A seamless connection method and system of indoor and outdoor maps of a robot

Abstract

This invention discloses a method and system for seamlessly connecting indoor and outdoor maps for a robot. The method includes the following steps: establishing a first positioning reference and a second positioning reference based on physical entities in the indoor and outdoor areas respectively, and constructing first feature information for identifying the first positioning reference and second feature information for identifying the second positioning reference; controlling the robot to travel along a preset transition path, and synchronously collecting first positioning information based on the first positioning reference and second positioning information based on the second positioning reference at multiple locations on the preset transition path to generate and store the correspondence between the first positioning information and the second positioning information; when the robot travels along the preset transition path and enters a preset switching interval, based on the stored correspondence, smoothly transitioning the robot's positioning output from the first positioning information based on the first positioning reference to the second positioning information based on the second positioning reference, or performing the reverse smooth transition, to improve stability.

Description

Technical Field

[0001] This invention relates to the field of robot application technology, specifically to a method and system for seamlessly connecting indoor and outdoor maps for robots. Background Technology

[0002] With the rapid development of robotics technology, the application scenarios of robots have expanded from single indoor environments to mixed indoor and outdoor scenarios, such as park inspections and indoor and outdoor logistics delivery. In these application scenarios, robots need to move continuously between indoor and outdoor environments. Therefore, this alternating indoor and outdoor work scenario places higher demands on the robot's positioning and navigation capabilities.

[0003] Currently, mainstream robot localization technologies are typically optimized for a single environment. In indoor environments, due to weak or unavailable Global Navigation Satellite System (GNSS) signals, Simultaneous Localization and Mapping (SLAM) technologies based on LiDAR or vision are commonly used to achieve localization by constructing a local map and matching it with environmental features. In outdoor environments, satellite-based localization technologies, such as Real-time Kinematic (RTK), are primarily relied upon to obtain high-precision global coordinates.

[0004] To achieve continuous localization for robots navigating indoors and outdoors, a common approach in existing technologies is to unify indoor and outdoor maps into a single global coordinate system. This method typically involves first establishing one or more anchor points with absolute coordinates outdoors using RTK technology. Then, through measurement or coordinate transformation, the local map constructed using indoor SLAM is forcibly registered to this global coordinate system, thereby creating a large, coordinate-unified map extending from indoors to outdoors.

[0005] However, the aforementioned method for establishing a unified coordinate system has significant drawbacks in practical applications. First, the differences between indoor and outdoor environments are substantial, and the error characteristics and drift trends of indoor SLAM positioning (relative coordinates) and outdoor RTK positioning (absolute coordinates) are completely different. Forcing an indoor map to align to the global coordinate system using a limited number of anchor points will cause significant nonlinear deformation and cumulative errors in areas far from the anchor points, making it difficult to guarantee positioning accuracy. Second, the environment is not static. Adjustments to the indoor layout, the growth of outdoor vegetation, or building repairs can all cause the original unified coordinate relationship to become invalid. The system requires frequent and complex manual calibration for maintenance, resulting in high maintenance costs and a lack of long-term stability. More importantly, when the robot moves in the boundary area between indoor and outdoor environments (such as doorways), due to the switching of positioning sources and the existence of coordinate system deviations, its positioning output is prone to jumps, leading to discontinuous robot pose perception, which in turn causes control oscillations or navigation failures, seriously affecting the robot's operational efficiency and safety. Summary of the Invention

[0006] To overcome the shortcomings of existing technologies, this invention provides a method and system for seamlessly connecting indoor and outdoor maps for robots, thereby solving the problems in existing technologies.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for seamlessly integrating indoor and outdoor maps for robots includes the following steps: A first positioning reference and a second positioning reference based on physical entities are established in the indoor area and the outdoor area respectively, and a first feature information for identifying the first positioning reference and a second feature information for identifying the second positioning reference are constructed; wherein, the first positioning reference and the second positioning reference are independent positioning references. Before the robot performs a task, the robot is controlled to travel along a preset transition path, and first positioning information based on the first positioning reference and second positioning information based on the second positioning reference are collected at multiple locations on the preset transition path simultaneously, so as to generate and store the correspondence between the first positioning information and the second positioning information. When the robot travels along the preset transition path and enters the preset switching interval, based on the stored correspondence, the robot's positioning output is smoothly transitioned from the first positioning information based on the first positioning reference to the second positioning information based on the second positioning reference, or the reverse smooth transition is performed.

