Robot autonomous mapping method and system in restricted space environment and robot

Abstract

The invention provides an autonomous mapping method and system for a robot in a limited space environment and the robot, and relates to the technical field of robot control, and the method comprises the steps: carrying out the front-end pose estimation and rear-end map optimization of multi-modal sensing data through an instant positioning and mapping module, and generating and updating a two-dimensional occupied grid map; based on a map and a robot pose, executing dynamic partitioning, identifying an unexplored boundary region, and determining a staged navigation target point sequence in combination with a multi-target optimization strategy; and generating a collision-free path meeting kinematics constraints by utilizing an environment perception self-adaptive local path planning algorithm, and taking the collision-free path as an advancing path for active mapping of the robot. According to the method, the technical problems of low mapping precision, low exploration efficiency and poor stability in a traditional method are solved, high-precision and high-efficiency real-time mapping and autonomous navigation are realized, and the mapping performance and the operation reliability of the robot in a complex and limited environment are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of robot control, and more specifically, to a method, system, and robot for autonomous mapping in a confined space environment. Background Technology

[0002] With the continuous development of intelligent robot technology, autonomous mobile robots are increasingly being used in confined space environments such as industrial inspection, emergency rescue, and facility maintenance. In these typical application scenarios, robots often need to complete autonomous exploration and map building tasks in environments with complex structures, narrow passages, unstable lighting conditions, and unknown prior information. To achieve effective environmental perception and navigation decision-making, Simultaneous Localization and Mapping (SLAM) technology, as a core supporting module, is widely integrated into various service and special-purpose robot systems.

[0003] However, in practical engineering applications, traditional mapping and exploration strategies often exhibit shortcomings such as insufficient mapping accuracy, severe pose estimation drift, and high redundancy in exploration paths when faced with narrow corridors, multi-branch passages, or areas with dense obstacles, due to limitations in sensor observation range, significant repetitiveness of environmental structures, and complex dynamic interference factors. Furthermore, existing SLAM systems still have limitations in terms of cumulative error control and closed-loop detection reliability during long-term operation, especially in scenarios with sparse features or strong dynamic interference, where map distortion or localization issues are prone to occur.

[0004] In summary, existing technologies still face technical challenges in complex and constrained environments, such as low accuracy in environmental modeling, poor pose stability, and limited ability to optimize exploration paths, making it difficult to meet the requirements for highly reliable and efficient autonomous mapping. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, and robot for autonomous mapping in confined space environments, so as to alleviate the technical problems of low accuracy, low efficiency, and poor stability in the existing technology of autonomous mapping of robots in confined spaces.

[0006] In a first aspect, embodiments of the present invention provide a method for autonomous mapping of a robot in a confined space environment. The method includes: inputting multimodal perception data collected in real-time by sensors mounted on the robot into an instant localization and mapping module for front-end pose estimation and back-end map optimization, generating and continuously updating a two-dimensional occupied grid map; performing dynamic partitioning processing on the map based on the two-dimensional occupied grid map and the robot's current pose information, identifying unexplored boundary areas, and determining a sequence of phased navigation target points according to a multi-objective optimization strategy; and generating a collision-free path that satisfies the robot's kinematic constraints using a local path planning algorithm based on improved environmental perception, based on the sequence of phased navigation target points and real-time sensing information; the collision-free path is used to represent the robot's travel route when performing an active mapping task.

[0007] In some optional implementations, the aforementioned front-end pose estimation includes: performing correlation scan matching between the current frame laser scan data and the local sub-map being constructed; and selecting high-confidence regions based on the matching confidence for secondary optimization alignment.

[0008] In some optional implementations, the above-mentioned selection of high-confidence regions based on matching confidence for secondary optimization alignment includes: calculating the confidence score of each point cloud matching result during the relevant scanning matching process; setting a threshold based on the confidence score, and extracting point sets higher than the threshold to form high-confidence regions; and re-performing the local optimal transformation solution based on the point cloud features in the high-confidence regions to correct the pose estimation value of the current frame.

[0009] In some optional implementations, the aforementioned backend map optimization includes: matching the current local subgraph with historical subgraphs to obtain pose constraints; and optimizing the pose graph based on the aforementioned pose constraints to correct accumulated errors.

