A method, control device and storage medium for robot localization and navigation without images
By using time matching and graph optimization of odometry frames and GPS positioning data in outdoor robots, navigation paths are dynamically established, solving the navigation problem in mapless scenarios. This enables robots to navigate autonomously without the need to pre-build environmental maps, ensuring the real-time performance and accuracy of the system.
Patent Information
- Application Number
- CN202611475638.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-21
- Publication Date
- 2026-11-10
AI Technical Summary
Existing outdoor robot navigation technologies require pre-built maps and cannot perform navigation tasks in scenarios without maps.
By acquiring the robot's odometer frames and GPS positioning data, a sliding window is used for time matching and graph optimization to establish the transformation relationship between the odometer coordinate system and the GPS coordinate system, and a local navigation path is dynamically generated.
It enables robots to navigate in mapless scenarios, solves the problem of inconsistency between third-party path coordinates and robot control coordinate system, realizes mapless navigation that can be used immediately after powering on, and ensures system real-time performance by controlling the amount of computation through a sliding window.
Smart Images

Figure CN123065443A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of robotics technology, and more specifically, to a method, control device, and storage medium for robot localization and navigation without a map. Background Technology
[0002] With the continuous development of robotics technology, outdoor mobile robots have been widely used in scenarios such as park delivery, unmanned cleaning, and security patrol. These robots typically need to move autonomously along designated routes and have the ability to bypass temporary obstacles in order to complete tasks such as transportation, cleaning, or patrolling.
[0003] Existing outdoor robot navigation solutions typically rely on pre-built maps for localization and navigation. The robot first constructs a point cloud map or grid map using SLAM technology, and then matches real-time perception data with the pre-built map during runtime to achieve localization and navigation.
[0004] The above navigation method requires pre-built maps. In scenarios where no pre-built maps are available, the robot cannot perform navigation tasks. Summary of the Invention
[0005] The purpose of this application is to address the shortcomings of the prior art by providing a robot localization and navigation method, control device, and storage medium that enable the robot to navigate normally even in mapless scenarios.
[0006] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows: In a first aspect, one embodiment of this application provides a map-free localization and navigation method for a robot, the method comprising: Acquire the robot's odometer frames and GPS positioning data; Store the odometer frame in a sliding window; The GPS positioning data is time-matched with the odometer frame, and the successfully matched GPS positioning data is bound to the corresponding odometer frame to obtain the GPS constraint frame. When the number of GPS constraint frames in the sliding window reaches a preset threshold, graph optimization is performed based on each odometer frame and each GPS constraint frame in the sliding window to align the pose of each odometer frame to the GPS coordinate system, thereby obtaining the pose of each odometer frame in the GPS coordinate system. Based on the pose of each odometer frame in the GPS coordinate system, the transformation relationship between the odometer coordinate system and the GPS coordinate system is determined; The system receives a target global path from a third-party map service and converts the target global path to the odometer coordinate system based on the conversion relationship to obtain a local navigation path.
[0007] In some embodiments, the step of time-matching the GPS positioning data with the odometer frame and binding the successfully matched GPS positioning data to the corresponding odometer frame to obtain a GPS constraint frame includes: Determine whether the timestamp of the GPS positioning data is within the preset time range of the sliding window; If so, then find the target odometer frame with the smallest time difference from the GPS positioning data in the sliding window; When the time difference between the target odometer frame and the GPS positioning data is less than a preset time difference threshold, the GPS positioning data is bound to the target odometer frame to obtain the GPS constraint frame.
[0008] In some embodiments, the step of time-matching the GPS positioning data with the odometer frame and binding the successfully matched GPS positioning data to the corresponding odometer frame to obtain a GPS constraint frame includes: The GPS positioning data is converted to the Northeast ENU coordinate system to obtain the first location coordinates; Obtain the second location coordinates in the ENU coordinate system from the previously bound GPS positioning data; Calculate the distance between the first position coordinates and the second position coordinates; When the distance is greater than the preset distance threshold, the GPS positioning data is bound to the corresponding odometer frame.
[0009] In some embodiments, the step of performing graph optimization based on each odometer frame and each GPS constraint frame in the sliding window to align the pose of each odometer frame to the pose in the Global Navigation Satellite System coordinate system, thereby obtaining the pose of each odometer frame in the GPS coordinate system, includes: Each odometer frame in the sliding window is treated as a pose node; Based on the coordinates of each odometer frame in the odometer coordinate system, add relative odometer constraints between adjacent pose nodes; Based on the coordinates of each GPS constraint frame in the GPS coordinate system, GPS position constraints are added at the pose nodes corresponding to each GPS constraint frame. The poses of each pose node are adjusted to minimize the difference between the pose of each odometer frame in the GPS coordinate system and the GPS coordinates of each GPS constraint frame, and to minimize the difference between the relative motion between each odometer frame and the odometer coordinates of each odometer frame, thereby obtaining the pose of each odometer frame in the GPS coordinate system.
[0010] In some embodiments, the step of transforming the global path to the odometry coordinate system based on the transformation relationship to obtain a local navigation path includes: Using the latest completed odometer frame in the sliding window as the anchor point, and based on the pose of the anchor point in the odometer coordinate system and the pose in the GPS coordinate system, each path point in the global path is transformed to the odometer coordinate system to obtain the local navigation path.
[0011] In some embodiments, before storing the odometer frame in the sliding window, the method includes: Determine whether the number of frame data included in the sliding window exceeds a preset threshold; If the number of frames exceeds the limit, then a portion of the frame data in the sliding window will be deleted.
[0012] In some embodiments, deleting a portion of the frame data in the sliding window includes: The odometer frames that are not keyframes and are not GPS constraint frames in the sliding window are identified as frame data to be deleted. Delete the frame data to be deleted in the sliding window.
[0013] In some embodiments, before receiving the target global path issued by the third-party map service, the method further includes: Receive the initial global path issued by the third-party map service; Obtain the robot's current GPS location; Calculate the offset based on the current GPS location and the first point in the initial global path; Based on the offset, each path point in the initial global path is offset to obtain the offset path points; The coordinates of each offset path point are transformed into the ENU coordinate system to obtain the target global path.
[0014] Secondly, another embodiment of this application provides a robot map-free localization and navigation device, the device comprising: an acquisition module, a matching module, an optimization module, a determination module, and a conversion module, wherein: The acquisition module is used to acquire the robot's odometer frames and GPS positioning data; and store the odometer frames in a sliding window; The matching module is used to perform time matching between the GPS positioning data and the odometer frame, and bind the successfully matched GPS positioning data to the corresponding odometer frame to obtain a GPS constraint frame. The optimization module is used to perform graph optimization based on each odometer frame and each GPS constraint frame in the sliding window when the number of GPS constraint frames in the sliding window reaches a preset threshold, and to align the pose of each odometer frame to the GPS coordinate system to obtain the pose of each odometer frame in the GPS coordinate system. The determining module is used to determine the transformation relationship between the odometer coordinate system and the GPS coordinate system based on the pose of each odometer frame in the GPS coordinate system; the transformation module is used to receive the target global path issued by a third-party map service, and transform the target global path to the odometer coordinate system based on the transformation relationship to obtain a local navigation path.