[0008] In one embodiment, a first positioning reference and a second positioning reference based on physical entities are established in an indoor area and an outdoor area, respectively, and first feature information for identifying the first positioning reference and second feature information for identifying the second positioning reference are constructed, including: The first positioning reference is established based on indoor fixed facilities, including charging piles. The charging piles have preset charging positions. The robot is controlled to stop at the preset charging positions, and the current pose information of the robot is set as the first positioning reference. The charging piles are then scanned using a lidar to generate the first feature information. The second positioning reference is established based on outdoor landmark environmental features, which are fixed landmark street scene structures. The global coordinates of the fixed landmark street scene structure are obtained through real-time dynamic differential positioning technology as the second positioning reference, and the street scene structure is scanned by LiDAR to generate the second feature information.

[0009] In one embodiment, the method further includes establishing a backup positioning reference, wherein establishing the backup positioning reference includes: Select at least one backup indoor fixed facility or backup outdoor landmark environmental feature, establish a backup positioning benchmark and construct the corresponding feature information in the same way.

[0010] In one embodiment, when controlling the robot to travel along the preset transition path, the robot pose information corresponding to each position point is recorded synchronously, and the pose information is used as a component of the first positioning information or the second positioning information.

[0011] In one embodiment, the robot is controlled to travel along a preset transition path, and simultaneously collects first positioning information based on the first positioning reference and second positioning information based on the second positioning reference at multiple locations along the preset transition path, including: The robot is controlled to travel back and forth along the preset transition path multiple times, and the positioning information collected multiple times is fused to generate the corresponding relationship.

[0012] In one embodiment, the correspondence is a coordinate mapping table, which records the one-to-one correspondence between the first coordinates and the second coordinates of discrete position points sampled at equal intervals of spatial distance on the preset transition path; The coordinate mapping table is generated in the following way: The location information collected multiple times is aligned in time and space. After removing outliers, the average of the first and second coordinates of the corresponding location points from each collection is taken as the coordinate value of the location point in the coordinate mapping table.

[0013] In one embodiment, the start and end points of the switching interval are determined based on the physical boundary features on the preset transition path, including door frames, thresholds, ground markings, or building edges.

[0014] In one embodiment, the switching interval is a continuous path segment from a first physical boundary feature to a second physical boundary feature, wherein the first physical boundary feature is located on the indoor area side and the second physical boundary feature is located on the outdoor area side.

[0015] In one embodiment, the smooth transition employs an interpolation algorithm to gradually adjust the weights of the first and second positioning information in the positioning output, so that the positioning output value changes continuously.

[0016] In one embodiment, within the switching interval, a normalized position coefficient is calculated based on the robot's current position on the preset transition path, and the weights of the first positioning information and the second positioning information are dynamically determined based on the normalized position coefficient, so that the weights change continuously as the robot moves.

[0017] A seamless indoor and outdoor map integration system for robots includes: An anchor point calibration module is used to establish a first positioning reference and a second positioning reference based on physical entities in indoor and outdoor areas, respectively, and to construct first feature information for identifying the first positioning reference and second feature information for identifying the second positioning reference; wherein, the first positioning reference and the second positioning reference are independent positioning references. The path mapping module is used to control the robot to travel along a preset transition path before the robot performs a task, and to simultaneously collect first positioning information based on the first positioning reference and second positioning information based on the second positioning reference at multiple locations on the preset transition path, so as to generate and store the correspondence between the first positioning information and the second positioning information. The smooth transition module is used to, when the robot travels along the preset transition path and enters the preset transition interval, smoothly transition the robot's positioning output from the first positioning information based on the first positioning reference to the second positioning information based on the second positioning reference, or perform the reverse smooth transition, based on the stored correspondence.

[0018] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention establishes independent first and second positioning references based on physical entities, both indoors and outdoors, enabling independent construction and maintenance of indoor and outdoor maps. This avoids the coordinate drift and calibration complexities caused by forced unification. Furthermore, the positioning references based on physical entities are repeatable and stable over the long term, thus providing a reliable reference system for subsequent steps.

[0019] The path mapping step of this invention involves controlling a robot to actually travel along a preset transition path, simultaneously collecting positioning information of each location point under two independent references, and generating and storing the correspondence between them. This correspondence is rigid; it directly originates from the robot's physical movement and records the true corresponding values ​​of the same physical location in the two coordinate systems, rather than an approximation result of mathematical fitting. By pre-generating and storing the information offline, the switching basis is not affected by the real-time environment, fundamentally ensuring the accuracy and reliability of the switching. Furthermore, through repeated data collection and data fusion processing during reciprocating travel, random errors are further eliminated, ensuring high precision in the mapping relationship.