[0010] In some optional implementations, unexplored boundary areas are identified, and a sequence of phased navigation target points is determined based on a multi-objective optimization strategy. This includes: identifying the set of boundary points between known and unknown areas in the aforementioned two-dimensional occupied grid map; selecting target sub-regions from the dynamically partitioned sub-regions that contain the aforementioned boundary points and whose internal unknown state grid ratio exceeds a first preset threshold as candidate exploration areas; prioritizing the aforementioned candidate exploration areas based on a preset strategy to generate a global exploration route; and sequentially extracting phased navigation target points from the aforementioned global exploration route.

[0011] In some optional implementations, the cost function of the aforementioned local path planning algorithm based on improved environmental perception includes weight coefficients, which include speed term weights and obstacle avoidance term weights. The adjustment method of the weight coefficients includes: real-time detection of environmental complexity in the local map to determine the trend of environmental state changes; when the environmental state is detected to change from a simple state to a complex state, the speed term weight is reduced while the obstacle avoidance term weight is increased; when the environmental state is detected to change from a complex state to a simple state, the speed term weight is increased while the obstacle avoidance term weight is decreased.

[0012] In some optional implementations, the environmental complexity in the local map is detected in real time, including: calculating an environmental complexity assessment value based on the density of obstacle points and / or the number of boundary points per unit area in the local map; determining an environmental complexity baseline, and determining a complex state when the assessment value rises above the upper limit of the baseline, and a simple state when the assessment value falls below the lower limit of the baseline; and determining whether the environmental state has switched based on the difference in environmental state between time periods.

[0013] Secondly, embodiments of the present invention provide a robot autonomous mapping system in a confined space environment. The system includes: a sensor component for real-time acquisition of multimodal perception data of the robot's surrounding environment; an instant localization and mapping (IMR) module, communicatively connected to the sensor component, configured to receive the multimodal perception data, perform front-end pose estimation and back-end map optimization, and generate and continuously update a two-dimensional occupancy grid map; an exploration decision module, communicatively connected to the IMR module, configured to perform dynamic partitioning processing on the map based on the two-dimensional occupancy grid map and the robot's current pose information, identify unexplored boundary areas, and determine a phased navigation target point sequence according to a multi-objective optimization strategy; a local path planning module, communicatively connected to the exploration decision module and the sensor component, configured to generate a collision-free path that satisfies robot kinematic constraints using an improved local path planning algorithm based on environmental perception, according to the phased navigation target point sequence and real-time sensing information; and a path execution interface module, configured to output the collision-free path to the robot motion control system as a travel route instruction for the robot during the active mapping task.

[0014] Thirdly, embodiments of the present invention provide a robot, including: a robot motion control system and the robot autonomous mapping system in a confined space environment as described in the second aspect.

[0015] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to perform the method described in any of the first aspects above.

[0016] This invention provides a method, system, and robot for autonomous mapping in confined spaces. The method utilizes a real-time localization and map building module to perform front-end pose estimation and back-end map optimization on multimodal perception data, generating and updating a two-dimensional occupancy grid map. Based on the map and robot pose, dynamic partitioning is performed to identify unexplored boundary areas, and a multi-objective optimization strategy is combined to determine a sequence of staged navigation target points. An environment-aware adaptive local path planning algorithm is used to generate collision-free paths that satisfy kinematic constraints, serving as the robot's active mapping route. This invention solves the technical problems of low mapping accuracy, low exploration efficiency, and poor stability in traditional methods, achieving high-precision, high-efficiency real-time mapping and autonomous navigation, significantly improving the robot's mapping performance and operational reliability in complex and confined environments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a method for autonomous mapping of a robot in a confined space environment, as provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of a robot autonomous mapping system in a confined space environment, provided by an embodiment of the present invention. Figure 3 This is a schematic diagram of a preferred robot autonomous mapping system in a confined space environment, provided by an embodiment of the present invention. Figure 4 This is a flowchart illustrating a preferred method for autonomous mapping of robots in a confined space environment, provided by an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] In existing technologies, most robots employ a sensor fusion-based SLAM framework for environmental mapping, combined with a hierarchical planning architecture for autonomous exploration. However, in practical deployment, traditional mapping and exploration strategies often exhibit shortcomings such as insufficient mapping accuracy, severe pose estimation drift, and high redundancy in exploration paths when faced with narrow corridors, multi-branch passages, or areas with dense obstacles, due to limitations in sensor observation range, high repetitiveness of environmental structures, and numerous dynamic interference factors. Furthermore, existing SLAM systems still have limitations in handling accumulated errors and loop closure detection reliability during long-term operation, especially in scenarios with sparse features or strong dynamic interference, where map distortion or localization failures are prone to occur.

