An autonomous navigation and obstacle avoidance system for a fire-fighting robot

Abstract

The application discloses an autonomous navigation and obstacle avoidance system of a fire-fighting robot, and relates to the technical field of navigation and obstacle avoidance.The application effectively guarantees the accuracy and feasibility of path planning by constructing a three-dimensional state unit set and eliminating inaccessible areas;identifies a task switching point, automatically divides path paragraphs according to the task type, enhances the logical clarity and control accuracy of the navigation task, and especially in the scene of path interruption or environmental mutation, the system can realize local reconstruction and continuity recovery of the path through a graph structure updating and path correction module, and avoid the delay and resource waste caused by global re-planning in the traditional system.The application also combines a space coincidence degree calculation and a jump detection method, effectively improves the reliability of key frame node identification and the stability of task switching.

Description

Technical Field

[0001] This invention relates to the field of navigation and obstacle avoidance technology, and in particular to an autonomous navigation and obstacle avoidance system for a firefighting robot. Background Technology

[0002] Against the backdrop of rapid urbanization, firefighting missions face multiple challenges, including high environmental complexity, tight emergency response times, and significant human risk. To effectively improve operational efficiency and safety in fire rescue, intelligent firefighting robots, as an important means of replacing manual labor, are gradually gaining attention. Especially in high-risk environments such as toxic, oxygen-deficient, or high-temperature environments, utilizing robots with strong autonomous navigation and obstacle avoidance capabilities to assist or replace firefighters in reconnaissance, firefighting, and search and rescue tasks has become a key direction for the development of intelligent firefighting systems. Currently, mainstream research focuses on improving the navigation and obstacle avoidance capabilities of firefighting robots through path planning, environmental perception, and obstacle recognition. Common path construction methods include grid maps, global shortest path search, and sensor-based 3D environmental modeling. Some solutions have achieved basic navigation and obstacle avoidance functions, but shortcomings remain in dynamic environment adaptability, task continuity control, and recovery capabilities after path interruption.

[0003] CN102359784B discloses an autonomous navigation and obstacle avoidance system and method for indoor mobile robots. It constructs a wireless sensor network and divides the indoor environment into several grid areas. Global path planning is achieved by combining dilatational and Dijkstra's algorithms. The robot coordinates with sensing nodes through a communication module to achieve localization and path execution. While this scheme has certain advantages in positioning accuracy and path planning, it relies on a pre-deployed static sensor network, resulting in high system deployment costs. Furthermore, in scenarios with sudden changes in environmental structure or sudden obstacle movement, the path update mechanism lags behind, and there is a lack of targeted path repair strategies. Therefore, it cannot meet the requirements for path reconstruction and keyframe segmentation control in dynamic environments such as fire scenes.

[0004] CN118623894B discloses an autonomous navigation and obstacle avoidance method for robots in dynamic scenes. It employs an image recognition model to process environmental images, distinguishes between dynamic and static obstacles, selects different obstacle avoidance algorithms, and dynamically adjusts the navigation path. This method demonstrates good adaptability in handling different types of obstacles, improving the robot's obstacle avoidance capabilities in complex scenes. However, this approach focuses on the obstacle avoidance strategy itself and lacks the ability to model the overall task path structure. Path control remains a single-segment continuous control, without addressing segmented management and state transition logic control of the path task. This makes it difficult to meet the requirements of multi-target state switching and automatic recovery after path interruption in firefighting missions. The system's global path robustness and controllability of critical path segments still need improvement. Summary of the Invention

[0005] In view of the problems existing in the navigation technology of fire-fighting robots, this invention is proposed.

[0006] Therefore, the problem to be solved by this invention is how to implement a flexible update mechanism in a dynamic and frequently changing emergency environment to cope with the risk of path interruption caused by sudden changes in obstacles.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, the present invention provides an autonomous navigation and obstacle avoidance system for a firefighting robot, comprising: a state graph construction and obstacle perception encoding module for dividing the work area and constructing a state transition probability graph; a key frame extraction module for calculating the optimal state path in the state transition probability graph based on the transition probability cost and extracting task type switching points as key frame state nodes to achieve segmented control of the path task; a graph structure update module for dynamically adjusting the state node and edge set in the state transition probability graph after obstacle detection or environmental changes, and reconstructing the key frame sequence to adapt to the new structure; and a path correction module for performing path backtracking and local reconstruction based on a pointer array after changes in the state transition probability graph structure or path interruption, updating the key frame sequence a second time, and restoring path continuity.