[0020] The smooth switching step of this invention, within a preset switching interval, gradually adjusts the output weights of the two positioning information types based on a fixed-stored correspondence using an interpolation algorithm. This ensures continuous variation in the positioning output value. During the switching process, the positioning output is no longer a simple fusion of real-time sensor data, but rather based on a pre-established rigid mapping. This achieves a seamless transition from positioning based on a first reference to positioning based on a second reference, fundamentally avoiding jumps caused by unstable sensor signals or coordinate system deviations at the switching point, thus ensuring the continuity and stability of the robot's pose perception. Furthermore, since the switching interval is determined based on physical boundary features, the accuracy of the switching timing and environmental adaptability are guaranteed. Attached Figure Description

[0021] Figure 1 This invention provides a flowchart of a method for seamlessly connecting indoor and outdoor maps for robots. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, the present invention provides a method for seamlessly connecting indoor and outdoor maps for robots, comprising the following steps: S100. Establish a first positioning reference and a second positioning reference based on physical entities in the indoor area and the outdoor area respectively, and construct a first feature information for identifying the first positioning reference and a second feature information for identifying the second positioning reference; wherein, the first positioning reference and the second positioning reference are independent positioning references. In this embodiment, by establishing independent first and second positioning references based on physical entities, the problem of long-term stable alignment between indoor and outdoor coordinate systems in the prior art is solved. The physical constraints of the indoor charging pile ensure the repeatability of the origin, while the RTK calibration of the outdoor street view structure ensures the accuracy of the global coordinates. The independent design of the two allows indoor and outdoor maps to be built and maintained separately, avoiding coordinate drift and calibration complexity problems caused by forced unification.

[0024] In one embodiment, a first positioning reference and a second positioning reference based on physical entities are established in an indoor area and an outdoor area, respectively, and first feature information for identifying the first positioning reference and second feature information for identifying the second positioning reference are constructed, including: S110. The first positioning reference is established based on an indoor fixed facility, which includes a charging pile. The charging pile has a preset charging position. The robot is controlled to stop at the preset charging position, and the current pose information of the robot is set as the first positioning reference. The charging pile is scanned by a laser radar to generate the first feature information. It should be noted that indoor fixed facilities are determined based on their fixed location, inability to move, and identifiable three-dimensional features. In this embodiment, facilities that the robot interacts with daily (such as charging stations) are prioritized to ensure their repeatable docking capability. By using charging facilities that are essential for the robot's daily operation, the problem of needing to deploy additional markers can be effectively avoided, thus achieving multiple uses for a single item. For example: The steps for establishing the first positioning reference (i.e., indoor anchor point calibration) include: controlling the robot to accurately stop at the standard charging position of the indoor charging pile through physical limits; starting laser SLAM and setting the origin of the robot's laser radar coordinate system at this time as the origin (0, 0, 0) of the indoor local coordinate system, i.e., the first positioning reference; controlling the robot to rotate in place and using the laser radar to scan the charging pile from multiple angles, collecting ≥50 sets of point cloud data; filtering, registering, and extracting features from the point cloud data to generate a charging pile feature template (such as the FPFH descriptor), i.e., the first feature information, which is stored in the anchor point feature library.

[0025] In this embodiment, controlling the robot to dock at a preset charging position ensures the repeatability of the robot's pose each time it docks, thereby providing stability for the first positioning reference and solving the problem that pure laser matching may fail due to feature changes. First feature information is generated through lidar scanning, providing a reliable matching basis for subsequent online anchor point identification, ensuring stable identification even under conditions of changing lighting or partial occlusion.

[0026] If there are no charging stations, permanent indoor pillars, shelf corners, etc. can be selected, but it is necessary to ensure that the robot can be stably identified by LiDAR.

[0027] As needed, the charging station has two positioning pins installed at the preset charging positions, and the robot's bottom has positioning holes corresponding to the positions of these pins. This ensures that the positioning pins can be stably inserted into the positioning holes each time the robot stops, thus fixing the robot's posture. Specifically, for example: Setting the first positioning reference (e.g., the origin of the first coordinate system) includes: the robot automatically navigating to the charging station until the positioning pin fully engages with the positioning hole and stops in place. The onboard controller sends a command to "set the origin of the indoor first coordinate system (i.e., the first positioning reference)". The system records the position of the robot's LiDAR coordinate system origin in the SLAM map at this time and forcibly sets it to (0, 0, 0) of the indoor local coordinate system. All indoor map data is updated to align its coordinate origin to this point.

[0028] The first feature information generation steps include: After setting the origin of the first coordinate system, the robot rotates 360° in place and scans the charging pile from multiple angles using a 32-line LiDAR. The real-time collected point cloud data undergoes preprocessing such as voxel filtering downsampling (voxel size 0.02m) and statistical filtering to remove outliers. The ICP algorithm is used to register and fuse multiple frames of point clouds to generate a complete 3D point cloud model of the charging pile. Key feature points are extracted from the point cloud model, and FPFH descriptors are calculated to form a charging pile feature template (i.e., the first feature information). The feature template (i.e., the first feature information) is stored in the anchor point feature library.