[0021] In summary, existing technologies still face challenges in terms of environmental modeling accuracy, pose stability, and exploration path optimization capabilities in complex and confined spaces, making it difficult to meet the requirements for highly reliable and efficient autonomous mapping.

[0022] Based on this, the present invention provides a method, system and robot for autonomous mapping in a confined space environment, to solve the technical problems of low mapping accuracy, low exploration efficiency and poor stability of robots in complex and confined environments in traditional methods.

[0023] To facilitate understanding of this embodiment, a detailed description of a robot autonomous mapping method in a confined space environment disclosed in this embodiment of the invention will be provided first. (See [link to relevant documentation]). Figure 1 The diagram shows a flowchart of a robot autonomous mapping method in a confined space environment. The method mainly includes the following steps S102 to S106: In step S102, the multimodal perception data collected in real time by the sensors on the robot is input into the instant localization and map building module for front-end pose estimation and back-end map optimization, generating and continuously updating a two-dimensional occupancy grid map.

[0024] In one embodiment, the aforementioned front-end pose estimation may include: performing correlation scanning matching between the current frame laser scanning data and the local sub-map being constructed; and selecting high-confidence regions based on the matching confidence for secondary optimization alignment.

[0025] Preferably, the above-mentioned selection of high-confidence regions based on matching confidence for secondary optimization alignment may include: calculating the confidence score of each point cloud matching result during the relevant scanning matching process; setting a threshold based on the confidence score, extracting point sets higher than the threshold to form high-confidence regions; and re-performing the local optimal transformation solution based on the point cloud features in the high-confidence regions to correct the pose estimation value of the current frame.

[0026] In one embodiment, the aforementioned backend map optimization may include: matching the current local subgraph with historical subgraphs to obtain pose constraints; and optimizing the pose graph based on the pose constraints to correct accumulated errors.

[0027] Preferably, matching the current local sub-image with historical sub-images to obtain pose constraints may include: fusing multiple consecutive frames of laser scanning data to generate the current local sub-image, the local sub-image containing a structured point cloud distribution; performing geometric consistency matching between the current local sub-image and the stored historical sub-images one by one; when a match is successful, generating the corresponding relative pose observation value as a closed-loop constraint input to the pose image optimization module.

[0028] Step S104: Based on the two-dimensional occupancy grid map and the robot's current pose information, perform dynamic partitioning processing on the map, identify unexplored boundary areas, and determine the sequence of phased navigation target points according to the multi-objective optimization strategy.

[0029] In one embodiment, the dynamic partitioning process of the map in step S104 above may include: dividing the global map into a uniform grid structure of M×N to generate multiple sub-regions; continuously monitoring the spatial size of each sub-region during the exploration process; and in response to any side length of any sub-region exceeding a preset spatial resolution threshold, increasing the number of partitions in the corresponding direction and recursively subdividing the sub-region to maintain the granularity accuracy of the environment representation.

[0030] In one embodiment, identifying unexplored boundary areas and determining a sequence of phased navigation target points based on a multi-objective optimization strategy may include: identifying the set of boundary points between known and unknown areas in the aforementioned two-dimensional occupied grid map; selecting target sub-regions from the dynamically partitioned sub-regions that contain the aforementioned boundary points and whose internal unknown state grid ratio exceeds a first preset threshold as candidate exploration areas; prioritizing the aforementioned candidate exploration areas based on a preset strategy to generate a global exploration route; and sequentially extracting phased navigation target points from the aforementioned global exploration route.