[0015] In some embodiments, the matching module is specifically used to determine whether the timestamp of the GPS positioning data is within a preset time range of the sliding window; if so, it searches for a target odometer frame with the smallest time difference from the GPS positioning data in the sliding window; when the time difference between the target odometer frame and the GPS positioning data is less than a preset time difference threshold, it binds the GPS positioning data to the target odometer frame to obtain the GPS constraint frame.
[0016] In some embodiments, the matching module is specifically used to convert the GPS positioning data to the ENU coordinate system to obtain a first location coordinate; obtain the second location coordinate of the previously bound GPS positioning data in the ENU coordinate system; calculate the distance between the first location coordinate and the second location coordinate; and bind the GPS positioning data to the corresponding odometer frame when the distance is greater than the preset distance threshold.
[0017] In some embodiments, the optimization module is specifically used to treat each odometer frame in the sliding window as a pose node; add odometer relative constraints between adjacent pose nodes according to the coordinates of each odometer frame in the odometer coordinate system; add GPS position constraints at the pose nodes corresponding to each GPS constraint frame according to the coordinates of each GPS constraint frame in the GPS coordinate system; adjust the pose of each pose node to minimize the difference between the pose of each odometer frame in the GPS coordinate system and the GPS coordinates of each GPS constraint frame, and minimize the difference between the relative motion between each odometer frame and the odometer coordinates of each odometer frame, thereby obtaining the pose of each odometer frame in the GPS coordinate system.
[0018] In some embodiments, the conversion module is specifically used to use the latest completed map optimization odometer frame in the sliding window as the anchor point, and based on the pose of the anchor point in the odometer coordinate system and the pose in the GPS coordinate system, convert each path point in the global path to the odometer coordinate system to obtain the local navigation path.
[0019] In some embodiments, the determining module is specifically used to determine whether the number of frame data included in the sliding window exceeds a preset threshold; if it does, then some of the frame data in the sliding window is deleted.
[0020] In some embodiments, the determining module is specifically used to determine that the odometer frame that is not a keyframe and is not a GPS constraint frame in the sliding window is the frame data to be deleted; and to delete the frame data to be deleted in the sliding window.
[0021] In some embodiments, the determining module is specifically configured to receive an initial global path issued by the third-party map service; obtain the robot's current GPS position; calculate an offset based on the current GPS position and the first point in the initial global path; offset each path point in the initial global path based on the offset to obtain the offset path points; and convert the coordinates of the offset path points to the ENU coordinate system to obtain the target global path.
[0022] Thirdly, another embodiment of this application provides a control device, including: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the control device is running, the processor communicates with the memory via the bus, and the processor executes the machine-readable instructions to perform the steps of any of the methods described in the first aspect above.
[0023] Fourthly, another embodiment of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of any of the methods described in the first aspect above.
[0024] The beneficial effects of this application are as follows: Odometer frames and GPS positioning data are received synchronously through a sliding window. Time-matched GPS data is bound to the corresponding odometer frame to form a GPS constraint frame. When the number of GPS constraint frames in the sliding window reaches a preset threshold, graph optimization is triggered. Based on GPS position constraints, the poses of each odometer frame within the sliding window are aligned to the GPS coordinate system, thereby dynamically establishing and continuously updating the transformation relationship between the odometer and GPS coordinate systems. Based on this transformation relationship, the global path of the target provided by the third-party map service is converted from the GPS coordinate system to the odometer coordinate system in real time, obtaining a local navigation path. The robot can directly use this local navigation path for navigation and movement. Using the method provided in this application, there is no need to pre-build an environmental map, solving the problem of inconsistency between the third-party path coordinates and the robot's control coordinate system. This achieves map-free navigation capability with automatic coordinate alignment upon startup, while controlling the computational load through the sliding window settings to ensure system real-time performance. Attached Figure Description
[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 This paper illustrates a schematic diagram of the architecture of a service system provided in an embodiment of this application. Figure 2 A schematic flowchart of a graph optimization process provided in an embodiment of this application is shown; Figure 3 A schematic diagram of the process for obtaining a GPS constraint frame according to an embodiment of this application is shown; Figure 4 A schematic diagram of the process for obtaining the target global path is shown in one embodiment of this application; Figure 5 This illustration shows a schematic diagram of a robot mapless positioning and navigation device according to an embodiment of this application; Figure 6 A schematic diagram of the structure of a computer device provided in an embodiment of this application is shown. Detailed Implementation
[0027] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.
[0028] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0029] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.
[0030] The method provided in this application can be applied to autonomous navigation of robots in outdoor environments, wherein the robot's work area needs to be able to receive, or at least for part of the time, a global navigation satellite system positioning signal, and has the ability to access third-party map services to obtain a global path.
[0031] The robots can be wheeled, tracked, or legged, including but not limited to park guide robots, park delivery robots, unmanned cleaning vehicles, security patrol robots, outdoor service robots, and humanoid robots. These robots must be equipped with odometry, such as wheeled odometry, visual odometry, or inertial odometry; a global navigation satellite system positioning module, such as a GPS / RTK module; and environmental sensors for local perception, such as LiDAR or visual sensors. Furthermore, the robots must have network connectivity to communicate with third-party map services.
[0032] The method provided in this application is applicable to outdoor navigation scenarios where it is impossible or unnecessary to pre-build environmental maps, such as open or semi-open environments like industrial parks, campuses, communities, parks, and urban roads. In these scenarios, the robot dynamically establishes local navigation capabilities using global navigation satellite system positioning data, odometer data, and real-time perception data, and autonomously walks along the global path provided by a third-party map service and performs corresponding tasks.
[0033] Figure 1 This application provides a schematic flowchart of a map-free positioning and navigation method according to an embodiment of the present application. Figure 1 As shown, the method includes steps 110-160, wherein: Step 110: Obtain the robot's odometer frames and GPS positioning data.
[0034] An odometry frame is a data set output by a robot's wheeled odometry at a specific moment, including at least the timestamp of that moment and the robot's pose (position and attitude angle) in the odometry coordinate system. The odometry coordinate system is a local coordinate system with the robot's starting position as the origin, obtained through integration by the wheeled encoder. It has high short-term accuracy but suffers from drift error that accumulates over time.
[0035] GPS positioning data is the positioning result output by the Global Navigation Satellite System receiver, which includes at least latitude, longitude, and elevation coordinates and a timestamp.