[0029] S120. The second positioning reference is established based on outdoor landmark environmental features, which are fixed landmark street scene structures. The global coordinates of the fixed landmark street scene structure are obtained through real-time dynamic differential positioning technology as the second positioning reference, and the street scene structure is scanned by LiDAR to generate the second feature information.

[0030] It should be noted that the fixed landmark street scene structure is determined based on uniqueness, identifiability, fixity, and accessibility. Uniqueness means that within the robot's working range, the feature should be unique, with no similar structures to avoid mismatches. Identifiability means it has a rich 3D structure, with a clear point cloud outline obtainable by LiDAR, and ≥20 feature points. Fixity means it is a permanent fixed facility, with no possibility of movement, modification, or demolition. Accessibility means the RTK mobile station can be safely set up at the feature center with a clear view of the sky. For example: The steps for establishing the second positioning reference (i.e., outdoor anchor point calibration) include: selecting a unique, fixed, and laser-identifiable street scene structure (such as a triple street light group) within the area as the outdoor anchor point. An RTK base station is set up, and the mobile station is placed at the center of the anchor point. At least 100 sets of WGS84 coordinates are continuously collected in fixed solution mode, and the average value is taken as the global coordinates of the anchor point, with an accuracy of ±2cm. This point is set as the origin (0, 0, 0) of the outdoor local coordinate system, i.e., the second positioning reference. The street scene structure is scanned using LiDAR to generate a point cloud feature template, which is then registered with the satellite map to ensure alignment, i.e., the second feature information.

[0031] In this embodiment, by utilizing the fixed streetscape features with unique three-dimensional structures in the environment, the vulnerability and maintenance difficulties of artificial landmarks are avoided. Global coordinates obtained through RTK provide absolute coordinates aligned with satellite maps for the outdoor reference, enabling the robot to locate itself in the global coordinate system. Feature templates generated by LiDAR scanning (i.e., second feature information) allow the robot to autonomously identify anchor points through perception without human intervention. Deep fusion of RTK and LiDAR overcomes the drawbacks of pure RTK signals being easily lost and pure LiDAR lacking absolute coordinates.

[0032] For example, the only "triple street light group" in the area was selected as the main anchor point. This street light group consists of three lamp posts arranged in an isosceles triangle, with each lamp post 6 meters high and a horizontal lamp arm at the top. LiDAR scanning can obtain rich point cloud features. RTK global coordinate acquisition includes: setting up an RTK base station in an open area near the street light cluster, ensuring the base station coordinates are known. Positioning the RTK rover at the center of the street light cluster (the geometric center of the three lamp posts), securing it with a tripod, and ensuring the rover antenna is horizontal. Powering on the rover and waiting for RTK to enter fixed solution mode, confirming the number of satellites is ≥20, C / N0 > 45dB, and PDOP < 2. Continuously acquiring 120 sets of WGS84 coordinate data at a sampling frequency of 1Hz, monitoring data stability in real time, and removing jump values. Statistical analysis of the acquired data, taking the average value as the global coordinates of the street light cluster.

[0033] The second positioning reference setting (such as the origin of the second coordinate system) includes: the robot moves to the vicinity of the street light group to ensure that the LiDAR can clearly scan the street light group. The geometric center point of the three lamp posts is manually marked in the LiDAR point cloud, and this point is set as the origin (0, 0, 0) of the outdoor local coordinate system (second coordinate system). The robot is controlled to move around the street light group for one revolution, and the LiDAR is used to scan from multiple angles to collect 100 frames of point cloud data. The point cloud data is preprocessed, registered and fused, and features are extracted to generate a street light group feature template (i.e., the second feature information).

[0034] Registration with satellite map: Load the campus satellite map and project the global coordinates of the street light group onto the satellite map. Overlay the local point cloud map built by the LiDAR onto the satellite map, and manually fine-tune to ensure alignment.

[0035] This invention provides dual protection through physical limiting (i.e., positioning pins and positioning holes) and laser features, ensuring that the first indoor positioning reference has both physical repeatability and perceptual identifiability; and through dual protection through RTK and laser, it provides the second outdoor positioning reference with both global absolute coordinates and local point cloud features, thereby significantly improving the reliability and robustness of the positioning reference of this invention.