[0031] In another embodiment, the selection method for candidate exploration regions may further include: uniformly sampling or traversing and counting the grid cells in each sub-region; calculating the proportion of grid cells in an "unknown" state to the total number of grid cells; if the proportion is greater than or equal to a preset proportion threshold, and there is at least one boundary point in the sub-region, then it is included in the candidate exploration region set.

[0032] In the above embodiments, the preset strategy may include: the principle of starting from the nearest point, the principle of the shortest path, the principle of consistent direction, and the principle of continuous backtracking to avoid.

[0033] Specifically, the principle of proximity at the start is as follows: the starting point of the route must be set to the sub-region where the robot is currently located to minimize initial movement costs and ensure rapid response. The principle of shortest path is as follows: the order of region visits should minimize the total length of the robot's movement trajectory to effectively reduce task execution time and energy consumption. The principle of consistent direction is as follows: during planning, regions with smaller deviations from the robot's current heading angle are prioritized as the next target to reduce unnecessary turns and improve motion smoothness. The principle of continuous backtracking and avoidance is as follows: the newly generated global planned route should maintain a certain structural similarity to the previous historical route. By introducing path continuity constraints, inefficient backtracking behavior is avoided, ensuring a smooth transition in the exploration process.

[0034] Step S106: Based on the phased navigation target point sequence and real-time sensing information, a local path planning algorithm based on environmental perception is used to generate a collision-free path that satisfies the robot's kinematic constraints; the collision-free path is used to represent the robot's travel route when performing an active mapping task.

[0035] In one embodiment, the cost function of the aforementioned local path planning algorithm based on improved environmental perception includes weight coefficients, which include a speed term weight and an obstacle avoidance term weight. Preferably, the adjustment method of the weight coefficients may include: real-time detection of environmental complexity in the local map to determine the trend of environmental state changes; when the environmental state is detected to change from a simple state to a complex state, the speed term weight is reduced while the obstacle avoidance term weight is increased; when the environmental state is detected to change from a complex state to a simple state, the speed term weight is increased while the obstacle avoidance term weight is decreased.

[0036] In one embodiment, real-time detection of environmental complexity in a local map may include: calculating an environmental complexity assessment value based on the density of obstacle points and / or the number of boundary points per unit area in the local map; determining an environmental complexity baseline, classifying the assessment value as a complex state when it rises above the upper limit of the baseline, and classifying the assessment value as a simple state when it falls below the lower limit of the baseline; and determining whether the environmental state has switched based on the difference in environmental state between previous and subsequent times.

[0037] Furthermore, after step S106 above, a collision-free path can be input to the underlying motion controller. The underlying motion controller generates linear and angular velocity control commands for the robot based on the current pose and path deviation, and sends them to the drive system to drive the robot to move. During the robot's movement, new environmental data is continuously collected and fed back to the SLAM module to dynamically update the two-dimensional occupancy grid map, thereby completing the closed-loop operation from environmental perception to action execution and then to map evolution.

[0038] The underlying motion controller can employ proportional-derivative (PD) control, pure pursuit, model predictive control (MPC), or other trajectory tracking algorithms suitable for nonholonomic constrained robots. The controller receives a sequence of reference points on the local path and calculates the required linear velocity v and angular velocity ω based on the lateral error and heading error between the robot's current pose and the nearest reference point, and outputs them to the motor drive unit.

[0039] This invention provides a method for autonomous mapping of robots in confined spaces. It utilizes a real-time localization and map building module to perform front-end pose estimation and back-end map optimization on multimodal perception data, generating and updating a two-dimensional occupancy grid map. Based on the map and robot pose, dynamic partitioning is performed to identify unexplored boundary areas, and a multi-objective optimization strategy is combined to determine a sequence of staged navigation target points. An environment-aware adaptive local path planning algorithm is used to generate a collision-free path that satisfies kinematic constraints, serving as the robot's active mapping route. This invention solves the technical problems of low mapping accuracy, low exploration efficiency, and poor stability in traditional methods, achieving high-precision, high-efficiency real-time mapping and autonomous navigation, significantly improving the robot's mapping performance and operational reliability in complex and confined environments.