[0036] In some embodiments, after the machine starts up, the data receiving module continuously receives GPS positioning data. Each time the system receives a frame of GPS data, it checks whether the current initial location (Home) point has been set and determines whether the GPS data frame is valid, such as whether the positioning status is a fixed solution and whether the positioning accuracy meets the requirements. If the Home point has not been set and the current GPS data is valid, the Home initialization module sets the latitude, longitude, and altitude coordinates of the GPS frame as the Home point and calls the geographic coordinate transformation module to establish an ENU local coordinate system with the Home point as the origin. The three axes of this ENU coordinate system point east, north, and sky, respectively. All subsequently received GPS latitude, longitude, and altitude coordinates can be projected onto this Cartesian coordinate system through geographic coordinate transformation to obtain (x, y, z) position coordinates in meters. Once the Home point is set, it is locked and will not be reset due to fluctuations in subsequent GPS data, ensuring the fixity of the ENU coordinate system origin and ensuring that the ENU coordinates of the same physical location remain consistent at different times.
[0037] When the Home point is not yet initialized, the system reports an uninitialized state to the external system, informing the downstream navigation module that the ENU coordinate system has not yet been established. Based on this state, the navigation module determines whether the transformed local navigation path can be used. In the uninitialized state, the navigation module does not use any path tracking instructions and can only perform stationary waiting or obstacle avoidance behavior within a limited range based on local perception data, waiting for the system to complete initialization.
[0038] Once the Home point is initialized, the system reports its initialization status, indicating that the ENU coordinate system has been established and any GPS latitude, longitude, or altitude can be converted to ENU coordinates. At this point, the system is capable of receiving global paths from third-party map services.
[0039] Through the aforementioned status reporting mechanism, the downstream navigation module does not need to care whether the coordinate system is ready. It only needs to decide whether to perform path tracking based on the initialization status reported by the system, thus achieving decoupling between system status and navigation behavior.
[0040] Step 120: Store the odometer frame in the sliding window.
[0041] A sliding window is a fixed-length data cache container used to store odometer frames from the most recent period in chronological order.
[0042] Before storing the odometer frames into the sliding window, it is necessary to detect the number of frames in the sliding window to determine whether the number of frame data included in the sliding window exceeds a preset threshold. If it does, some frame data in the sliding window will be deleted.
[0043] The sliding window setting allows graph optimization to always be based on recent local data, avoiding the computational explosion caused by using all historical data, and also avoiding the problem of decreased alignment accuracy due to excessively outdated data with significant odometry drift. Furthermore, by limiting the window size, real-time operation can be achieved on embedded platforms or low-computing-power robot controllers.
[0044] As an example, the maximum number of frames in the sliding window can be set to 50. If the robot outputs odometry frames at a frequency of 50Hz, the time span covered by the sliding window is approximately 1 second. When the 51st frame is received, the sliding window is full and no new odometry frames can be added. At this point, the deletion of some frame data within the sliding window is triggered, such as randomly selecting the frame data from the first frame. The 51st frame is then stored at the end of the sliding window, ensuring that the sliding window always maintains a capacity of 50 frames. During straight-line travel, the pose of the odometry frames is continuously updated, and the spatiotemporal range covered by the sliding window continues to slide forward.
[0045] Step 130: Time-match the GPS positioning data with the odometer frame, and bind the successfully matched GPS positioning data to the corresponding odometer frame to obtain the GPS constraint frame.
[0046] The timestamp of the GPS positioning data is compared with the timestamp of each odometer frame in the sliding window. The odometer frame with the smallest time difference is found. If the time difference is less than the preset time difference threshold (such as 0.2 seconds), it is considered that the two are successfully matched in time, and the GPS positioning data is bound to that odometer frame.
[0047] The GPS constraint frame contains both the pose from the odometer coordinate system and the position from the GPS coordinate system, and is the key anchor point for subsequent connection between the odometer coordinate system and the GPS coordinate system.
[0048] The aforementioned method of associating GPS positioning data with odometer frames through time matching allows each GPS constraint frame to simultaneously possess coordinate information in two coordinate systems, providing the foundational data for establishing coordinate system transformation relationships in subsequent graph optimization. Furthermore, the preset time difference threshold avoids erroneous matching, ensuring the accuracy of the constraints.
[0049] In some embodiments, it can be first determined whether the timestamp of the GPS positioning data is within the preset time range of the sliding window; if so, the target odometer frame with the smallest time difference from the GPS positioning data is found in the sliding window; when the time difference between the target odometer frame and the GPS positioning data is less than the preset time difference threshold, the GPS positioning data is bound to the target odometer frame to obtain the GPS constraint frame.
[0050] As an example, the system receives a GPS data frame with a timestamp of 08:00:02.500. The system checks that this timestamp is within the time range of the sliding window (08:00:01.000 to 08:00:03.000), and then iterates through the timestamps of 50 odometer frames in the sliding window. It finds an odometer frame with a timestamp of 08:00:02.480. The time difference between the two is 0.020 seconds, which is less than the preset threshold of 0.2 seconds. At this point, the match is considered successful, the GPS data is bound to this odometer frame, and the odometer frame with bound GPS data is marked as a GPS constraint frame.
[0051] Step 140: When the number of GPS constraint frames in the sliding window reaches a preset threshold, graph optimization is performed based on each odometer frame and each GPS constraint frame in the sliding window to align the pose of each odometer frame to the GPS coordinate system, thereby obtaining the pose of each odometer frame in the GPS coordinate system.
[0052] Alignment is the process of mapping the poses of all odometer frames in a sliding window from an odometer coordinate system to a GPS coordinate system through graph optimization. After alignment, the position representation of the odometer frame changes from coordinates in the odometer coordinate system to coordinates in the GPS coordinate system.
[0053] The preset threshold setting ensures that graph optimization is triggered only after a sufficient number of medium GPS constraint frames have been accumulated, thus avoiding inaccurate alignment results caused by performing graph optimization directly when there are insufficient medium GPS constraint frames.
[0054] Figure 2 A schematic flowchart of a graph optimization process provided in an embodiment of this application is shown, as follows: Figure 2 As shown, step 140 may include steps 141-144, wherein: Step 141: Treat each odometer frame in the sliding window as a pose node.
[0055] Pose nodes are the basic units in graph optimization, representing the robot's pose (position and attitude angle) at a given moment. Each pose node corresponds to an odometry frame and is the variable to be optimized. The goal of graph optimization is to adjust the poses of these pose nodes to minimize the total error across all constraints.
[0056] Step 142: Based on the coordinates of each odometer frame in the odometer coordinate system, add relative odometer constraints between adjacent pose nodes.
[0057] Adjacent pose nodes are pose nodes corresponding to two consecutive odometry frames in time sequence, such as node 1 and node 2, or node 2 and node 3. The relative motion between adjacent nodes is provided by odometry, and this relative relationship is maintained during the constraint optimization process.
[0058] In graph optimization, the edges connecting adjacent pose nodes represent the relative motion relationship between two adjacent odometry frames. This constraint is directly provided by the odometry; the relative pose change (translation and rotation) between two adjacent frames is obtained by odometry integration, and the relative pose of two adjacent nodes in the constraint graph optimization should be consistent with this value. The odometry relative constraint reflects the relative relationship between adjacent pose nodes, rather than the absolute position of a particular frame. This constraint method, leveraging the short-term high accuracy of the odometry, adds odometry relative constraints between adjacent pose nodes, ensuring that the relative motion relationship between frames within the sliding window is maintained during optimization, preventing distortion of the relative relationship between frames due to the introduction of GPS position constraints.
