Robot initial positioning-free navigation method based on pose hypothesis motion disambiguation

Abstract

The invention discloses a robot initial positioning-free navigation method based on pose hypothesis motion disambiguation, and the method comprises the steps: generating an initial pose hypothesis set based on laser radar scanning data and a prior map; selecting an initial reference pose, and planning a broken line path comprising a plurality of end points on the priori map by taking the initial reference pose as a starting point and the target pose as an end point; controlling the robot to move along the broken line path, and performing safety inspection in the process of moving towards the next end point: if the inspection is not passed, re-planning the broken line path based on the initial reference pose; and if the inspection is passed and the robot arrives at the next end point, static observation is executed at the end point, and the pose hypothesis set is updated based on an observation result until the robot arrives at the target pose, and motion disambiguation is completed. According to the method, the robot can safely reach a given target pose after being started at any unknown position, and progressive pruning of initial passive multiple hypotheses is completed in the reaching process.

Description

Technical Field

[0001] This application relates to the field of autonomous localization and navigation of mobile robots, and more specifically, to a method for robot navigation without initial localization based on pose assumption motion disambiguation. Background Technology

[0002] When mobile robots perform navigation tasks in structured indoor environments (building inspection, warehousing and logistics, rail transit station halls, etc.), they typically rely on laser SLAM algorithms to achieve localization on a priori map M. Generally, laser SLAM needs to be started at the robot's mapping origin or manually given an initial pose prior in order to enter continuous pose tracking. However, in engineering practice, robots often encounter scenarios such as "powering on at any location," "recovering from lost localization," and "re-powering on after being moved or dragged." In these situations, the robot's initial pose is unknown, requiring global relocalization before or during operation to achieve initial-localization-free navigation that allows the robot to continue working without remotely returning to the origin or manually giving an initial pose. Existing similar technologies can be broadly classified into three categories, each with its own shortcomings: (1) Pure passive global relocalization: When the robot is stationary, the single-frame laser scan Z is directly used to perform global matching with the prior map M to output the optimal pose estimate to initialize the robot SLAM system so that the robot can continue the navigation task. However, when the laser radar FoV is small and there are repetitive structures in the environment (symmetrical rooms, mirrored corridors, repetitive compartments), multiple different poses can produce approximately consistent observations (i.e., generate perceptual aliases), making it difficult to distinguish or even completely consistent multiple candidate poses in terms of confidence during the passive relocalization process. It is impossible to directly determine the unique initial pose, which in turn leads to relocalization errors and navigation task failure. (2) Pure active relocation / exploratory relocation: When the initial pose is unknown, the robot expands the observation coverage through exploratory motion, and gradually converges to the correct pose by combining AMCL particle filtering or SLAM loop closure detection or global matching. However, the convergence time of this method is uncontrollable. Under the small FoV configuration of LiDAR, the robot's perception of obstacles in front is limited. If the curvature of the motion trajectory is large or the path shape is complex, there is a higher risk of collision and positioning drift. Moreover, in the pose convergence process, it often relies on odometry diffusion or kinematic model, which further introduces uncertainty. At the same time, the motion of this type of relocation method does not attempt to move towards the target pose. Its exploratory motion requires additional time, which further reduces efficiency. (3) Recovery relying on external beacons or historical trajectory logs: such as UWB / visual markers, or recording pose logs before shutdown for restart initialization. However, if external beacons are used, the environmental modification cost is high and the maintenance is complex. If log recovery is used, it cannot cover the scenario of "the robot being moved to an arbitrary unknown location and then powered on". Summary of the Invention

[0003] To address the aforementioned issues, this application provides a robot navigation method without initial localization based on pose hypothesis motion disambiguation. The aim is to achieve motion disambiguation of the initial pose hypothesis through a target-driven safe navigation process without relying on the initial pose or odometry kinematics diffusion, ultimately obtaining a unique global pose that can be used for navigation and task execution.