[0009] As a preferred embodiment of the autonomous navigation and obstacle avoidance system for the firefighting robot described in this invention, the construction of the state transition probability graph includes: constructing a set of state units in a three-dimensional space based on the building structure diagram of the firefighting scenario, measured laser point cloud data, and preset grid size, and removing inaccessible areas and occluded elements from each spatial unit; based on the task type set and the carrier set, using a pre-constructed task-carrier adaptation table to filter state units that meet the carrier constraints, generating a set of state nodes; in the set of state nodes, if any pair of state nodes satisfies that the Euclidean distance between the state nodes is less than the jump threshold, and the state units have a structural connectivity relationship, then a transition edge is added to the set of transition edges, and a state transition probability graph is constructed.

[0010] As a preferred embodiment of the autonomous navigation and obstacle avoidance system for the firefighting robot described in this invention, the keyframe extraction module includes: classifying and labeling each state node based on the set of state nodes, according to the set of task types and the set of carriers, and selecting state nodes representing the start and end points of the task; in the state transition probability graph, using the selected start and end point state nodes as the two ends, and combining the transition probability cost of the transition edges to perform path planning operations, resulting in an ordered state node access sequence; in the state node access sequence, identifying state nodes corresponding to changes in task type, and selecting state nodes with task segmentation characteristics as a candidate set of task keyframe state nodes, and dividing them into continuous task segments; in the candidate set of task keyframe state nodes, combining the spatial overlap of the cross-connection transition edges, scoring the task switching intensity between adjacent state nodes, selecting state nodes with scores exceeding the stability threshold, and aggregating them into the final task keyframe state node sequence.

[0011] As a preferred embodiment of the autonomous navigation and obstacle avoidance system of the firefighting robot described in this invention, the path planning operation includes: identifying and selecting the starting state node and ending state node in the state transition probability graph, and recording them as boundary nodes for path planning; constructing a graph structure with state nodes as vertices and transition probability costs as edge weights for all accessible paths between boundary nodes, and calculating the path cost by accumulating costs; for each accessible path, summing the negative logarithmic probability costs of the transition edges, and recording the state node access sequence corresponding to the path with the minimum total cost.

[0012] As a preferred embodiment of the autonomous navigation and obstacle avoidance system of the firefighting robot described in this invention, the step of screening state nodes with task segmentation characteristics includes: based on the task type of each state node in the state node access sequence, identifying the location where the task type changes, and marking the corresponding state node as a candidate task segment boundary node; for each candidate task segment boundary node, evaluating the task switching intensity by combining the jump amplitude of the transition probability cost of the preceding and following state nodes, and filtering out state nodes whose change amplitude is lower than the change threshold; gathering the remaining candidate task segment boundary nodes into a set of candidate task keyframe state nodes, and dividing the continuous task segments between adjacent state nodes according to their location; for each candidate task keyframe state node, identifying the set of adjacent transition edges formed with the preceding and following state nodes, and extracting the pairs of transition edges with cross-connection relationships; for each pair of transition edges, calculating the ratio of the number of overlapping voxels to the total number of adjacent voxels in the spatial coordinate system, as the spatial overlap score of the task switching point;

[0013] The spatial overlap score is compared with the stability threshold, and state nodes with scores higher than the stability threshold are selected as high-confidence task key frame state nodes, forming a key frame state node sequence.

[0014] As a preferred embodiment of the autonomous navigation and obstacle avoidance system for the firefighting robot described in this invention, the dynamic adjustment of the state node and edge set in the state transition probability graph includes: after acquiring environmental change information or obstacle detection results, locating the affected area in the state unit set and marking the affected state node set; performing a validity check on the affected state node set, and if the state unit where the state node is located is occluded or impassable, removing it from the affected state node set and simultaneously removing the associated transition edge; for areas where a new passable path appears due to structural occlusion, regenerating the state node based on the connectivity of the state unit and supplementing the transition edge according to the spatial topological relationship with surrounding nodes; re-executing the path planning operation on the adjusted state transition probability graph, and adding, deleting, and correcting the state node sequence of the task key frame to adapt to the state transition logic after the structural update.

[0015] As a preferred embodiment of the autonomous navigation and obstacle avoidance system for the firefighting robot described in this invention, the following steps are included: When the path structure in the state transition probability graph is interrupted or altered, the range of affected subsequences in the original state node access sequence is determined, and the start and end state nodes are identified; based on the forward pointer array, backtracking is performed from the end node of the affected subsequence to reconstruct the shortest path subsequence to the starting point; in the backtracked path, combined with the unaffected task key frame state nodes in the original state node access sequence, local connections and task segment reorganization are performed to restore the continuity of the path structure; the backtracked and repaired state node access sequence is output, and the task key frame state node sequence is updated a second time to support subsequent task segmented navigation control operations.

[0016] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the autonomous navigation and obstacle avoidance system of the fire-fighting robot as described in the first aspect of the present invention are implemented.

[0017] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the autonomous navigation and obstacle avoidance system of the fire-fighting robot as described in the first aspect of the present invention are implemented.