[0059] Step 143: Based on the coordinates of each GPS constraint frame in the GPS coordinate system, add GPS position constraints at the pose nodes corresponding to each GPS constraint frame.
[0060] GPS position constraints are absolute position constraints connected to pose nodes in graph optimization, representing the absolute position of the node in the GPS coordinate system. This constraint is provided by GPS positioning data. The GPS constraint frame, after transforming the GPS data to the ENU coordinate system, yields position coordinates (x, y, z), constraining the node's optimized position to be consistent with these GPS coordinates. GPS position constraints are absolute constraints, not relative constraints, and are used to provide a global reference. Introducing absolute GPS position information into graph optimization as constraints allows the odometry frame's pose to be pulled towards GPS coordinates during optimization, thus achieving overall alignment from the odometry coordinate system to the GPS coordinate system.
[0061] Step 144: Adjust the pose of each pose node to minimize the difference between the pose of each odometer frame in the GPS coordinate system and the GPS coordinates of each GPS constraint frame, and minimize the difference between the relative motion between each odometer frame and the odometer coordinates of each odometer frame, so as to obtain the pose of each odometer frame in the GPS coordinate system.
[0062] By jointly optimizing the relative constraints of the odometer and the position constraints of GPS, and by using the short-term high-precision relative motion information of the odometer and the absolute position information of GPS, the relative relationship between each frame within the sliding window is accurate, and the absolute accuracy of the overall pose in the GPS coordinate system is guaranteed.
[0063] As an example, the system continuously receives odometer frames and GPS data, and the number of GPS constraint frames in the sliding window gradually accumulates. When the number of GPS constraint frames in the sliding window reaches four, graph optimization is triggered. The four GPS constraint frames in the sliding window are located at times t1, t2, t4, and t7, respectively. The graph optimizer treats all 50 odometer frames as pose nodes, adds relative odometer constraints between adjacent odometer frames, and adds GPS position constraints at the four GPS constraint frames t1, t2, t4, and t7. The graph optimizer adjusts the pose of all nodes to minimize the sum of the GPS coordinate error of the GPS constraint frames and the relative motion error between adjacent frames, thereby obtaining the pose of each odometer frame in the GPS coordinate system.
[0064] Step 150: Based on the pose of each odometer frame in the GPS coordinate system, determine the transformation relationship between the odometer coordinate system and the GPS coordinate system.
[0065] The transformation relationship refers to the coordinate transformation parameters between the odometer coordinate system and the GPS coordinate system, including the rotation matrix or rotation angle and the translation vector. Multiplying any coordinate in the odometer coordinate system by the rotation matrix and adding the translation vector yields the corresponding coordinate in the GPS coordinate system. This transformation relationship forms the basis for subsequently transforming third-party paths from the GPS coordinate system to the odometer coordinate system, enabling the global path in the GPS coordinate system to be converted to the odometer coordinate system actually used by the robot chassis, thereby achieving path tracking.
[0066] Step 160: Receive the target global path from the third-party map service, and convert the target global path to the odometry coordinate system based on the transformation relationship to obtain the local navigation path.
[0067] The target global path is a complete driving route planned by a third-party map service based on the user-specified starting point and destination. It consists of a series of latitude, longitude, and altitude coordinates. The target global path reflects the pre-planned driving direction from the starting point to the destination, but does not include information on local obstacles.
[0068] A local navigation path is a sequence of path points obtained by transforming the global path of the target from the GPS coordinate system to the odometry coordinate system. Each point in the local navigation path is expressed as a planar coordinate in the odometry coordinate system, which can be directly used by the robot chassis's local planner to generate specific speed control commands.
[0069] The aforementioned path conversion method allows the paths provided by third-party map services to be directly used by the robot chassis, achieving a seamless connection from cloud planning to local execution without the need for manual intervention or additional coordinate calibration.
[0070] In the embodiments of this application, the latest completed odometer frame in the sliding window can be used as the anchor point. Based on the pose of the anchor point in the odometer coordinate system and the pose in the GPS coordinate system, each path point in the global path can be transformed to the odometer coordinate system to obtain the local navigation path.
[0071] Anchor points are the most recently optimized odometry frames within the sliding window, serving as reference points for coordinate transformation. The odometry frame corresponding to an anchor point simultaneously possesses the pose in both the odometry coordinate system and the GPS coordinate system, thus acting as a bridge between the two systems. The selection of anchor points determines the accuracy of subsequent path transformations; the closer the anchor point is to the robot's current position, the closer the connection between the transformed path and the robot's actual position.
[0072] Each path point in the global path is a path point in the target global path issued by a third-party map service. Before the conversion, these path points are expressed as ENU coordinates in the GPS coordinate system, and after the conversion, they are expressed as coordinates in the odometer coordinate system for use by the local planner.
[0073] The above method of using the latest completed map-optimized odometer frame as the anchor point ensures that the transformation reference is always based on the latest coordinate system alignment result, making the connection between the starting point of the path and the robot's current position the tightest and the error the smallest.
[0074] As an example, taking a robot vacuum cleaner as an example, the system plans a global path from the east gate to the west gate using a third-party map service, containing approximately 200 latitude, longitude, and altitude coordinates. After receiving this path, the vacuum cleaner first obtains its current GPS position (108.95°E, 34.27°N) and calculates the offset with the first point on the path (108.95°E, 34.27°N). The offset is 0, indicating that the current GPS position and the first point coincide, and no offset correction is needed. Subsequently, the system converts each path point from latitude, longitude, and altitude to ENU coordinates, calculates the heading angle for adjacent points, and writes it into the attitude information. The system obtains the transformation relationship determined by the most recent map optimization, uses the most recently aligned odometer frame in the sliding window as the anchor point, and converts the ENU path to the odometer coordinate system to obtain the local navigation path. The local planner generates speed commands based on the local navigation path, and the vacuum cleaner begins to travel along the path.
[0075] In the embodiments of this application, the current odometry pose, positioning status, confidence level, GPS alignment status, robot's current GPS coordinates, origin, third-party global path, and local navigation path in the odometry coordinate system can be simultaneously published to the navigation module by defining a unified downstream publishing message. This allows the navigation module to choose whether to perform path tracking or local obstacle avoidance based on the status.
[0076] Specifically, the above information is encapsulated into a unified message format and periodically published to downstream navigation modules. After subscribing to this message, the navigation module does not need to handle multi-source coordinate transformation itself; it can directly read the fields in the message to obtain the current complete positioning and path information. Specifically, when the GPS alignment status is "not aligned," the navigation module can choose to perform only local obstacle avoidance without performing path tracking; when the alignment status is "aligned," the navigation module can switch to path tracking mode and drive along the transformed local navigation path.