[0040] Based on the same inventive concept, this invention also provides a robot autonomous mapping system for confined space environments. See [link to relevant documentation]. Figure 2 As shown, the system mainly includes the following parts: The sensor assembly 210 mounted on the robot is used to collect multimodal perception data of the robot's surrounding environment in real time; The real-time positioning and mapping module 220 is communicatively connected to the aforementioned sensor component 210 and is configured to receive the aforementioned multimodal sensing data, perform front-end pose estimation and back-end map optimization, and generate and continuously update a two-dimensional occupied grid map. The exploration decision module 230 is communicatively connected to the aforementioned real-time localization and map building module 220. It is configured to perform dynamic partitioning processing on the map based on the aforementioned two-dimensional occupied grid map and the robot's current pose information, identify unexplored boundary areas, and determine a phased navigation target point sequence according to a multi-objective optimization strategy. The local path planning module 240 is communicatively connected to the exploration decision module 230 and the sensor component 210, and is configured to generate a collision-free path that satisfies the robot's kinematic constraints by using a local path planning algorithm based on environmental perception improvement, according to the above-mentioned phased navigation target point sequence and real-time sensing information. The path execution interface module 250 is configured to output the above-mentioned collision-free path to the robot motion control system as the travel route instruction of the robot in the process of performing the active mapping task.

[0041] The autonomous mapping system for robots in confined spaces provided in this invention embodiment can be specific hardware on a device or software or firmware installed on the device. The system provided in this invention embodiment has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment section can be referred to the corresponding content in the aforementioned method embodiment. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can all be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0042] To facilitate understanding, the methods described above will be explained in detail below using a preferred autonomous mapping system for robots in confined spaces as an example. The structure of this system can be found in [reference needed]. Figure 3 As shown.

[0043] The system provided in this embodiment can be used to perform the steps in the robot autonomous mapping method in the confined space environment described in the above embodiment. Specifically, the robot can continuously collect multimodal data of the surrounding environment through various onboard sensors and input it into the real-time localization and mapping (SLAM) module. In this module, the system can fuse information from LiDAR, vision, etc., to achieve real-time modeling of the robot's environment and finally output a dynamically updated two-dimensional occupancy grid map that clearly marks the feasible area and the distribution of obstacles.

[0044] Based on the two-dimensional occupancy grid map information and the robot's current pose obtained through the positioning system, the exploration module can carry out an autonomous decision-making process. This exploration module can comprehensively consider the global environmental structure and local perception information, use multi-objective optimization strategies to plan a reasonable exploration sequence, and generate corresponding global reference paths and phased navigation target points.

[0045] Subsequently, the system continuously evaluates the completion status of the current exploration task and transmits the determined navigation target to the motion planning layer. The motion planner within this layer, based on the adopted path search algorithm and considering robot motion constraints and dynamic obstacles, generates a smooth and collision-free executable path. Finally, this path is sent to the underlying controller for parsing. Through real-time calculation and output of the required linear and angular velocity commands, the robot is driven to move precisely and stably along the predetermined trajectory, thus achieving closed-loop autonomous navigation and exploration throughout the entire process.

[0046] Furthermore, the specific processes involved in the above preferred embodiments can be found in [reference needed]. Figure 4 As shown.

[0047] During the global path planning phase, the system first uses the robot's current position... The root node is used to run the RRT (Rapid Exploratory Random Tree) algorithm in the configuration space. This algorithm continuously grows a random tree through random sampling and expansion. When a branch of the tree touches the boundary of an unexplored region, the algorithm terminates and outputs a candidate set of "global boundary points" distributed throughout the space.

[0048] The system then proceeds to the target point selection phase. First, it divides the global map into regions based on the structural characteristics of the environment. Then, it integrates heuristic information such as distance and expected exploration benefits to evaluate and select the most valuable boundary points from each region and prioritize them.

[0049] Finally, during the route generation phase, these ordered boundary points are linked together to form a global exploration route. This route, acting as a high-level guide, transforms the abstract exploration task into a concrete sequence of waypoints. This hierarchical planning mechanism effectively avoids the robot's ineffective wandering and repetitive traversal within already explored areas, thereby significantly improving the overall system's exploration efficiency and autonomy.

