that the SLAM module receives updated information about the environment. The SLAM process 104 may result in a point cloud map 112 and a pose of the robot. A traversability analysis 105 may be carried out based on the point cloud map 112, to mark out which regions in the environment are traversable, i.e. passable, and which regions are non-traversable, i.e. non-passable. The output of the traversability analysis 105 may be referred herein as a traversability map 114. The method 100 may further include generating a motion path based on the traversability map 114 and further based on the pose of the robot as determined through the SLAM process 104, in a path planning process 106. The pose of the robot is also referred herein as “robot pose” and refers to the location and orientation of the robot. The path planning process 106 may generate a motion path 116 for the robot. The motion path 116 may be transmitted to the robot hardware 108. The motion path 116 may include commands for the robot, for example, velocity commands, for the robot to travel from its current position to a desired destination.

[0023] FIG. 2 shows a block diagram 200 of an implementation of the method 100. The embodiment shown in FIG. 2 may be combined with any above-described embodiment or with any below described further embodiment. In this implementation, the robot is equipped with a 3D LiDAR 204 and a low-cost inertial measurement unit (IMU). The robot is placed in an environment that includes an access ramp. The environment is referred herein as a ramp scenario 202. The 3D LiDAR 204 may perceive the ramp scenario 202, and thereby generate 3D sensor data 110. The 3D sensor data 110 may include a 3D point cloud. The 3D sensor data 110 may be transmitted to a SLAM module 210.

[0024] The SLAM module 210 may segment the 3D point cloud into different categories by clustering to result in segmented point cloud data. The SLAM module 210 may then extract features of the current frame, from the 3D point cloud and the segmented point cloud data. The SLAM module 210 may then match the point cloud data by the position relationship between the front and back frames to generate a 3D point cloud map 214, using for example, the Levenberg-Marquardt method and the features of the current frame extracted from the 3D point cloud and the segmented point cloud data. In addition, based on the point cloud matched pose and real-time pose received from the IMU, the SLAM module 210 may estimate the robot pose 216. As an example, the SLAM module 210 may be implemented using the Lego-LOAM algorithm. Details of the Lego-LOAM algorithm are disclosed in “LeGO-LOAM: Lightweight and ground-optimized lidar odometry and mapping on variable terrain” by Shan et. al. in 2018 IEEE / RSJ Int. Conf. Intell. Robots Syst., which is incorporated herein by reference.

[0025] Next, the outputs of the SLAM module 210 may be transmitted to a traversability analysis module 220. The traversability analysis module 220 may generate an elevation map 224 based on the 3D point cloud map 214. To this end, the traversability analysis module 220 may generate a planar map comprising a plurality of grids. The planar map may initially be a blank grid map. The traversability analysis module 220 may then extract from the 3D point cloud map 214, height information, also referred herein as elevation information. The traversability analysis module 220 may add the relevant height information to the respective grids of the planar map, to generate an elevation map 224. The elevation map 224 is also referred to as a “2.5 dimensional (2.5D)” map, as it is a 2D map with appended height (third dimension) data. As an example, the traversability analysis module 220 may include a Baynesian generalized kernel (BGK) inference module 222. Details of the BGK inference algorithm are disclosed in “Bayesian generalized kernel inference for terrain traversability mapping,” by Shan et. al., in Proc. 2nd Conf. Robot Learn., which is incorporated herein by reference.

[0026] The BGK inference module 222 may use the BGK inference algorithm to predict elevation data for areas in the point cloud map 214 that has sparse sensor data, thereby densifying the elevation data of the 3D point cloud map 214. The BGK inference module 222 may update the elevation data of the elevation map 224, based on elevation data of the grid from a previous frame, and further based on elevation data estimated based on the 3D point cloud map of the current frame. Next, the BGK inference module 222 may classify regions of the elevation map 224 as being passable or impassable, based on elevation, slope and roughness. As a result, a binary costmap 226 is generated, where each region is classified as either passable or impassable. The binary costmap 226 is a 2D map, that may be used as an input to 2D navigation algorithms.