[0018] The beneficial effects of this invention are as follows: By constructing a three-dimensional state unit set and eliminating impassable areas, this invention effectively ensures the accuracy and feasibility of path planning; by identifying task switching points and automatically dividing path segments according to task type, it enhances the logical clarity and control precision of navigation tasks. Especially in scenarios of path interruption or sudden environmental changes, this system can achieve local reconstruction and continuity restoration of the path through graph structure update and path correction modules, avoiding the delays and resource waste caused by global replanning in traditional systems. This invention also integrates spatial overlap calculation and jump detection methods, effectively improving the reliability of key frame node identification and the stability of task switching. In summary, this invention significantly improves the navigation adaptability and task execution continuity of firefighting robots in dynamic, high-risk environments, providing strong technical support for intelligent emergency operations in high-risk environments. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the autonomous navigation and obstacle avoidance system for a firefighting robot. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] As mentioned in the background section, with the rapid advancement of modern urbanization, firefighting missions face multiple challenges, including high environmental complexity, tight emergency response times, and significant human risk. To effectively improve operational efficiency and safety in fire rescue, intelligent firefighting robots, as an important means of replacing manual labor, are gradually gaining attention. Especially in high-risk environments such as toxic, oxygen-deficient, or high-temperature environments, utilizing robots with strong autonomous navigation and obstacle avoidance capabilities to assist or replace firefighters in reconnaissance, firefighting, and search and rescue tasks has become a key direction for the development of intelligent firefighting systems. Currently, mainstream research focuses on improving the navigation and obstacle avoidance capabilities of firefighting robots through path planning, environmental perception, and obstacle recognition. Common path construction methods include grid maps, global shortest path search, and sensor-based 3D environmental modeling. Some solutions have achieved basic navigation and obstacle avoidance functions, but they still have shortcomings in dynamic environment adaptability, task continuity control, and recovery capabilities after path interruption.

[0025] Figure 1 This is a flowchart of an autonomous navigation and obstacle avoidance system for a firefighting robot according to an embodiment of the present invention. Figure 1 As shown, the autonomous navigation and obstacle avoidance system of the firefighting robot includes:

[0026] The state graph construction and obstacle perception coding module 100 is used to divide the work area and construct a state transition probability graph.

[0027] First, a set of state units in a three-dimensional space is constructed based on the building structure diagram of the fire scene, measured laser point cloud data, and preset grid size. Then, inaccessible areas and obstructing elements are removed from each spatial unit.

[0028] Specifically, in actual fire safety scenarios, building structural drawings are two-dimensional or three-dimensional structural plans or volumetric models, typically provided by fire departments, urban planning departments, or construction companies. They include various structural information such as room boundaries, passageway directions, stair slopes, and ventilation opening locations. However, building structural drawings alone cannot dynamically reflect on-site access conditions; therefore, it is necessary to combine them with on-site laser point cloud data to obtain the actual space occupancy.

[0029] Laser point cloud data is collected using lidar installed on firefighting robots or pre-survey equipment. The resulting data reflects the specific coordinates and density distribution of obstacle objects in three-dimensional space. Conventional three-dimensional space partitioning methods include octree spatial partitioning, voxel block algorithms, and regular grid partitioning strategies. However, these methods suffer from information redundancy or ambiguity when dealing with complex occlusion and unstructured firefighting environments. To address this, this invention employs a joint modeling approach based on grid size, point cloud density threshold, and structural topology verification. The laser point cloud is projected onto the coordinate system corresponding to the building structure diagram. Regular grid units are divided with a set size (e.g., 0.5-meter side length). The point cloud density of each unit is then statistically analyzed. If the density exceeds the occlusion threshold (e.g., 100 points / cubic meter), it is identified as an occlusion object and is not included in subsequent state diagrams.

[0030] In addition, it is necessary to perform topological validity verification on whether there are structural passages between continuous units. If there are staircases connecting the upper and lower floors, vertical accessibility relationships need to be manually added.

[0031] The above processing ultimately forms a set of state units that can be accessed in three-dimensional space, providing a basic spatial framework for the subsequent construction of state transition probability diagrams.

[0032] Secondly, based on the task type set and carrier set, a pre-built task-carrier adaptation table is used to filter state units that meet the carrier constraints and generate a set of state nodes.

[0033] It should be noted that before constructing the state transition diagram, functional attributes and task adaptability must be defined for each state unit. In this operation, the task type set consists of preset task execution categories, such as "high-temperature area monitoring," "smoke boundary detection," and "search for key personnel." Each task type has specific operational requirements, such as movement speed limits, target recognition range requirements, or continuous dwell time limits. The carrier set includes the physical platform types carried by the firefighting robot, such as tracked robots, quadrupedal walking robots, and multi-rotor drones. Each carrier has unique capabilities such as mobility, climbing limitations, and minimum turning radius. The task-carrier adaptability table is constructed in a matrix manner, where each unit corresponds to a logical judgment flag indicating whether a task-carrier combination matches a spatial unit. For example, a multi-rotor drone is deemed unsuitable in a spatial unit with a roof obstruction, and a tracked robot is similarly eliminated in a high-slope area. Based on this table, each spatial unit in the state unit set is traversed one by one, and spatial units that satisfy at least one task-carrier combination are selected by combining their structural attributes and environmental factors.