[0004] The first aspect of this invention provides a robot navigation method without initial localization based on pose assumption motion disambiguation, comprising: When the robot's initial pose is unknown, an initial pose hypothesis set is generated based on LiDAR scanning data and prior maps; An initial reference pose is selected from the initial pose hypothesis set, and a polyline path with multiple endpoints is planned on the prior map, starting from the initial reference pose and ending at the target pose. Control the robot to move along the zigzag path, and perform a safety check during the movement to the next endpoint: If the test fails, the polyline path is replanned based on the initial reference pose; If the test passes and the robot reaches the next endpoint, static observation is performed at that endpoint, and the pose hypothesis set is updated based on the observation results until the robot reaches the target pose, thus completing motion disambiguation.

[0005] In one optional implementation, selecting an initial reference pose from the initial pose hypothesis set specifically includes: Obtain the initial confidence level of each pose hypothesis in the initial pose hypothesis set; Calculate the distance representation from each pose assumption to the target pose, wherein the distance representation is Euclidean distance or path length calculated based on the prior map; The initial reference pose is selected based on the initial confidence level and distance representation, and the formula is as follows:

[0006] in, As the initial reference pose, To select a hypothesis with higher confidence, To select the weights of hypotheses that are farther from the target pose, As the initial confidence level, This is a distance representation.

[0007] In one alternative implementation, when planning the polyline path, the endpoint sequence is determined with the premise of ensuring the robot passes the safety detection and the objective of maximizing the scoring function:

[0008] in, For the scoring function, Where N represents the amount of observable map information at the endpoints, and N is the number of line segments in the polyline path. For information content weight, The endpoint number weight.

[0009] In one optional implementation, the safety check during the movement to the next endpoint specifically includes: Let the line segment from the robot's current position to the next endpoint be:

[0010] in, This is the geometric description of the current line segment. This is the robot's current position. The position of the next endpoint; Determine if there is an obstacle whose distance to the line segment is less than the safe radius:

[0011] in, The distance from the line segment to the nearest obstacle on the map. The safe radius for the robot; If so, the test is deemed unsuccessful, and the polyline path is replanned based on the initial reference pose; If the replanning is successful, control the robot to continue moving along the newly planned path; If replanning fails, the current reference pose is determined to be infeasible, and the pose hypothesis set is updated based on the current lidar scanning data. The current reference pose is then reselected to plan a new polyline path.

[0012] In one optional implementation, updating the pose hypothesis set based on the observation results specifically includes: For each pose hypothesis in the set of pose hypotheses, its predicted pose is calculated through pose composition operations based on the relative pose increment from the initial reference pose to the current endpoint:

[0013] in, for Compound operations Let i be the i-th pose hypothesis in the pose hypothesis set. From the reference starting point to the endpoint The relative pose increment; Acquire lidar scan data at the current endpoint and calculate the consistency score between this data and the prior map observations under the predicted pose:

[0014] in, In position The score for matching the lidar scan data with the prior map. For the current scan data, For the robot's pose, A priori map; Based on the observation consistency score, the global confidence of the pose hypothesis is updated using the fusion operator:

[0015] in, Assuming h j After accumulating the global confidence level of the j-th observation, for h j After accumulating the global confidence level observed for the (j-1)th time, For fusion operators.

[0016] In an optional implementation, the method further includes: after updating the global confidence of the pose hypothesis, if the global confidence of the pose hypothesis is lower than a preset removal threshold, then removing it from the pose hypothesis set.

[0017] In an optional implementation, the method further includes: after updating the global confidence of the pose hypothesis, if there is another pose hypothesis other than the current initial reference pose whose global confidence exceeds the global confidence of the current initial reference pose, then the initial reference pose is updated to the pose hypothesis with the highest global confidence, and the subsequent polyline path is replanned based on the updated reference pose.