[0077] The conversion formula from the odometer (odom) coordinate system to the GPS coordinate system can be:
[0078] in, This represents the transformation matrix that transforms the nth point in the local coordinate system of the kth frame from the odometer coordinate system to the GPS coordinate system. This represents the pose transformation matrix of the k-th frame in the GPS coordinate system, i.e., the position and orientation of the origin of this frame in the GPS coordinate system. express The inverse matrix is used to transform points from the global odometry coordinate system to the local coordinate system of the k-th frame; This represents the pose transformation matrix of the k-th frame in the odometer coordinate system, i.e., the position and orientation of the origin of this frame in the odometer coordinate system. This represents the transformation matrix from the nth point in the local coordinate system of the kth frame to the origin of the odometer coordinate system.
[0079] The conversion formula for transforming third-party global path GPS coordinates to ODOM coordinates can be:
[0080] This represents the transformation matrix that converts the k-th path point from the GPS coordinate system to the odometer coordinate system. This represents the pose transformation matrix of the nth keyframe in the odometer coordinate system, that is, the position and orientation of the origin of the keyframe in the odometer coordinate system. This represents the transformation matrix from the nth keyframe to the nth path point; express The inverse matrix is used to transform the path points from the ENU coordinate system to a local coordinate system with the path points as the origin; This represents the pose transformation matrix of the nth path point in the GPS coordinate system with the Home point as the origin, i.e., the position of the path point in the ENU coordinate system; This represents the pose transformation matrix of the k-th path point in the GPS coordinate system with the Home point as the origin.
[0081] The above formula means using the anchor point, i.e., the latest completed odometer frame, as a bridge to transform the path points from the GPS coordinate system to the odometer coordinate system. Specifically, the path points are first transformed from the GPS coordinate system to the local coordinate system where the anchor point is located, and then the path points are transformed to the odometer coordinate system using the pose of the anchor point in the odometer coordinate system.
[0082] The robot localization and navigation method provided in this application synchronously receives odometry frames and GPS positioning data through a sliding window. Time-matched GPS data is bound to the corresponding odometry frame to form a GPS constraint frame. When the number of GPS constraint frames in the sliding window reaches a preset threshold, graph optimization is triggered. Based on GPS position constraints, the poses of each odometry frame within the sliding window are aligned to the GPS coordinate system, thereby dynamically establishing and continuously updating the transformation relationship between the odometry and GPS coordinate systems. Based on this transformation relationship, the target global path provided by a third-party map service is converted from the GPS coordinate system to the odometry coordinate system in real time to obtain a local navigation path. The robot can directly use the local navigation path for navigation and movement. Using the method provided in this application, there is no need to pre-build an environmental map, solving the problem of inconsistency between third-party path coordinates and the robot's control coordinate system. This achieves mapless navigation capability with automatic coordinate alignment and immediate robot startup. Furthermore, the computational load is controlled through the sliding window settings to ensure system real-time performance.
[0083] In some embodiments, when deleting some frame data in a sliding window, the odometer frames that are not key frames and are not GPS constraint frames in the sliding window can be identified as the frame data to be deleted; the frame data to be deleted can then be deleted in the sliding window.
[0084] Keyframes are odometry frames marked as priority to be retained within a sliding window. Keyframes are typically set based on the robot's motion state; for example, when the robot moves more than a certain distance, rotates more than a certain angle, or passes a preset time interval, the currently acquired odometry frame is marked as a keyframe.
[0085] Setting keyframes allows the sliding window to cover a larger spatiotemporal range within the same window size by retaining keyframes, avoiding the problem of insufficient time span and constraint information caused by continuous and dense ordinary frames occupying the window space of the sliding window.
[0086] The method described above, which prioritizes deleting ordinary frames that are neither keyframes nor GPS constraint frames, ensures that the critical constraint information required for graph optimization is not lost while controlling the sliding window size. Keyframes guarantee the breadth of coverage of the sliding window in the time dimension, while GPS constraint frames guarantee the number of anchor points required for coordinate system alignment. Both keyframes and GPS constraint frames are preferentially retained in the sliding window, thereby avoiding insufficient GPS constraints due to continuous deletion of GPS constraint frames and optimization trigger delays caused by keyframe deletion.
[0087] As an example, the maximum capacity of the sliding window is still set to 50 frames. The sliding window currently contains 50 frames, of which frames 5 and 12 are keyframes, marked when the robot starts and turns, respectively; frames 8, 23, 31, and 42 are GPS constraint frames; and the remaining 44 frames are ordinary frames. When the system receives frame 51, the sliding window is full, triggering the deletion of some frame data within the window. The system traverses the sliding window, identifying 44 ordinary frames that are neither keyframes nor GPS constraint frames. The earliest frame from these ordinary frames is selected as the frame to be deleted, and this frame is deleted from the window. Frame 51 is then stored at the end of the sliding window. After deletion, the sliding window still retains 2 keyframes and 4 GPS constraint frames, ensuring that the constraint information required for subsequent graph optimization remains unaffected.
[0088] Figure 3 This application provides a schematic diagram of a process for obtaining GPS constraint frames according to an embodiment of the present application. Figure 3 As shown, step 130 may include: steps 131-133, wherein: Step 131: Determine whether the timestamp of the GPS positioning data is within the preset time range of the sliding window.
[0089] The preset time range is the time interval covered by the timestamps of the earliest and latest odometer frames within the sliding window. For example, if the oldest frame in the sliding window has a timestamp of 08:00:01.000 and the latest frame has a timestamp of 08:00:03.000, then the preset time range is [08:00:01.000, 08:00:03.000]. Only GPS data with timestamps falling within this range can potentially be matched with odometer frames in the sliding window; GPS data outside this range is discarded because no sufficiently close odometer frames can be found.
[0090] By using the aforementioned filtering method with preset time ranges, GPS data that does not match the time span of the sliding window can be quickly excluded, avoiding invalid traversal and matching operations. When the timestamp of GPS data is earlier than the latest frame of the sliding window or later than the oldest frame of the window, the GPS data cannot establish a valid time match with any odometer frame, and therefore can be discarded in advance, thereby saving computing resources.
[0091] If so, proceed to step 132: find the target odometer frame with the smallest time difference from the GPS positioning data in the sliding window.
[0092] The time difference is the absolute value of the difference between the timestamp of GPS positioning data and the timestamp of a certain odometer frame. The smaller the time difference, the closer the GPS data and the odometer frame are in time, and the more likely they are to correspond in physical location.
[0093] The target odometer frame is the odometer frame with the smallest time difference from the current GPS positioning data in the sliding window, that is, the odometer frame most likely to be time-aligned with the GPS data.
[0094] The method described above, which involves traversing all odometer frames within a sliding window to find the odometer frame closest to the current GPS positioning data time, provides matching candidates for subsequent GPS binding. Since the number of frames in the sliding window is controlled by the upper limit of the sliding window, the computational cost of the traversal operation is predictable and bounded, ensuring real-time performance of the search.
[0095] Step 133: When the time difference between the target odometer frame and the GPS positioning data is less than the preset time difference threshold, the GPS positioning data is bound to the target odometer frame to obtain the GPS constraint frame.