[0036] Specifically, this invention solves the problems of high deployment costs and difficult maintenance caused by relying on manually placed markers in existing technologies by utilizing existing physical entities in the environment as positioning references. Indoor fixed facilities (such as charging piles) are essential facilities for robots on a daily basis, enabling functional reuse; outdoor landmark environmental features (such as street scene structures) are permanent or long-term features in the environment, requiring no additional deployment or maintenance. Since indoor fixed facilities are usually fixed in location, and outdoor landmark street scene structures are not easily moved or modified, the long-term reliability of the first and second positioning references is guaranteed. This effectively solves the problems of easy drift of calibration parameters and the need for frequent calibration in existing technologies. Furthermore, different types of references are used indoors and outdoors to adapt to the characteristics of their respective environments. Indoor environments have limited space and abundant features, making them suitable for using nearby fixed facilities; outdoor environments are open and sparse in features, making them suitable for using distant landmark structures.

[0037] In one embodiment, the method further includes establishing a backup positioning reference, wherein establishing the backup positioning reference includes: Select at least one backup indoor fixed facility or backup outdoor landmark environmental feature, establish a backup positioning benchmark and construct the corresponding feature information in the same way.

[0038] In this embodiment, by establishing a backup positioning reference, the problem of positioning interruption caused by single-point failure in the prior art is solved. When the primary anchor point fails due to obstruction, damage, environmental changes, etc., the system can automatically switch to the backup anchor point, maintaining the continuity of positioning output and significantly improving the robustness and availability of the system. Establishing the backup reference in the same way ensures that the accuracy and characteristics of the backup reference are consistent with the primary reference, making the switching process smooth and reliable, avoiding positioning jumps caused by inconsistency in the reference. The redundancy design enables the system to cope with unpredictable environmental changes in actual working conditions (such as construction obstruction, facility modification), reducing the frequency of manual intervention and improving operation and maintenance efficiency. Specifically: Establishing a backup indoor reference point: If there are multiple charging stations indoors, the remaining charging stations can be designated as backup reference points. If there are no multiple charging stations, fixed shelf columns, equipment cabinets, etc., can be selected, but it must be ensured that the robot can stably identify them using LiDAR. The calibration method is the same as the primary reference point, including origin setting, feature acquisition, and template generation.

[0039] Establishment of a backup outdoor benchmark: Following the method described above, another permanent streetscape structure is selected for calibration. In this embodiment, a flower bed sculpture is selected as a backup anchor point. This sculpture is a concrete structure with a unique shape, can be clearly scanned by LiDAR, and its position is permanently fixed. The calibration steps are the same as for the main anchor point, including RTK coordinate acquisition, origin setting, feature acquisition, template generation, and satellite map registration.

[0040] The system records the characteristic information, priority, and switching conditions of the primary and backup anchor points. During operation, the matching score of each anchor point is monitored in real time. When the primary anchor point's matching score is below a threshold (e.g., 0.6) for 10 consecutive frames, it is determined to be a failed primary anchor point, and the system automatically switches to the backup anchor point with the highest current score.

[0041] The handover process uses weighted interpolation for a smooth transition, avoiding abrupt changes. Events are recorded, and alarm information is sent to the operations and maintenance center.

[0042] S200. Before the robot performs the task, control the robot to travel along a preset transition path, and simultaneously collect first positioning information based on the first positioning reference and second positioning information based on the second positioning reference at multiple locations on the preset transition path, so as to generate and store the correspondence between the first positioning information and the second positioning information. In this embodiment, by collecting and storing the corresponding relationships offline before the robot performs its task, the problems of online conversion or fusion being susceptible to environmental interference and exhibiting poor stability in existing technologies are solved. Actual physical travel ensures the authenticity of the corresponding relationships, and multiple data collections and fusion processes during reciprocating travel eliminate random errors, ensuring high accuracy of the mapping relationships. Furthermore, the method of storing the data offline ensures that the switching criteria are unaffected by the real-time environment, fundamentally guaranteeing the reliability of the switching.

[0043] In one embodiment, the robot is controlled to travel along a preset transition path, and simultaneously collects first positioning information based on the first positioning reference and second positioning information based on the second positioning reference at multiple locations along the preset transition path, including: The robot is controlled to travel back and forth along the preset transition path multiple times, and the positioning information collected multiple times is fused to generate the corresponding relationship.

[0044] In this embodiment, by collecting positioning information multiple times, the random errors of a single trip (such as brief vibrations caused by uneven ground or instantaneous sensor noise) are eliminated, thereby improving the accuracy and reliability of the mapping table.

[0045] In one embodiment, the correspondence is a coordinate mapping table, which records the one-to-one correspondence between the first coordinates and the second coordinates of discrete position points sampled at equal intervals of spatial distance on the preset transition path; The coordinate mapping table is generated in the following way: The location information collected multiple times is aligned in time and space. After removing outliers, the average of the first and second coordinates of the corresponding location points from each collection is taken as the coordinate value of the location point in the coordinate mapping table.