[0027] The outputs of the traversability analysis module 220 include the 2D costmap 226, also referred herein as a traversability map 226. The traversability map 226 may be provided to a path planner 230. The path planner 230 may include a global planner 232 and a local planner 234. The global planner 232 may receive positional information of a destination 208, and may receive the robot pose 216. The global planner 230 may generate a global path 236 based on these information. The global path 236 may indicate a passable path from the current position of the robot to the destination. The local planner 234 may process the global path 236 to result in a motion path 116. The robot may not be able to fully align with the global path 236, due to its hardware limitations. The local planner 234 addresses this misalignment, by adjusting the global path 236 for the robot according to its hardware capabilities, to result in an achievable trajectory. The local planner 234 may fine tune the global path 236 to ensure that the robot obtains a smooth trajectory while following the global path 236 as much as possible. The motion path 116 may be provided to the robot hardware 108. The robot hardware 108 may include components such as driving controllers, wheels, and engine transmission. The motion path 116 may include driving instructions, for example, velocity commands. The robot may drive to the destination based on the motion path 116.

[0028] As an example, the global planner 232 may include a dual-tree rapidly exploring random tree (RRT) planner. Details of the dual-tree RRT planner are disclosed in “A fast and efficient double-tree RRT*-like sampling-based planner applying on mobile robotic systems,” by Chen et. al., in IEEE-ASME Trans. Mechatron., which is incorporated herein by reference. As an example, the local planner 234 may include a dynamic window approach (DWA) planner. Details of the DWA planner are disclosed in “Vision based sidewalk navigation for last-mile delivery robot” by Wen et. al., in 2022 17th Int. Conf. Control Automat. Robot. Vision (ICARCV), which is incorporated herein by reference. The dual-tree RRT planner may obtain an optimal passable path faster than the DWA planner, while the DWA planner may generate a smooth trajectory.

[0029] The method 100 using the implementation shown in FIG. 2, was validated in the real-world environment, using a last-mile delivery robot. A four-wheel differential-driven last-mile delivery robot manufactured by Continental was used for a ramp access navigation task. In the following, the experiment is described.

[0030] FIGS. 3 A to 3C show the maps used in the experiment to validate the method 100. First, the effectiveness of the mapping and traversability algorithms were verified using pre-collected ramp access point cloud data. FIG. 3A shows the 3D sensor data 110 collected by the 3D LiDAR 204. The 3D sensor data 110 is a 3D point cloud. The 3D point cloud is provided to the SLAM module 210 that generates a 3D point cloud map 214 as shown in FIG. 3B. The 3D point cloud map 214 is more accurate that the 3D sensor data 110. The 3D point cloud map 214 is imported into the traversability analysis module 220 to generate a 2.5D elevation map 224, as shown in FIG. 3B. The traversability analysis module 220 may determine traversability based on a set of traversability criterion and its corresponding threshold.

[0031] The traversability criteria may include height ah, slope as, and roughness ar. The criterion may be expressed as follows:

[0032] Tc = — + k2 - + , L ah as ar

[0033] where ah, as, and ar are the maximum allowable threshold of ah, as, and ar respectively. k±, k2, and k3 are the weights satisfying k± + k2 + k3 = 1. The value of Tc may indicate whether the terrain is flat, rough, or impassable. Furthermore, if one of the criteria is larger than its maximum allowable threshold, the corresponding grid may be set as impassable. The relevant thresholds may be set manually. Based on the above traversability analysis and the classification of the grid elevation, the grid map of the robot may be binarized to distinguish the passable / impassable areas, to result in the 2D binary costmap 226 as shown in FIG. 3C.

[0034] The last-mile delivery robot was tested on an actual handicapped ramp access and the last-mile delivery robot passed through the handicapped ramp access smoothly without collision. The 3D point cloud map 214, the 2.5D elevation map 224, and the 2D costmap 226 were generated in real-time from the sensor data 110 collected by the robot via an onboard 3D LiDAR. The method 100 was able to result in a motion path that enabled the robot to navigate the ramp access successfully.

[0035] FIG. 4 shows a flow diagram of the method 100 for navigating a robot in an environment, according to various embodiments. The method 100 may include processes 402, 404, 406, 408 and 410. The process 402 may include constructing a point cloud map representing the environment based on 3D sensor data. The process 404 may include determining pose of the robot based on the 3D sensor data. The process 406 may include generating an elevation map based on the point cloud map. The process 408 may include generating a traversability map based on the elevation map. The process 410 may include planning a motion path for the robot based on the traversability map.

[0036] The traversability map is a 2D map. By using a 2D map to determine the motion path, the method 100 is computationally less demanding than 3D navigation methods. At the same time, the method 100 includes adding elevation data to the elevation map, and the traversability map is derived using the elevation map, such that the method 100 takes into consideration elevation data in its planning of the motion path. As such, the method 100 is more effective than existing 2D navigation methods, as it can differentiate an accessible path with elevation such as a ramp, from an obstacle.