[0018] A second aspect of this invention provides a robot navigation device for initial localization without interference based on pose assumption motion disambiguation, the device comprising: The pose hypothesis initialization module is used to generate an initial pose hypothesis set based on LiDAR scanning data and prior maps when the robot's initial pose is unknown. The path planning module is used to select an initial reference pose from the initial pose hypothesis set, and plan a polyline path with multiple endpoints on the prior map, starting from the initial reference pose and ending at the target pose; and The motion verification module is used to control the robot to move along the polygonal path and to perform safety checks during the movement to the next endpoint. If the test fails, the polyline path is replanned based on the initial reference pose; If the test passes and the robot reaches the next endpoint, static observation is performed at that endpoint, and the pose hypothesis set is updated based on the observation results until the robot reaches the target pose, thus completing motion disambiguation.

[0019] A third aspect of the present invention provides an electronic device, characterized in that it includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements a robot navigation method without initial localization based on pose hypothesis motion disambiguation.

[0020] A fourth aspect of this invention provides a computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, a robot initial localization navigation method based on pose assumption motion disambiguation is provided.

[0021] This application has at least the following advantages or beneficial effects: (1) This invention generates a limited number of pose assumptions through passive relocation. In the subsequent target navigation process, it innovatively introduces planning control feasibility constraints and multiple discrete observation constraints to progressively disambiguate and prune the pose assumptions. Finally, when the robot arrives at the target pose, a unique global pose is accurately determined, enabling the robot to directly execute the target navigation task after being powered on at any unknown location, truly achieving navigation without initial positioning. This capability greatly expands the applicable boundaries of the robot system in actual engineering scenarios, especially adapting to rapid task recovery scenarios after the robot is moved, restarted, or loses its positioning, significantly improving the system's scenario adaptability and practicality.

[0022] (2) This invention fundamentally solves the technical pain point of small field-of-view lidar in environments with repetitive or mirror structures where passive relocation is prone to multiple solutions and difficult to accurately distinguish. By deeply embedding the relocation disambiguation process into the target-driven navigation motion, and with the help of the planning and control feasibility constraints introduced by the real motion process, as well as the consistency constraints of multiple discrete static observations, the pose assumptions that were originally difficult to distinguish at the local observation level gradually show essential differences at the global motion scale, thereby achieving efficient and thorough elimination of perception aliases and ensuring the accuracy of positioning.

[0023] (3) This invention uses a broken line path composed of straight line segments as the navigation path, and performs static observations at the path endpoints by stopping and orienting, which significantly improves the stability and discriminative power of laser observation under small field-of-view conditions. Compared with active relocation methods that rely on continuous curve motion or random exploration, this invention has higher safety and controllability in the process of parallel navigation and disambiguation, effectively reducing the risk of positioning divergence and collision caused by insufficient observation or complex motion trajectory, and improving the reliability of system operation.

[0024] (4) This invention achieves an integrated design of relocalization and navigation processes. Relocalization no longer exists as a separate preprocessing stage or an additional exploration stage, but is naturally integrated into the navigation process of the robot reaching the target pose. This design allows the system to obtain reliable global pose estimation while completing the predetermined navigation task without the need for additional relocalization convergence criteria or exploration strategies. On the basis of ensuring localization robustness, it greatly improves the overall system's operating efficiency and engineering practical value. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of a robot navigation method without initial localization based on pose assumption motion disambiguation proposed in an embodiment of this application; Figure 2 This is a structural diagram of a robot navigation device without initial localization based on pose assumption motion disambiguation proposed in an embodiment of this application; Figure 3 This is a schematic diagram of an electronic device according to this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] Please refer to Figure 1 , Figure 1 This is a flowchart of a robot navigation method without initial localization based on pose assumption motion disambiguation, proposed in an embodiment of this application. Figure 1 As shown, a robot navigation method without initial localization based on pose hypothesis motion disambiguation includes: S100: When the robot's initial pose is unknown, generate a set of initial pose assumptions based on LiDAR scanning data and prior maps; In this embodiment, after the robot is powered on with an unknown initial pose, it acquires at least one frame of LiDAR scanning data and, combined with a prior map, generates an initial pose hypothesis set using a passive global relocalization method. The passive global relocalization method can be any of the following: (1) Sparse pose sampling is performed in the robot's reachable space, and a limited number of high-confidence pose hypotheses are selected through a multi-stage matching and scoring mechanism; (2) Passive sampling of poses is performed within the global map area, and the rationality of each sampled pose is quickly evaluated by laser scanning and map matching. Through the above passive relocalization method, an initial set of pose assumptions is obtained:

[0029] Each pose assumption From two-dimensional position and orientation angle Composition, and This is a configurable finite integer. The passive phase simultaneously outputs the initial confidence level for each hypothesis. .