[0050] For sub-region division, map segmentation can employ a dynamically adjusted gridding method. Initially, the entire map is divided into a uniform 2×2 grid. This grid contains M×N sub-regions, where M and N are the number of partitions in the horizontal and vertical directions, respectively.

[0051] During autonomous exploration, the system continuously monitors the size of each sub-region. Once the side length (length or height) of a sub-region exceeds twice the diameter d of the LiDAR field of view, it means that the current grid granularity is insufficient to accurately describe the local features of the region. At this point, the system will increase the number of partitions by 1 in the corresponding direction (M or N), dynamically subdividing the region to ensure the accuracy of environmental representation.

[0052] During the sub-region selection process, candidate regions can be screened after the initial division of the environment. The screening criteria are as follows: if a sub-region has boundary points and its state is determined to be "internal region" or "mostly unknown," then that sub-region is considered a candidate.

[0053] To accurately quantify the fuzzy state of "most areas unknown," this invention designs a region state evaluation function. This function achieves this by uniformly sampling the grids within a sub-region and calculating the proportion of grids in an unknown state. When this proportion exceeds a preset threshold, the function returns "true," indicating that the region contains a large amount of unknown space and possesses high exploration value. Finally, the system selects the K sub-regions with the best evaluation results from all candidate sub-regions to form an initial set of sub-regions to be explored.

[0054] Intelligent global path sorting: After determining the initial set of sub-regions to be explored, they need to be intelligently sorted to generate the final global planning route. To achieve a balance between exploration efficiency and real-time planning, the generation of this global planning route comprehensively optimizes the following four principles: The principle of starting from the nearest point: the starting point of the route must be set to the sub-region where the robot is currently located, in order to minimize the initial movement cost and ensure a rapid response.

[0055] The principle of shortest path: The order of area visits should minimize the total movement trajectory length of the robot to effectively reduce task execution time and energy consumption.

[0056] Principle of consistent direction: When planning, prioritize areas with smaller deviations from the robot's current heading angle as the next target to reduce unnecessary turns and improve motion smoothness.

[0057] Continuous backtracking avoidance principle: The newly generated global planning route should maintain a certain structural similarity with the previous historical route. By introducing path continuity constraints, inefficient backtracking behavior is avoided, and a smooth transition in the exploration process is ensured.

[0058] During the local path planning phase, an environment perceiver can be introduced to dynamically configure the cost function weights by invoking the Local Path Planning Algorithm (Dynamic Window Approach, DWA).

[0059] Local path planning (DWA) algorithms are widely used for real-time obstacle avoidance and trajectory generation in dynamic environments for mobile robots. The basic principle is to define a "dynamic window" in the robot's velocity space (linear velocity v, angular velocity ω), which is determined by both robot dynamics constraints and sensor range. Then, several feasible velocity combinations are sampled within this window, and the optimal velocity command is selected by evaluating indicators such as safety, target proximity, and velocity smoothness of each candidate trajectory. In this embodiment, an environmental sensor can be introduced to dynamically adjust the weight parameters in the DWA cost function, enabling the planning behavior to adaptively respond to changes in environmental complexity.

[0060] Its decision-making logic is as follows: when the environment changes from "simple" to "complex", the velocity coefficient is reduced and the obstacle distance coefficient is increased. The purpose is to prioritize safety and obstacle avoidance response in areas with dense obstacles, reduce the travel speed, enhance obstacle avoidance ability, and shorten the decision time.

[0061] Conversely, when the environment changes from "complex" to "simple", increasing the velocity weight and decreasing the obstacle distance weight aims to maximize movement efficiency in open areas, improve operational efficiency, accelerate exploration progress, and use inertia to help the robot escape local minima (such as U-shaped traps).

[0062] In another embodiment, the detection of environmental state transitions described above can be achieved by analyzing the local occupancy grid map output by the SLAM module and the current frame LiDAR data to extract multiple geometric and topological features related to the spatial structure, and comprehensively judging the complexity of the current environment. Preferably, the calculation of the environmental complexity can be performed by a lightweight environmental sensor module.