[0096] Preset time difference threshold: The maximum allowed time difference between GPS positioning data and odometer frames, for example, 0.2 seconds. If this threshold is exceeded, the two are considered not close enough in time, and the binding is not very meaningful because the robot may have already made significant movements during this period.
[0097] By binding GPS positioning data to the target odometer frame, the odometer frame can simultaneously possess the pose in the odometer coordinate system and the position in the GPS coordinate system.
[0098] The aforementioned method of determining the time difference threshold avoids binding GPS data with excessively large time differences to odometry frames, thus ensuring the accuracy of GPS position constraints. If the time difference is too large, the robot may have already moved during that time, and binding GPS data to that frame would lead to excessive constraint errors, affecting the quality of subsequent graph optimization.
[0099] In some embodiments, to avoid the burden of subsequent optimization caused by excessively dense GPS positioning data, the GPS positioning data can be further filtered based on spatial distance. For example, after the current GPS positioning data is converted to the ENU coordinate system, the positional distance between it and the ENU coordinates corresponding to the GPS positioning data previously inserted into the sliding window is determined. Only when the aforementioned positional distance is greater than a preset distance threshold can the current GPS positioning data be bound to the corresponding mileage frame in the sliding window, thereby reducing computational overhead while ensuring the effectiveness of constraints.
[0100] The ENU coordinate system is a locally tangent Cartesian coordinate system with the Home point as the origin, pointing east, north, and zenith respectively. All GPS latitude, longitude, and altitude coordinates are projected onto the ENU coordinate system through geographic coordinate transformation, so that the position can be represented in (x, y, z) meters, which facilitates the calculation of positional distances between ENU coordinates and subsequent coordinate transformations.
[0101] In the embodiments of this application, GPS positioning data can be converted to the ENU coordinate system to obtain the first location coordinates; the second location coordinates of the previously bound GPS positioning data in the ENU coordinate system can be obtained; the distance between the first location coordinates and the second location coordinates can be calculated; when the distance is greater than a preset distance threshold, the GPS positioning data can be bound to the corresponding odometer frame.
[0102] The first position coordinates are the position coordinates obtained after converting the current GPS positioning data to the ENU coordinate system, denoted as (x1, y1, z1), in meters, and are used to compare the distance with the previously bound GPS position.
[0103] The second position coordinates are the position coordinates of the GPS positioning data that was successfully bound to the odometer frame last time in the ENU coordinate system, denoted as (x0, y0, z0), in meters, and serve as a reference point for distance comparison.
[0104] The preset distance threshold is the minimum distance allowed between two GPS binding locations, for example, 0.5 meters. If the distance between the current location and the previous binding location is less than the preset distance threshold, the physical locations corresponding to the two GPS data are considered too close, and the current GPS data does not provide new valid constraint information; therefore, the system discards it. Conversely, if the distance between the current location and the previous binding location is greater than or equal to the preset distance threshold, the current GPS data is considered to provide new location information, and binding is allowed.
[0105] The previously bound GPS location data is the most recent GPS location data successfully bound to a specific odometer frame in chronological order. The system saves this data and its position coordinates in the ENU coordinate system as a comparison benchmark before the next GPS binding.
[0106] The aforementioned spatial distance filtering mechanism avoids increasing the computational burden of graph optimization due to an excessive number of GPS constraint frames in the sliding window caused by overly dense GPS data. Since the output frequency of GPS positioning data is typically lower than that of odometry data, but its absolute position accuracy is higher, inserting every frame of GPS data into the sliding window could result in an overly dense array of GPS constraint frames, especially when the robot is moving at low speed or stationary. This would lead to almost no new positional information between adjacent GPS constraint frames, yet graph optimization would still need to process all frames, increasing computational overhead. By allowing new GPS bindings only after the robot has moved a sufficient distance, the number of constraint frames in the sliding window is controlled while ensuring the effectiveness of the constraints, thereby improving computational efficiency.
[0107] Figure 4 This application provides a schematic diagram of a process for obtaining a target global path according to an embodiment of the present application. Figure 4As shown, before step 150, the method further includes steps 1501-1505, wherein: Step 1501: Receive the initial global path issued by the third-party map service.
[0108] The initial global path is raw path data directly provided by a third-party map service, consisting of a series of latitude, longitude, and altitude coordinates, without any coordinate correction or coordinate system transformation. This path reflects the macroscopic driving direction from the starting point to the destination and does not include information on local obstacles.
[0109] By interfacing with a third-party map service, the robot no longer needs to build and maintain global road network information on its own. This third-party map service provides rich information such as road topology and traffic rules, and the initial global path output by the service is imported into the system as the initial basis for the robot's navigation.
[0110] Step 1502: Obtain the robot's current GPS location.
[0111] The current GPS position is the robot's latest GPS positioning data at the current moment, generated by GPS / Real-time Dynamic Differential Positioning (Real-time Dynamic Differential Positioning). The Time Kinematic (RTK) module outputs real-time coordinates, including at least latitude, longitude, and altitude.
[0112] Step 1503: Calculate the offset based on the current GPS location and the first point in the initial global path.
[0113] The starting point in the initial global path, the first latitude and longitude coordinate point, which is the starting position of the path planned by the third-party map service.
[0114] Offset is the coordinate difference between the current GPS position and the first point on the initial global path. Since there may be system offsets between third-party map coordinates and robot GPS coordinates, offset calculation and compensation can eliminate this deviation. Offset can be expressed directly as a difference in latitude and longitude, or as a difference in planar coordinates (in meters) after converting both to the ENU coordinate system.
[0115] By calculating the offset between the robot's current GPS position and the first point in the initial global path, a basis for subsequent path correction is provided. This offset reflects the deviation between the third-party map coordinate system and the robot's GPS coordinate system. After offset compensation for the first point in the initial global path, the relative relationship between the robot's current position and the first point in the initial global path is consistent with the actual situation, avoiding the problem of incorrect position of the first point in the initial global path due to coordinate system offset.
[0116] Step 1504: Offset each path point in the initial global path based on the offset to obtain the offset path points.
[0117] After offsetting each path point in the initial global path, the fixed deviation between the initial global path output by the third-party map service and the robot's GPS coordinates is eliminated. The relative relationship between the position of each path point in the GPS coordinate system after offset and the robot's current GPS position is accurate, avoiding the problem of incorrect path position due to coordinate system offset.
[0118] Step 1505: Transform the coordinates of each offset path point to the Northeast ENU coordinate system to obtain the target global path.
[0119] The target global path is the global path after offset correction and ENU coordinate transformation. Each point in the target global path is expressed as planar coordinates (x, y, z) in the ENU coordinate system, in meters. The target global path is the input data ultimately used for transformation to the odometry coordinate system. Planar coordinates in the ENU coordinate system can directly participate in map optimization and coordinate transformation calculations without the need for complex conversions between latitude, longitude, and altitude coordinates.