[0030] S200: Select an initial reference pose from the initial pose hypothesis set, and plan a polyline path with multiple endpoints on the prior map, starting from the initial reference pose and ending at the target pose. Specifically, selecting an initial reference pose from the initial pose hypothesis set includes: S210: Obtain the initial confidence level of each pose hypothesis in the initial pose hypothesis set; S220: Calculate the distance representation from each pose assumption to the target pose, wherein the distance representation is Euclidean distance or path length calculated based on the prior map; In this embodiment, the target pose is set as follows: For the hypothesis set Each hypothesis Calculate the Euclidean distance or reachable path length from it to the target. This is used to characterize the approximate running distance from this assumption to the target pose.

[0031] S230: Based on the initial confidence level and distance representation, an initial reference pose is selected, and the formula is as follows:

[0032] in, As the initial reference pose, To select a hypothesis with higher confidence, To select the weights of hypotheses that are farther from the target pose, As the initial confidence level, This is a distance representation.

[0033] In this embodiment, the formula can be used to comprehensively weigh confidence level and disambiguation potential to select the optimal initial reference pose.

[0034] S300: Control the robot to move along the zigzag path and perform a safety check during the movement to the next endpoint: In this embodiment, on the map The above planning starts from arrive global path ,in , This path can be obtained by algorithms including but not limited to A* and Dijkstra.

[0035] Select the endpoint index set from the discrete path. Obtain the endpoints of the polyline path. And define the direction of the line segment before the endpoint number as the previous line segment:

[0036] Therefore, the broken line path can be represented as:

[0037] in .

[0038] set up In position Above, the fan-shaped field of view formed by the robot's LiDAR encompasses the area on the prior map. The amount of information that can be covered here, including but not limited to, the number of occupies in the state, then the total score of the polyline path is defined as:

[0039] in, For the scoring function, Where N represents the amount of observable map information at the endpoints, and N is the number of line segments in the polyline path. For information content weight, The endpoint number weight.

[0040] The premise is that the constraints are satisfied:

[0041] Therefore, the selection of endpoints for the polyline path is based on the principle of balancing higher information gain and fewer endpoints while satisfying robot drivability constraints, thereby balancing the potential for relocation disambiguation and navigation efficiency.

[0042] Specifically, the safety check during the movement to the next endpoint includes: Let the line segment from the robot's current position to the next endpoint be [ , Considering the robot's current position Tracking sub-segments in the short term [ , Its parameterized expression is:

[0043] in, This is the geometric description of the current line segment. This is the robot's current position. The position of the next endpoint; Determine if there is an obstacle whose distance to the line segment is less than the safe radius:

[0044] in, The distance from the line segment to the nearest obstacle on the map. The safe radius for the robot; If so, the test is deemed unsuccessful, and the polyline path is replanned based on the initial reference pose; If the replanning is successful, control the robot to continue moving along the newly planned path; If replanning fails, the current reference pose is determined to be infeasible, and the pose hypothesis set is updated based on the current lidar scanning data. The current reference pose is then reselected to plan a new polyline path.

[0045] S310: If the test fails, the polyline path is replanned based on the initial reference pose; In this embodiment, if the safety check fails, i.e., the distance from an obstacle to a future sub-segment is less than the safe radius, then the segment is considered unsafe to execute at the current moment, triggering a replanning process. A new polyline path from the current position to the target is recalculated under the current reference assumptions. If a polyline path satisfying the constraints cannot be obtained under given planning constraints (such as maximum planning time or maximum number of replanning attempts), then the current reference assumption is considered inconsistent with the geometric constraints of the real environment. This reference assumption is then discarded, and the reference pose is switched for replanning.