[0046] In this embodiment, discretization facilitates rapid lookup, and equal-interval sampling ensures sufficient spatial resolution. A one-to-one correspondence ensures the determinism of the mapping, avoiding fuzzy matching. Through a refined data processing flow, the accuracy and reliability of the mapping table are further guaranteed, eliminating various sources of error.

[0047] In one embodiment, when controlling the robot to travel along the preset transition path, the robot pose information corresponding to each position point is recorded synchronously, and the pose information is used as a component of the first positioning information or the second positioning information.

[0048] In this embodiment, by using robot pose information as first positioning information and second positioning information, a data basis is provided for the continuity of posture during smooth switching, avoiding sudden changes in robot direction during switching and ensuring the integrity of pose perception.

[0049] Specifically, the S200 steps include: The pre-planned transition path includes: selecting a fixed passage connecting the indoor charging station and the outdoor street scene sign (such as indoor corridor → gate → outdoor sidewalk), ensuring that the path has no branches, the environment is stable, and the length can be 5-20 meters.

[0050] The data acquisition steps include: Before the robot performs its daily tasks (e.g., at night or during idle periods), the robot is controlled to travel at a low speed (e.g., 0.1-0.2 m / s) along a preset transition path from indoors to outdoors. During the journey, the indoor coordinates (calculated by laser SLAM based on the charging pile origin) and outdoor coordinates (calculated by RTK+laser joint calculation, or by laser SLAM based on the street view marker origin) are recorded simultaneously at each sampling time, with a sampling frequency of 10 Hz. IMU / odometer data is also recorded for subsequent trajectory interpolation and error correction. The robot travels back and forth ≥ 5 times to obtain multiple sets of dense coordinate sequences.

[0051] The mapping generation process includes: time and spatial alignment of multiple sets of data. Preset transition paths are sampled at 0.1-meter intervals. For each sampling point, the average of multiple indoor and outdoor coordinates is calculated, and outliers are removed. A coordinate mapping table is generated in the format: path location (meters) → (indoor coordinates x, y, z, outdoor coordinates x, y, z). The mapping table is then stored locally on the robot in YAML or binary file format.

[0052] S300. When the robot travels along the preset transition path and enters the preset switching interval, based on the stored correspondence, the robot's positioning output is smoothly transitioned from the first positioning information based on the first positioning reference to the second positioning information based on the second positioning reference, or the reverse smooth transition is performed.

[0053] In this embodiment, by performing weighted interpolation based on the fixed-stored correspondence within a preset switching interval to smoothly transition, the problem of coordinate jumps at switching points in the prior art can be effectively solved. Determining the switching interval based on physical boundary features ensures the accuracy of the switching timing; by using a smooth transition to make the positioning output change continuously, a seamless transition is achieved, and the continuity and stability of the robot's pose perception are guaranteed.

[0054] In one embodiment, the start and end points of the switching interval are determined based on the physical boundary features on the preset transition path, including door frames, thresholds, ground markings, or building edges. The switching interval is a continuous path segment from the first physical boundary feature to the second physical boundary feature, wherein the first physical boundary feature is located on the indoor area side and the second physical boundary feature is located on the outdoor area side.

[0055] In this embodiment, by using stable and easily identifiable physical features in the environment as the basis for switching triggers, the situation where triggers based on mileage or time are easily affected by accumulated errors is avoided, ensuring the absolute accuracy of the switching timing. The double boundary ensures that the switching process is completed within a defined physical space, avoiding switching too early or too late, and making the switching interval strictly correspond to the physical space.

[0056] In one embodiment, the smooth transition employs an interpolation algorithm to gradually adjust the weights of the first positioning information and the second positioning information in the positioning output, so that the positioning output value changes continuously. Within the switching interval, a normalized position coefficient is calculated based on the robot's current position on the preset transition path, and the weights of the first positioning information and the second positioning information are dynamically determined based on the normalized position coefficient, so that the weights change continuously as the robot moves.

[0057] In this embodiment, an interpolation algorithm is used to fundamentally eliminate coordinate jumps at switching points, achieving a truly "seamless" transition and ensuring the continuity and stability of the robot's pose perception. By mapping weights to the robot's actual position, a smooth, spatially adaptive transition is achieved, ensuring that the switching process precisely matches the physical position.

[0058] In one embodiment, the interpolation algorithm is either linear interpolation or S-curve interpolation. When S-curve interpolation is used, the rate of change of weights is smaller at the beginning and end of the switching interval and larger in the middle. This further optimizes the smoothness of the switching process, avoids small acceleration abrupt changes caused by linear changes in weights, and achieves a smoother transition experience.