[0037] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, planning the motion path, i.e. the process 410, may include planning the motion path for the robot to access a ramp in the environment. This enables the robot to make use of available ramps to move between positions of different heights, instead of treating a ramp as an obstacle. This may allow the robot to enter buildings, overhead bridges and tunnels, thereby enhancing the usability of the robot in an urban setting.

[0038] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, at least one of constructing the point cloud and determining pose of the robot includes extracting features from the 3D sensor data using SLAM technique. This enables the robot to map out its surroundings in real time and also to determine its position in the environment. The robot map therefore navigate in a new environment for which there may be no predefined map or terrain information.

[0039] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, the pose of the robot is determined based on the point cloud map and odometry data. The combination of the odometry data and the point cloud map make improve accuracy of localizing the robot based on either the point cloud map alone or the odometry data alone. Using the point cloud map alone for localization could include errors, for example introduced due to temporary obstruction of view, whereas localization using odometry data alone is inaccurate as errors accumulate overtime.

[0040] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, generating the elevation map includes discretizing a planar map into a plurality of grid cells, determining elevation data based on the point cloud map, and assigning respective elevation data to each grid cell of the plurality of grid cells. The resulting elevation map is 2D while still containing elevation information. This elevation map can be analysed by 2D navigation algorithms which use less computational power as compared to 3D navigation algorithms.

[0041] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, determining elevation data comprises applying Baynesian Generalized Kernel inference on the point cloud data to estimate elevation of positions in the point cloud map with sparse sensor data. The BGK inference algorithm may densify the point cloud map with elevation data, and thereby provide an accurately identify whether the surroundings are passable or impassable regions.

[0042] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, the traversability map is a two-dimensional binary cost map. The 2D cost map can be analysed by 2D navigation algorithms which use less computational power as compared to 3D navigation algorithms.

[0043] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, the traversability map is generated further based on traversability criterion and passable threshold.

[0044] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, the 3D sensor data is generated by a sensor mounted on the robot, and wherein the sensor is a LiDAR sensor. LiDAR sensor is able to output accurate geometric locations in 3D space, and the geometric information is useful for determining height, slope, continuity of the elevation, which are inputs to the traversability module.

[0045] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, planning the motion path comprises generating velocity commands for the robot based on the pose of the robot, a destination and the traversability map. The robot may then travel to the destination based on these velocity commands, while avoiding impassable regions in its surroundings.

[0046] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, planning the motion path comprises generating a global path based on the pose of the robot and a destination. The global path may be generated at faster speed, to determine a shortest path between the robot and destination.

[0047] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, planning the motion path further comprises generating a local path based on the global path and the traversability map. The local path may take into consideration the traversability of the path to ensure that the path towards the destination is safe.

[0048] FIG. 5 shows a block diagram of a navigation system 500 for navigating a robot in an environment according to various embodiments. The navigation system 500 may include a mapping module, a pose module, an elevation module, a traversability module and a path planner. The mapping module may be configured to construct a point cloud map based on 3D sensor data. The pose module may be configured to determine pose of the robot based on the 3D sensor data. The elevation module may be configured to generate an elevation map based on the point cloud map. The traversability module may be configured to generate a traversability map based on the elevation map. The path planner may be configured to plan a motion path for the robot based on the traversability map.

[0049] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, the navigation system 500 may include a SLAM module. The SLAM module may include the mapping module and the pose module.

[0050] FIG. 6 shows a block diagram of a robot 600 according to various embodiments. The robot 600 may include the navigation system 500.

[0051] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, the robot 600 may further include a sensor configured to generate the 3D sensor data

[0052] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, the robot may further include an IMU. The IMU may be configured to generate the odometiy data, that may be used to determine the pose of the robot.

[0053] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, the robot 600 may be a last mile delivery robot. The last mile delivery robot may deliver goods from outdoors into an indoor environment, to the customer's doorstep.

[0054] According to an embodiment which may be combined with any above-described embodiment or with any below described further embodiment, the robot 600 may be an autonomous vehicle. The navigation system 500 in the autonomous vehicle may guide the vehicle to navigate in complex environments, such as to move up carpark ramps.

[0055] While embodiments of the invention have been particularly shown and described with reference to specific embodiments, it should be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. The scope of the invention is thus indicated by the appended claims and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced. It will be appreciated that common numerals, used in the relevant drawings, refer to components that serve a similar or the same purpose.