[0046] S320: If the test passes and the robot reaches the next endpoint, then perform static observation at that endpoint and update the pose hypothesis set based on the observation results until the robot reaches the target pose and completes motion disambiguation.

[0047] In this embodiment, if the safety check is passed and the robot reaches the next endpoint... The robot then stops moving at that endpoint and adjusts to face the direction. Then, collect endpoint scan data. Then, based on the observation, the pose hypothesis set is updated, and this process is repeated until the robot reaches the target pose. This completes motion disambiguation, causing the hypothesis set to eventually converge to a unique reasonable pose.

[0048] Specifically, updating the pose hypothesis set based on the observation results includes: S321: For each pose hypothesis in the set of pose hypotheses, calculate its predicted pose through pose composition operation based on the relative pose increment from the initial reference pose to the current endpoint:

[0049] in, for Compound operations Let i be the i-th pose hypothesis in the pose hypothesis set. From the reference starting point to the endpoint The relative pose increment; In this embodiment, the pose hypothesis set is addressed. Each pose assumption in , Based on the position from the initial reference pose to the current endpoint relative pose increment (Determined by the geometry of the polyline path), through the SE (2) composite operation. Calculate its predicted pose The predicted pose characterizes the pose assumptions under the current reference assumptions. The corresponding robot is at the endpoint The theoretical pose of the location.

[0050] S322: Acquire lidar scan data at the current endpoint and calculate the consistency score between the data and the prior map observations under the predicted pose:

[0051] in, In position The score for matching the lidar scan data with the prior map. For the current scan data, For the robot's pose, A priori map; In this embodiment, the lidar scanning data at the current endpoint is acquired. Then, the data is calculated in the predicted pose. Below and prior map Observational consistency score ,in For scanning - ground Figure 1 The consistency scoring function, whose form is not limited, may include scores based on average distance features, scores based on likelihood fields, and other scan matching scores. This score is used to reflect the pose assumption. The degree of matching with actual observations.

[0052] S323: Based on the observation consistency score, update the global confidence of the pose hypothesis using the fusion operator:

[0053] in, Assuming h j After accumulating the global confidence level of the j-th observation, for h j After accumulating the global confidence level observed for the (j-1)th time, For fusion operators.

[0054] In this embodiment, the observation consistency score is used as the basis. Update pose assumption When calculating the global confidence level, a fusion operator is used. The recursive update is performed, and the form of the fusion operator F is not limited. It can include arithmetic average, weighted average based on endpoint information, exponential filtering, log-domain likelihood accumulation, etc.

[0055] Furthermore, it also includes: after updating the global confidence of the pose hypothesis, if the global confidence of the pose hypothesis is lower than a preset removal threshold, then it is removed from the pose hypothesis set.

[0056] In this embodiment, after updating the global confidence of the pose hypothesis, a preset rejection threshold is set. If a certain pose hypothesis is rejected... If the global confidence level is less than the removal threshold, the hypothesis is considered inconsistent with the actual environment or motion constraints, and it is removed from the pose hypothesis set. This involves deleting the hypothesis set to achieve progressive pruning.

[0057] Furthermore, it also includes: after updating the global confidence of the pose hypothesis, if there is another pose hypothesis other than the current initial reference pose whose global confidence exceeds the global confidence of the current initial reference pose, then the initial reference pose is updated to the pose hypothesis with the highest global confidence, and the subsequent polyline path is replanned based on the updated reference pose.

[0058] In this embodiment, after updating the global confidence of the pose hypothesis, the global confidence of all pose hypotheses is compared. If there is another pose hypothesis other than the current initial reference pose whose global confidence exceeds that of the current initial reference pose, the initial reference pose is updated to the pose hypothesis with the highest global confidence. Based on the updated reference pose, the subsequent polyline path from the current endpoint to the target pose is replanned to ensure the accuracy of the navigation and disambiguation process.