[0059] In one embodiment, a multi-sensor data fusion step is also included, which dynamically adjusts the fusion weights of LiDAR, real-time dynamic differential positioning module, inertial measurement unit and odometry in positioning calculation according to the indoor area, outdoor area or switching interval where the robot is located; wherein, LiDAR has the highest weight in indoor area, RTK has the highest weight in outdoor area, and the corresponding relationship query results are used as the main positioning basis in switching interval.

[0060] In this embodiment, by using the corresponding sensors as the main sensors in the optimal indoor and outdoor scenarios and using fixed mapping as the main sensors in the switching interval, the problem of unstable sensor signals in the transition area is solved, ensuring the continuity and reliability of positioning in all scenarios.

[0061] In one embodiment, a periodic calibration step is also included, which includes recalibrating the positioning reference and / or reacquiring the path update correspondence.

[0062] In this embodiment, by actively correcting errors introduced by environmental changes or sensor drift, the system maintains high accuracy over a long operating period (such as several months or even several years).

[0063] In one embodiment, the periodic calibration process includes three levels: daily calibration, periodic calibration, and quarterly calibration, with different calibration frequencies and depths at each level.

[0064] In this embodiment, tiered calibration optimizes resource utilization while ensuring system stability. Routine calibration enables rapid verification, periodic calibration provides moderate updates, and quarterly calibration ensures comprehensive recalibration, thereby avoiding resource waste caused by over-calibration.

[0065] In one embodiment, routine calibration is achieved by a robot automatically returning to the charging station. The characteristics of the charging station are matched with LiDAR to verify the deviation of the first positioning reference. If the deviation exceeds a preset threshold, the first positioning reference is automatically updated with the current matching result.

[0066] In this embodiment, automated calibration reduces human intervention and ensures that the benchmark is always accurate; threshold-based grading (fine-tuning for small deviations and alarms for large deviations) enables refined management.

[0067] In one embodiment, the method further includes loop closure detection and global optimization steps: loop closure detection is achieved by robot repositioning in indoor areas, and loop closure detection is achieved by global descriptor matching in outdoor areas, and multiple factors are uniformly modeled using factor graph optimization.

[0068] In this embodiment, strong loopback is achieved indoors using charging piles, and weak loopback is achieved outdoors using scene recognition; factor graph unified optimization effectively suppresses long-term cumulative drift and ensures global consistency of indoor and outdoor maps.

[0069] In one embodiment, the method further includes: switching to a backup reference when the primary positioning reference fails, adjusting the fusion weights when the sensor fails, triggering calibration when the correspondence deviation exceeds the standard, and pausing the switching when the path is temporarily blocked.

[0070] In this embodiment, a robust fault-tolerance mechanism is established to cover various abnormal situations such as reference failure, sensor failure, data deviation, and environmental interference, thereby enabling the system to operate reliably when facing complex working conditions, thus improving robustness and practicality.

[0071] A seamless indoor and outdoor map integration system for robots includes: An anchor point calibration module is used to establish a first positioning reference and a second positioning reference based on physical entities in indoor and outdoor areas, respectively, and to construct first feature information for identifying the first positioning reference and second feature information for identifying the second positioning reference; wherein, the first positioning reference and the second positioning reference are independent positioning references. The path mapping module is used to control the robot to travel along a preset transition path before the robot performs a task, and to simultaneously collect first positioning information based on the first positioning reference and second positioning information based on the second positioning reference at multiple locations on the preset transition path, so as to generate and store the correspondence between the first positioning information and the second positioning information. The smooth transition module is used to, when the robot travels along the preset transition path and enters the preset transition interval, smoothly transition the robot's positioning output from the first positioning information based on the first positioning reference to the second positioning information based on the second positioning reference, or perform the reverse smooth transition, based on the stored correspondence.

[0072] The foregoing description of specific exemplary embodiments of the present invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it is obvious that many changes and variations can be made based on the above teachings. Although embodiments of the invention have been shown and described, these specific embodiments are merely explanations of the invention and are not intended to limit it. The specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. The purpose of selecting and describing exemplary embodiments is to explain the specific principles of the invention and its practical application, so that those skilled in the art, after reading this specification, can make modifications, substitutions, variations, and various choices and changes to the embodiments as needed without departing from the principles and spirit of the invention, provided that such modifications, substitutions, variations, and choices and changes are within the scope of the claims of the invention and are protected by patent law.