[0120] In the embodiments of this application, during the robot's navigation based on a local navigation path, the local planner can use the local path as a reference and combine it with occupancy (OCC) data to detect the distribution of obstacles ahead and around the path in real time. When the OCC grid indicates the presence of obstacles on the path, such as temporarily parked vehicles, pedestrians, or construction barriers, the planner generates a detour trajectory in real time, avoiding the obstacle and returning to the original path to continue driving, while ensuring that it does not deviate from the global path direction. When the OCC grid indicates that the path ahead is clear, the local planner directly generates speed control commands along the local navigation path. When the OCC occupancy grid detects an obstacle ahead, the robot generates a detour trajectory in real time based on the local navigation path and the OCC occupancy grid, avoiding the obstacle and returning to the local navigation path to continue driving. The above navigation is continuously performed during the robot's movement, achieving safe obstacle avoidance while achieving target guidance.
[0121] In some embodiments, in startup scenarios where both GPS and heading are reliable, after the robot starts, the GPS module can output fixed solution positioning data, while the inertial navigation system or dual-antenna GPS can provide reliable heading angle information. In this scenario, the system immediately sets the Home point and establishes the ENU coordinate system after receiving valid GPS data for the first time. Since the heading is known, the angular alignment between the ENU coordinate system and the odometer coordinate system can be directly determined, and the position alignment is initially established by matching the current GPS position with the current odometer position. When the number of GPS constraint frames in the sliding window reaches a preset threshold, graph optimization is triggered, and convergence is rapid based on the existing good initial values, thereby accurately determining the complete transformation relationship between the two coordinate systems. Third-party paths can be converted to the odometer coordinate system earlier for the robot to use.
[0122] In some embodiments, in startup scenarios where only a fixed GPS solution exists but no reliable heading is available—such as scenarios with only a single-antenna GPS, incomplete inertial navigation system initialization, or high heading angle noise—the robot can receive the fixed GPS solution to obtain position information upon startup, but lacks a reliable source of heading information. In this scenario, the system can still normally set the Home point to establish the ENU coordinate system, but due to the lack of accurate heading angles, the rotational relationship between the ENU and odometry coordinate systems cannot be directly established, and the global path cannot be temporarily converted to the odometry coordinate system. During robot movement, the system continuously receives odometry frames and GPS data, binding the time-matched GPS data to the corresponding odometry frame to obtain GPS constraint frames. Odometry data records the robot's position changes during movement, while GPS data records the robot's absolute position in the ENU coordinate system. As the robot moves, the number of GPS constraint frames in the sliding window gradually accumulates. When the number of GPS constraint frames reaches a preset threshold, graph optimization is triggered. By comprehensively optimizing the odometry relative constraints and GPS position constraints, the system can simultaneously estimate the position offset and angular rotation between the ENU and Odom coordinate systems. Once coordinate system alignment is complete, third-party path conversion can be enabled normally. Throughout the process, the robot moved normally without waiting for the heading information to be ready before starting, achieving the ability to start moving immediately upon power-on and automatically complete alignment during movement.
[0123] In some embodiments, in scenarios where there is no GPS signal upon startup, the robot is in an area without GPS signal coverage, such as indoors, underground parking garages, or densely built-up areas, and cannot obtain valid GPS positioning data. In this scenario, due to the lack of GPS positioning information, the system cannot set a Home point or complete positioning in the ENU coordinate system. The robot can first perform local exploration, obstacle avoidance, or waiting in place based on odometry data and real-time local perception grid data. For example, in an unknown environment, it can use odometry for short-term positioning, or use LiDAR or OCC perception grid data to detect surrounding obstacles in real time and perform basic obstacle avoidance behaviors. Once the robot moves to an area where GPS signal is restored and successfully receives valid GPS positioning data, the system immediately sets a Home point and establishes the ENU coordinate system, then gradually completes GPS / Odom coordinate alignment, enables third-party path conversion functions, and continues to perform subsequent navigation tasks.
[0124] In some embodiments, during a brief GPS loss scenario, the robot's GPS signal may be temporarily blocked during normal navigation due to areas such as buildings, tunnels, or tree-lined paths, preventing the system from receiving new GPS positioning data. In this scenario, the system has already completed GPS / Odom coordinate system alignment, and the sliding window stores the latest completed map-optimized odometry frames (i.e., aligned frames). When the GPS signal is lost, the system uses the precise pose of the aligned frames in the GPS coordinate system as a reference, combined with the relative motion between subsequent odometry frames, to continuously calculate the pose in the ENU coordinate system at subsequent times, and then converts it into latitude, longitude, and altitude coordinates for output, ensuring uninterrupted positioning results. The robot can continue to travel along the local navigation path during GPS loss. When the GPS signal is restored, the system re-receives GPS positioning data and resumes the normal binding and map optimization process. A new round of map optimization corrects the positioning errors accumulated due to odometry drift during the GPS loss, bringing the positioning output back to the correct position.
[0125] Based on the same inventive concept, this application also provides a robot mapless positioning and navigation device corresponding to the robot mapless positioning and navigation method. Since the principle of the device in this application is similar to the robot mapless positioning and navigation method described above in this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.
[0126] Figure 5 This application provides a schematic diagram of the structure of a robot mapless localization and navigation device according to an embodiment of the present application. Figure 5 As shown, the device includes: an acquisition module 210, a matching module 220, an optimization module 230, a determination module 240, and a conversion module 250, wherein: The acquisition module 210 is used to acquire the robot's odometer frames and GPS positioning data; and to store the odometer frames in a sliding window. Matching module 220 is used to time match GPS positioning data with odometer frames and bind the successfully matched GPS positioning data to the corresponding odometer frames to obtain GPS constraint frames; The optimization module 230 is used to perform graph optimization based on each odometer frame and each GPS constraint frame in the sliding window when the number of GPS constraint frames in the sliding window reaches a preset threshold, and to align the pose of each odometer frame to the GPS coordinate system to obtain the pose of each odometer frame in the GPS coordinate system. The determination module 240 is used to determine the transformation relationship between the odometer coordinate system and the GPS coordinate system based on the pose of each odometer frame in the GPS coordinate system. The conversion module 250 is used to receive the target global path sent by the third-party map service, and convert the target global path to the odometer coordinate system based on the conversion relationship to obtain the local navigation path.
[0127] In some embodiments, the matching module 220 is specifically used to determine whether the timestamp of the GPS positioning data is within a preset time range of the sliding window; if so, it searches for the target odometer frame with the smallest time difference from the GPS positioning data in the sliding window; when the time difference between the target odometer frame and the GPS positioning data is less than a preset time difference threshold, it binds the GPS positioning data to the target odometer frame to obtain a GPS constraint frame.
[0128] In some embodiments, the matching module 220 is specifically used to convert GPS positioning data to the ENU coordinate system to obtain a first position coordinate; obtain the second position coordinate of the previously bound GPS positioning data in the ENU coordinate system; calculate the distance between the first position coordinate and the second position coordinate; and bind the GPS positioning data to the corresponding odometer frame when the distance is greater than a preset distance threshold.