[0056] It will be appreciated to a person skilled in the art that the terminology used herein is for the purpose of describing various embodiments only and is not intended to be limiting of the present invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0057] It is understood that the specific order or hierarchy of blocks in the processes / flowcharts disclosed is an illustration of exemplary approaches. Based upon design preferences, it is understood that the specific order or hierarchy of blocks in the processes / flowcharts may be rearranged. Further, some blocks may be combined or omitted. The accompanying method claims present elements of the various blocks in a sample order, and are not meant to be limited to the specific order or hierarchy presented.

[0058] The previous description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects. Thus, the claims are not intended to be limited to the aspects shown herein, but is to be accorded the full scope consistent with the language claims, wherein reference to an element in the singular is not intended to mean “one and only one” unless specifically so stated, but rather “one or more.” The word “exemplary” is used herein to mean “serving as an example, instance, or illustration.” Any aspect described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects. Unless specifically stated otherwise, the term “some” refers to one or more. Combinations such as “at least one of A, B, or C,” “one or more of A, B, or C,” “at least one of A, B, and C,” “one or more of A, B, and C,” and “A, B, C, or any combination thereof’ include any combination of A, B, and / or C, and may include multiples of A, multiples of B, or multiples of C. Specifically, combinations such as “at least one of A, B, or C,” “one or more of A, B, or C,” “at least one of A, B, and C,” “one or more of A, B, and C,” and “A, B, C, or any combination thereof’ may be A only, B only, C only, A and B, A and C, B and C, or A and B and C, where any such combinations may contain one or more member or members of A, B, or C. All structural and functional equivalents to the elements of the various aspects described throughout this disclosure that are known or later come to be known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be encompassed by the claims.

Claims

1. A computer-implemented method (100) for navigating a robot in an environment, the method (100) comprising:constructing a point cloud map representing the environment, based on 3D sensor data (402);determining pose of the robot based on the 3D sensor data (404);generating an elevation map based on the point cloud map (406);generating a traversability map based on the elevation map (408); and planning a motion path for the robot based on the traversability map (410).

2. The method (100) of any preceding claim, wherein planning the motion path comprises planning the motion path for the robot to access a ramp in the environment.

3. The method (100) of any preceding claim, wherein at least one of constructing the point cloud and determining pose of the robot comprises extracting features from the 3D sensor data using SLAM technique.

4. The method (100) of any preceding claim, wherein the pose of the robot is determined based on the point cloud map and odometry data.

5. The method (100) of any preceding claim, wherein generating the elevation map comprises discretizing a planar map into a plurality of grid cells, determining elevation data based on the point cloud map, and assigning respective elevation data to each grid cell of the plurality of grid cells.

6. The method (100) of claim [0040], wherein determining elevation data comprises applying Baynesian Generalized Kernel inference on the point cloud data to estimate elevation of positions in the point cloud map with sparse sensor data.

7. The method (100) of any preceding claim, wherein the traversability map is a two-dimensional binary cost map.

8. The method (100) of any preceding claim, wherein the traversability map is generated further based on traversability criterion and passable threshold.

9. The method (100) of any preceding claim, wherein the 3D sensor data is generated by a sensor mounted on the robot, and wherein the sensor is a LiDAR sensor.

10. The method (100) of any preceding claim, wherein planning the motion path comprises generating velocity commands for the robot based on the pose of the robot, a destination and the traversability map.

11. The method (100) of any preceding claim, wherein planning the motion path comprises generating a global path based on the pose of the robot and a destination.

12. The method (100) of claim [0045], wherein planning the motion path further comprises generating a local path based on the global path and the traversability map.

13. A computer program product comprising instructions, which, when the program is executed by a computer, cause the computer to carry out the method (100) of any preceding claim.

14. A data carrier signal carrying the computer program product of claim 13.

15. A navigation system (500) for navigating a robot in an environment, the navigationsystem (500) comprising:a mapping module (502) configured to construct a point cloud map based on 3D sensor data;a pose module (504) configured to determine pose of the robot based on the 3D sensordata;an elevation module (506) configured to generate an elevation map based on the point cloud map;a traversability module (508) configured to generate a traversability map based on the elevation map; anda path planner (230) configured to plan a motion path for the robot based on the traversability map.

16. The navigation system (500) of claim 15, wherein the navigation system (500) comprises a SLAM module comprising the mapping module (502) and the pose module (504).

17. A robot (600) comprising:the navigation system (500) of any one of claims 15 to 16.

18. The robot (600) of claim 17, further comprising: a sensor (602) configured to generate the 3D sensor data.

19. The robot (600) of any one of claims 17 to 18, further comprising: an inertial measurement unit (604) configured to generate odometry data.

20. The robot (600) of any one of claims 17 to 19, wherein the robot (600) is a last mile delivery robot.

Images