Claims

1. A method for seamlessly connecting indoor and outdoor maps for a robot, characterized in that, Includes the following steps: A first positioning reference and a second positioning reference based on physical entities are established in the indoor area and the outdoor area respectively, and a first feature information for identifying the first positioning reference and a second feature information for identifying the second positioning reference are constructed; wherein, the first positioning reference and the second positioning reference are independent positioning references. Before the robot performs a task, the robot is controlled to travel along a preset transition path, and first positioning information based on the first positioning reference and second positioning information based on the second positioning reference are collected at multiple locations on the preset transition path simultaneously, so as to generate and store the correspondence between the first positioning information and the second positioning information. When the robot travels along the preset transition path and enters the preset switching interval, based on the stored correspondence, the robot's positioning output is smoothly transitioned from the first positioning information based on the first positioning reference to the second positioning information based on the second positioning reference, or the reverse smooth transition is performed.

2. The method for seamlessly connecting indoor and outdoor maps of a robot according to claim 1, characterized in that, A first positioning reference and a second positioning reference based on physical entities are established in the indoor area and the outdoor area respectively, and first feature information for identifying the first positioning reference and second feature information for identifying the second positioning reference are constructed, including: The first positioning reference is established based on indoor fixed facilities, including charging piles. The charging piles have preset charging positions. The robot is controlled to stop at the preset charging positions, and the current pose information of the robot is set as the first positioning reference. The charging piles are then scanned using a lidar to generate the first feature information. The second positioning reference is established based on outdoor landmark environmental features, which are fixed landmark street scene structures. The global coordinates of the fixed landmark street scene structure are obtained through real-time dynamic differential positioning technology as the second positioning reference, and the street scene structure is scanned by LiDAR to generate the second feature information.

3. The method for seamlessly connecting indoor and outdoor maps of a robot according to claim 1, characterized in that, When controlling the robot to travel along the preset transition path, the robot pose information corresponding to each position point is recorded synchronously, and the pose information is used as a component of the first positioning information or the second positioning information.

4. The method for seamlessly connecting indoor and outdoor maps of a robot according to claim 1, characterized in that, Controlling the robot to travel along a preset transition path, and simultaneously collecting first positioning information based on the first positioning reference and second positioning information based on the second positioning reference at multiple locations along the preset transition path, including: The robot is controlled to travel back and forth along the preset transition path multiple times, and the positioning information collected multiple times is fused to generate the corresponding relationship.

5. The method for seamlessly connecting indoor and outdoor maps of a robot according to claim 4, characterized in that, The correspondence is a coordinate mapping table, which records the one-to-one correspondence between the first coordinate and the second coordinate of discrete position points sampled at equal intervals of spatial distance on the preset transition path; The coordinate mapping table is generated in the following way: The location information collected multiple times is aligned in time and space. After removing outliers, the average of the first and second coordinates of the corresponding location points from each collection is taken as the coordinate value of the location point in the coordinate mapping table.

6. The method for seamlessly connecting indoor and outdoor maps of a robot according to claim 1, characterized in that, The starting and ending points of the switching interval are determined based on the physical boundary features on the preset transition path, including door frames, thresholds, ground markings, or building edges.

7. The method for seamlessly connecting indoor and outdoor maps of a robot according to claim 6, characterized in that, The switching interval is a continuous path segment from the first physical boundary feature to the second physical boundary feature, wherein the first physical boundary feature is located on the indoor area side and the second physical boundary feature is located on the outdoor area side.

8. The method for seamlessly connecting indoor and outdoor maps of a robot according to claim 1, characterized in that, The smooth transition employs an interpolation algorithm to gradually adjust the weights of the first and second positioning information in the positioning output, so that the positioning output value changes continuously.

9. A method for seamlessly connecting indoor and outdoor maps of a robot according to claim 8, characterized in that, Within the switching interval, a normalized position coefficient is calculated based on the robot's current position on the preset transition path, and the weights of the first positioning information and the second positioning information are dynamically determined based on the normalized position coefficient, so that the weights change continuously as the robot moves.

10. A seamless indoor and outdoor map integration system for robots, characterized in that, include: An anchor point calibration module is used to establish a first positioning reference and a second positioning reference based on physical entities in indoor and outdoor areas, respectively, and to construct first feature information for identifying the first positioning reference and second feature information for identifying the second positioning reference; wherein, the first positioning reference and the second positioning reference are independent positioning references. The path mapping module is used to control the robot to travel along a preset transition path before the robot performs a task, and to simultaneously collect first positioning information based on the first positioning reference and second positioning information based on the second positioning reference at multiple locations on the preset transition path, so as to generate and store the correspondence between the first positioning information and the second positioning information. The smooth transition module is used to, when the robot travels along the preset transition path and enters the preset transition interval, smoothly transition the robot's positioning output from the first positioning information based on the first positioning reference to the second positioning information based on the second positioning reference, or perform the reverse smooth transition, based on the stored correspondence.

Images