[0129] In some embodiments, the optimization module 230 is specifically used to treat each odometer frame in the sliding window as a pose node; add odometer relative constraints between adjacent pose nodes according to the coordinates of each odometer frame in the odometer coordinate system; add GPS position constraints at the pose nodes corresponding to each GPS constraint frame according to the coordinates of each GPS constraint frame in the GPS coordinate system; adjust the pose of each pose node to minimize the difference between the pose of each odometer frame in the GPS coordinate system and the GPS coordinates of each GPS constraint frame, and minimize the difference between the relative motion between each odometer frame and the odometer coordinates of each odometer frame, thereby obtaining the pose of each odometer frame in the GPS coordinate system.
[0130] In some embodiments, the conversion module 250 is specifically used to convert each path point in the global path to the odometer coordinate system based on the pose of the latest completed map optimization frame in the sliding window and the pose of the anchor point in the GPS coordinate system, thereby obtaining a local navigation path.
[0131] In some embodiments, the determining module 240 is specifically used to determine whether the number of frame data included in the sliding window exceeds a preset threshold; if it does, then some of the frame data in the sliding window is deleted.
[0132] In some embodiments, the determining module 240 is specifically used to determine that the odometer frames that are not keyframes and are not GPS constraint frames in the sliding window are frame data to be deleted; and to delete the frame data to be deleted in the sliding window.
[0133] In some embodiments, the determining module 240 is specifically used to receive an initial global path issued by a third-party map service; obtain the robot's current GPS position; calculate the offset based on the current GPS position and the first point in the initial global path; offset each path point in the initial global path based on the offset to obtain the offset path points; and transform the coordinates of the offset path points to the ENU coordinate system to obtain the target global path.
[0134] The processing flow of each module in the device and the interaction flow between each module can be referred to the relevant descriptions in the above method embodiments, and will not be detailed here.
[0135] This application also provides a computer device. Figure 6 This application provides a schematic diagram of the structure of a computer device according to an embodiment of the present application. Figure 6 As shown, it includes: a processor 61, a memory 62, and optionally, a bus 63. The memory 62 stores machine-readable instructions that can be executed by the processor 61. When the computer device is running, the processor 61 and the memory 62 communicate via the bus 63. When the machine-readable instructions are executed by the processor 61, the steps of the above-described robot mapless localization and navigation method are performed.
[0136] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the above-described robot localization and navigation method without map.
[0137] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some communication interfaces; the indirect coupling or communication connection of devices or modules can be electrical, mechanical, or other forms.
[0138] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.
[0139] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.
Claims
1. A map-free localization and navigation method for robots, characterized in that, The method includes: Acquire the robot's odometer frames and GPS positioning data; Store the odometer frame in a sliding window; The GPS positioning data is time-matched with the odometer frame, and the successfully matched GPS positioning data is bound to the corresponding odometer frame to obtain the GPS constraint frame. When the number of GPS constraint frames in the sliding window reaches a preset threshold, graph optimization is performed based on each odometer frame and each GPS constraint frame in the sliding window to align the pose of each odometer frame to the GPS coordinate system, thereby obtaining the pose of each odometer frame in the GPS coordinate system. Based on the pose of each odometer frame in the GPS coordinate system, the transformation relationship between the odometer coordinate system and the GPS coordinate system is determined; The system receives a target global path from a third-party map service and converts the target global path to the odometer coordinate system based on the conversion relationship to obtain a local navigation path.
2. The method according to claim 1, characterized in that, The step of matching the GPS positioning data with the odometer frame in time, and binding the successfully matched GPS positioning data to the corresponding odometer frame to obtain a GPS constraint frame, includes: Determine whether the timestamp of the GPS positioning data is within the preset time range of the sliding window; If so, then find the target odometer frame with the smallest time difference from the GPS positioning data in the sliding window; When the time difference between the target odometer frame and the GPS positioning data is less than a preset time difference threshold, the GPS positioning data is bound to the target odometer frame to obtain the GPS constraint frame.
3. The method according to claim 1, characterized in that, The step of matching the GPS positioning data with the odometer frame in time, and binding the successfully matched GPS positioning data to the corresponding odometer frame to obtain a GPS constraint frame, includes: The GPS positioning data is converted to the Northeast ENU coordinate system to obtain the first location coordinates; Obtain the second location coordinates in the ENU coordinate system from the previously bound GPS positioning data; Calculate the distance between the first position coordinates and the second position coordinates; When the distance is greater than a preset distance threshold, the GPS positioning data is bound to the corresponding odometer frame.
4. The method according to claim 1, characterized in that, The step of graph optimization based on each odometer frame and each GPS constraint frame in the sliding window, aligning the pose of each odometer frame to the GPS coordinate system, to obtain the pose of each odometer frame in the GPS coordinate system, includes: Each odometer frame in the sliding window is treated as a pose node; Based on the coordinates of each odometer frame in the odometer coordinate system, add relative odometer constraints between adjacent pose nodes; Based on the coordinates of each GPS constraint frame in the GPS coordinate system, GPS position constraints are added at the pose nodes corresponding to each GPS constraint frame. The poses of each pose node are adjusted to minimize the difference between the pose of each odometer frame in the GPS coordinate system and the GPS coordinates of each GPS constraint frame, and to minimize the difference between the relative motion between each odometer frame and the odometer coordinates of each odometer frame, thereby obtaining the pose of each odometer frame in the GPS coordinate system.
5. The method according to claim 1, characterized in that, The process of transforming the global path to the odometry coordinate system based on the transformation relationship to obtain the local navigation path includes: Using the latest completed odometer frame in the sliding window as the anchor point, and based on the pose of the anchor point in the odometer coordinate system and the pose in the GPS coordinate system, each path point in the global path is transformed to the odometer coordinate system to obtain the local navigation path.
6. The method according to claim 1, characterized in that, Before storing the odometer frame in the sliding window, the method includes: Determine whether the number of frame data included in the sliding window exceeds a preset threshold; If the number of frames exceeds the limit, then a portion of the frame data in the sliding window will be deleted.
7. The method according to claim 6, characterized in that, The deletion of a portion of the frame data in the sliding window includes: The odometer frames that are not keyframes and are not GPS constraint frames in the sliding window are identified as frame data to be deleted. Delete the frame data to be deleted in the sliding window.
8. The method according to claim 1, characterized in that, Before receiving the target global path sent by the third-party map service, the method further includes: Receive the initial global path issued by the third-party map service; Obtain the robot's current GPS location; Calculate the offset based on the current GPS location and the first point in the initial global path; Based on the offset, each path point in the initial global path is offset to obtain the offset path points; The coordinates of each offset path point are transformed into the ENU coordinate system to obtain the target global path.
9. A control device, characterized in that, include: The device includes a processor and a memory, the memory storing machine-readable instructions executable by the processor, which, when the control device is in operation, are executed by the processor to perform the steps of the robot mapless localization and navigation method as described in any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the robot mapless localization and navigation method as described in any one of claims 1 to 8.