[0063] That is, in one embodiment, the local planner may further include an environment sensor, which monitors the structural characteristics of the robot's environment and generates corresponding environmental state signals. Preferably, the environment sensor includes at least a local map analysis unit, a geometric feature extraction unit, and an environment state recognition module.

[0064] The environmental state recognition module is configured to determine whether the current environment has changed from a simple state to a complex state, or vice versa, based on the extracted obstacle density, number of free passage directions, and local curvature index, combined with a hysteresis comparison mechanism.

[0065] In another embodiment, the system may further include an environment state recognition module, which is used to calculate the current environment complexity based on the obstacle distribution density, the number of free passage directions and the local geometric curvature in the local map, and to determine whether the environment state has changed from simple to complex in combination with a hysteresis comparison mechanism; the state signal is used to adjust the cost weights of the velocity term and the obstacle avoidance term in the DWA local planner in real time, thereby realizing intelligent adaptation of navigation behavior.

[0066] Traditional DWA (Directed Exploration and Mapping) methods rely on manually setting fixed cost weights, which cannot cope with dynamically changing exploration environments. This invention introduces a hierarchical environment sensor, particularly an integrated environment state recognition module, enabling semantic-level understanding of the environmental context. This allows path planning behavior to adaptively adjust as the scene evolves, significantly improving the robot's navigation robustness and exploration efficiency in confined spaces.

[0067] Furthermore, the SLAM module has been improved upon Cartographer-SLAM, specifically including: (1) To alleviate the “zero drift” phenomenon caused by the cumulative error of the odometer during the mapping process, this embodiment of the invention introduces a real-time CSM matcher (RTCSM). To this end, the matcher has made key improvements to the traditional correlation scanning matching method: it performs real-time matching between the current lidar pose and the local sub-map being constructed, and prioritizes the selection of regions with high matching confidence for secondary optimization and alignment.

[0068] The above method effectively maintains map structure consistency in complex environments and avoids error propagation caused by single-frame matching failures. By executing this matching strategy frame by frame, the system significantly suppresses drift accumulation during long-term operation, thereby reducing overall mapping errors and ultimately improving the usability, quality, and overall reliability of the constructed map.

[0069] (2) To address the problem that traditional "Scan-to-Map" methods are prone to data association errors in feature-sparse or dynamic scenarios, this embodiment of the invention proposes a sub-map matching strategy. The core of this strategy is to match the local sub-map formed by the current laser data with existing historical sub-maps. As a fusion product of multiple frame scans, the sub-map's point cloud distribution and structural features are far more stable and discriminative than those of a single frame scan.

[0070] This rich structural information provides stronger constraints for pose optimization, significantly improving matching accuracy and anti-interference capabilities. Experiments have confirmed that this strategy is an effective solution that simultaneously achieves real-time performance and high accuracy, greatly enhancing the robustness of the SLAM system in challenging environments.

[0071] (3) Set the center of the lidar as the initial starting point for the robot's movement, and define all the distance data collected during one rotation as one frame of lidar data. By processing the multi-directional distance information in real time through the control board, the local position of the surrounding obstacles relative to the lidar can be calculated.

[0072] (4) Introducing delayed decision-making: when a potential loop is detected, pose graph optimization is not performed immediately. Instead, the matching results of subsequent subgraphs are continuously collected. The final loop correction is triggered after the confidence accumulation condition is met, so as to improve the accuracy of the algorithm in loop detection in the constructed map.

[0073] Based on the same inventive concept, embodiments of the present invention also provide a robot, specifically, the robot includes: a robot motion control system and a robot autonomous mapping system in a confined space environment as described in any of the above embodiments.

[0074] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium storing machine-executable instructions. When the machine-executable instructions are invoked and executed by a processor, the machine-executable instructions cause the processor to perform the steps of the above-described method.