[0059] Please refer to Figure 2 , Figure 2 This is a structural diagram of a robot navigation device for motion disambiguation based on pose assumptions, proposed in one embodiment of this application, that eliminates the need for initial localization. Figure 2 As shown in the figure, this disclosure also provides a robot navigation device without initial localization based on pose hypothesis motion disambiguation. The device includes: a pose hypothesis initialization module 201, a path planning module 202, and a motion verification module 203; wherein, The pose hypothesis initialization module 201 is used to generate an initial pose hypothesis set based on LiDAR scanning data and prior map when the robot's initial pose is unknown. Path planning module 202 is used to select an initial reference pose from the initial pose hypothesis set, and plan a polyline path with multiple endpoints on the prior map, starting from the initial reference pose and ending at the target pose; and The motion verification module 203 is used to control the robot to move along the polygonal path and to perform a safety check during the movement to the next endpoint. If the test fails, the polyline path is replanned based on the initial reference pose; If the test passes and the robot reaches the next endpoint, static observation is performed at that endpoint, and the pose hypothesis set is updated based on the observation results until the robot reaches the target pose, thus completing motion disambiguation.

[0060] This disclosure also provides an electronic device, please refer to... Figure 3 , Figure 3 This is a schematic diagram of an electronic device illustrated in an embodiment of this disclosure. For example... Figure 3As shown, the electronic device 100 includes a memory 110 and a processor 120. The memory 110 and the processor 120 are connected via a bus for communication. The memory 110 stores a computer program that can run on the processor 120 to implement the steps in the robot initial positioning navigation method based on pose assumption motion disambiguation disclosed in this embodiment.

[0061] Example 1: A wheeled differential robot is used, equipped with a 220° field-of-view LiDAR; the initial pose assumption set size is... Initial reference pose selection weight parameters , The minimum obstacle distance requirement follows the preset standard of the implementation example.

[0062] The above experiments show that the robot can start at any point on a map containing the same rooms, without needing to provide an initial pose. It can achieve convergence of localization estimation and reach the target pose using the method of this application, thus improving the long-term autonomy of the robot's operation.

[0063] Example 2: A quadruped robot equipped with a 180° field-of-view lidar (smaller than the field of view in Example 1); initial pose assumptions set size. Initial reference pose selection weight parameters , The minimum obstacle distance requirement is the same as in Example 1.

[0064] The above experiments show that the robot dog can start at any point on an industrial site map with a large number of similar features, or after a positioning failure, and can achieve convergence of positioning estimation and reach the target pose through the method of this application, without the need for on-site personnel to manually take over the robot.

[0065] The disclosed embodiments also provide a computer-readable storage medium that, when executed by a processor of a computer device, enables the computer device to perform steps in the robot initial localization navigation method based on pose assumption motion disambiguation according to the embodiments of this disclosure.

[0066] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, electronic devices, and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0067] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0068] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0069] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0070] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0071] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0072] The above provides a detailed description of a robot navigation method without initial localization based on pose assumption motion disambiguation provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A robot navigation method without initial localization based on pose assumption motion disambiguation, characterized in that, include: When the robot's initial pose is unknown, an initial pose hypothesis set is generated based on LiDAR scanning data and prior maps; An initial reference pose is selected from the initial pose hypothesis set, and a polyline path with multiple endpoints is planned on the prior map, starting from the initial reference pose and ending at the target pose. Control the robot to move along the zigzag path, and perform a safety check during the movement to the next endpoint: If the test fails, the polyline path is replanned based on the initial reference pose; If the test passes and the robot reaches the next endpoint, static observation is performed at that endpoint, and the pose hypothesis set is updated based on the observation results until the robot reaches the target pose, thus completing motion disambiguation.

2. The robot navigation method without initial localization based on pose assumption motion disambiguation according to claim 1, characterized in that, Selecting an initial reference pose from the initial pose hypothesis set specifically includes: Obtain the initial confidence level of each pose hypothesis in the initial pose hypothesis set; Calculate the distance representation from each pose assumption to the target pose, wherein the distance representation is Euclidean distance or path length calculated based on the prior map; The initial reference pose is selected based on the initial confidence level and distance representation, and the formula is as follows: in, As the initial reference pose, To select a hypothesis with higher confidence, To select the weights of hypotheses that are farther from the target pose, As the initial confidence level, This is a distance representation.