[0075] In the embodiments provided by this invention, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0076] Furthermore, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0077] Furthermore, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0078] It should be noted that if the functionality is implemented as a software module and sold or used as an independent product, it 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 various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0079] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0080] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for autonomous mapping of robots in a confined space environment, characterized in that, include: Multimodal perception data collected in real time by the sensors on the robot is input into the instant localization and map building module for front-end pose estimation and back-end map optimization, generating and continuously updating a two-dimensional occupancy grid map. Based on the two-dimensional occupied grid map and the robot's current pose information, dynamic partitioning processing is performed on the map to identify unexplored boundary areas, and a phased navigation target point sequence is determined according to a multi-objective optimization strategy. Based on the phased navigation target point sequence and real-time sensing information, a local path planning algorithm based on environmental perception is used to generate a collision-free path that satisfies the robot's kinematic constraints. The collision-free path is used to represent the route the robot takes when performing an active mapping task.

2. The method according to claim 1, characterized in that, The front-end pose estimation includes: Perform relevant scan matching between the current frame laser scan data and the local submap being constructed; Based on the matching confidence level, high-confidence regions are selected for secondary optimization and alignment.

3. The method according to claim 2, characterized in that, The step of selecting high-confidence regions based on matching confidence for secondary optimization alignment includes: During the relevant scan matching process, the confidence score of each point cloud matching result is calculated; A threshold is set based on the confidence score, and a set of points higher than the threshold is extracted to form a high-confidence region; Based on the point cloud features within the high-confidence region, the local optimal transformation solution is re-executed to correct the pose estimate of the current frame.

4. The method according to claim 2, characterized in that, The backend map optimization includes: Match the current local subgraph with the historical subgraphs to obtain pose constraints; Pose graph optimization is performed based on the pose constraints to correct accumulated errors.

5. The method according to claim 1, characterized in that, Identify unexplored boundary areas and determine a phased navigation target point sequence based on a multi-objective optimization strategy, including: Identify the set of boundary points between known and unknown regions in the two-dimensional occupied grid map; From the sub-regions after dynamic partitioning, select target sub-regions that contain the boundary points and whose internal unknown state grid ratio exceeds a first preset threshold, as candidate exploration regions; Based on a preset strategy, the candidate exploration areas are prioritized and a global exploration route is generated. Extract the phased navigation target points sequentially from the global exploration route.

6. The method according to claim 1, characterized in that, The cost function of the local path planning algorithm based on environmental perception improvement includes weight coefficients, which include speed term weights and obstacle avoidance term weights. The adjustment methods for the weighting coefficients include: Real-time detection of environmental complexity in local maps to determine trends in environmental state changes; When the environmental state is detected to change from a simple state to a complex state, the weight of the speed term is reduced, while the weight of the obstacle avoidance term is increased. When the environmental state is detected to change from a complex state to a simple state, the weight of the speed term is increased, while the weight of the obstacle avoidance term is decreased.

7. The method according to claim 6, characterized in that, Real-time detection of environmental complexity in a local map, including: The environmental complexity assessment value is calculated based on the density of obstacle points and / or the number of boundary points per unit area in the local map. A baseline for environmental complexity is established. When the evaluation value rises above the upper limit of the baseline, it is determined to be a complex state. When the evaluation value falls below the lower limit of the baseline, it is determined to be a simple state. Determine whether the environmental state has switched based on the difference in environmental state between the previous and next times.

8. A robot autonomous mapping system for confined space environments, characterized in that, include: Sensor components are used to acquire multimodal perception data of the robot's surrounding environment in real time; The real-time positioning and mapping module is communicatively connected to the sensor component and configured to receive the multimodal sensing data, perform front-end pose estimation and back-end map optimization, and generate and continuously update a two-dimensional occupancy grid map. The exploration decision module is communicatively connected to the real-time positioning and mapping module. It is configured to perform dynamic partitioning processing on the map based on the two-dimensional occupied grid map and the robot's current pose information, identify unexplored boundary areas, and determine a phased navigation target point sequence according to a multi-objective optimization strategy. The local path planning module is communicatively connected to the exploration decision module and the sensor components. It is configured to generate a collision-free path that satisfies the robot's kinematic constraints by using a local path planning algorithm based on environmental perception improvement, according to the phased navigation target point sequence and real-time sensing information. The path execution interface module is configured to output the collision-free path to the robot motion control system as a travel route instruction for the robot during the active mapping task.

9. A robot, characterized in that, include: The robot motion control system and the robot autonomous mapping system in a confined space environment as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method according to any one of claims 1 to 7.

Images