3. The robot navigation method without initial localization based on pose assumption motion disambiguation according to claim 1, characterized in that, When planning the polyline path, the endpoint sequence is determined with the premise of ensuring the robot passes the safety detection and the objective of maximizing the scoring function: in, For the scoring function, Where N represents the amount of observable map information at the endpoints, and N is the number of line segments in the polyline path. For information content weight, The endpoint number weight.

4. The robot navigation method without initial localization based on pose assumption motion disambiguation according to claim 1, characterized in that, The safety check during the movement to the next endpoint specifically includes: Let the line segment from the robot's current position to the next endpoint be: in, This is the geometric description of the current line segment. This is the robot's current position. The position of the next endpoint; Determine if there is an obstacle whose distance to the line segment is less than the safe radius: in, The distance from the line segment to the nearest obstacle on the map. The safe radius for the robot; If so, the test is deemed unsuccessful, and the polyline path is replanned based on the initial reference pose; If the replanning is successful, control the robot to continue moving along the newly planned path; If replanning fails, the current reference pose is determined to be infeasible, and the pose hypothesis set is updated based on the current lidar scanning data. The current reference pose is then reselected to plan a new polyline path.

5. The robot navigation method without initial localization based on pose assumption motion disambiguation according to claim 1, characterized in that, The process of updating the pose hypothesis set based on observation results specifically includes: For each pose hypothesis in the set of pose hypotheses, its predicted pose is calculated through pose composition operations based on the relative pose increment from the initial reference pose to the current endpoint: in, for Compound operations Let i be the i-th pose hypothesis in the pose hypothesis set. From the reference starting point to the endpoint The relative pose increment; Acquire lidar scan data at the current endpoint and calculate the consistency score between this data and the prior map observations under the predicted pose: in, In position The score for matching the lidar scan data with the prior map. For the current scan data, For the robot's pose, A priori map; Based on the observation consistency score, the global confidence of the pose hypothesis is updated using the fusion operator: in, Assuming h j After accumulating the global confidence level of the j-th observation, for h j After accumulating the global confidence level observed for the (j-1)th time, For fusion operators.

6. The robot navigation method without initial localization based on pose assumption motion disambiguation according to claim 5, characterized in that, Also includes: After updating the global confidence of the pose hypothesis, if the global confidence of the pose hypothesis is lower than the preset removal threshold, it is removed from the pose hypothesis set.

7. The robot navigation method without initial localization based on pose assumption motion disambiguation according to claim 5, characterized in that, Also includes: After updating the global confidence of the pose hypothesis, if there is another pose hypothesis other than the current initial reference pose whose global confidence exceeds that of the current initial reference pose, then the initial reference pose is updated to the pose hypothesis with the highest global confidence, and the subsequent polyline path is replanned based on the updated reference pose.

8. The robot navigation device for initial localization without initial positioning based on pose assumption motion disambiguation according to any one of claims 1-7, characterized in that, The device includes: The pose hypothesis initialization module is used to generate an initial pose hypothesis set based on LiDAR scanning data and prior maps when the robot's initial pose is unknown. The path planning module is used to select an initial reference pose from the initial pose hypothesis set, and plan a polyline path with multiple endpoints on the prior map, starting from the initial reference pose and ending at the target pose; and The motion verification module is used to control the robot to move along the polygonal path and to perform safety checks during the movement to the next endpoint. If the test fails, the polyline path is replanned based on the initial reference pose; If the test passes and the robot reaches the next endpoint, static observation is performed at that endpoint, and the pose hypothesis set is updated based on the observation results until the robot reaches the target pose, thus completing motion disambiguation.

9. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the robot initial localization navigation method based on pose assumption motion disambiguation as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the robot initial localization navigation method based on pose hypothesis motion disambiguation as described in any one of claims 1 to 7.

Images