[0034] This screening process creates a set of task-adaptive state nodes. Each state node, in addition to its three-dimensional spatial coordinates, must also include five fields: node number, spatial unit number, list of supported task types, list of allowed carrier types, and risk factor identifier. These fields support subsequent path deduction and state transformation calculations.

[0035] Finally, in the set of state nodes, if any pair of state nodes satisfies that the Euclidean distance between the state nodes is less than the jump threshold and that the state units they belong to have a structural connectivity relationship, then a transition edge is added to the set of transition edges, and a state transition probability graph is constructed.

[0036] Specifically, based on the already generated set of state nodes, it is necessary to determine whether any pair of state nodes has the conditions for a transition edge.

[0037] The criteria for judgment include that the Euclidean distance does not exceed the jump threshold and that there is spatial connectivity between the state units of the nodes. Connectivity includes spatial continuity and structural interoperability. The former determines whether two units share a boundary surface or adjacent voxels, while the latter determines whether there are gate, passage, or platform connection structures.

[0038] When two state nodes satisfy the above two conditions, a transition edge record is added to the state transition graph. Furthermore, a transition probability weight is assigned to each transition edge, reflecting the reliability and risk cost of the robot's autonomous navigation from one state node to another. Conventional transition probability assignment methods are based on pure path distance or historical traversal frequency, but do not consider task characteristics, vehicle capabilities, and scene dynamics. To enhance the matching degree between path decision-making and environmental adaptation, this invention assigns a passability evaluation score, a task matching index, and a dynamic passability assessment factor, respectively, and these three are normalized to synthesize the final transition probability value. Finally, a state transition probability graph is constructed with state nodes as vertices and transition probability weighted edges as edge sets to support keyframe extraction and path planning modules.

[0039] The keyframe extraction module 200 is used to calculate the optimal state path in the state transition probability graph based on the transition probability cost, and extract the task type switching point as the keyframe state node to realize the path task segmentation control.

[0040] The keyframe extraction module includes the following sub-units:

[0041] Subunit 201: Based on the set of state nodes, classify and label each state node according to the set of task types and the set of carriers, and select the state nodes that represent the start and end points of the task.

[0042] Specifically, firstly, the task scenario attributes required for each task in the task type set are extracted, such as target recognition field of view requirements, dwell time minimum limit, and platform stability requirements. At the same time, the capability parameters of different carriers in the carrier set are extracted, such as maximum reachable distance, minimum turning radius, and maximum crossing capability, to form a task-carrier adaptation condition set.

[0043] Based on this, each state node is subjected to dual screening: on the one hand, it is determined whether the spatial location and attributes of the state node support the execution of a specific task; on the other hand, it is determined whether the state node allows a specific type of carrier to enter.

[0044] If the above two dimensions of adaptation requirements are met at the same time, the corresponding task type and carrier category attributes are marked for the state node. In the entire set of state nodes, the node groups marked as task start flags and task end flags are identified respectively, forming the starting point set and ending point set in the task path, which serve as boundary anchor points for subsequent path deduction.

[0045] Subunit 202: In the state transition probability graph, the path planning operation is performed by selecting the state nodes at the starting point and the ending point as the two ends, and combining the transition probability cost of the transition edge, to obtain an ordered sequence of state node visits.

[0046] In a better approach, the starting and ending state nodes are identified and recorded as boundary nodes for path planning in the state transition probability graph. For all accessible paths between boundary nodes, a graph structure is constructed with state nodes as vertices and transition probability costs as edge weights. Path costs are calculated by accumulating costs. For each accessible path, the negative logarithmic probability costs of the transition edges are summed, and the state node access sequence corresponding to the path with the minimum total cost is recorded.

[0047] It should be noted that since solving the maximum probability product path is not suitable for direct multiplication operations due to numerical overflow and computational instability, a negative logarithmic transformation is performed on each transition probability edge before path planning. The cost of each edge is defined as the negative logarithmic value of the original probability. That is, the maximum probability path problem is transformed into the minimum cost path problem, which satisfies the non-negative weighted graph model requirement of shortest path algorithms (such as Dijkstra's algorithm or dynamic programming).

[0048] In this graph structure, starting from the starting node of each task segment, the system performs layer-by-layer cost accumulation calculations on the entire graph, constructs a dynamic cost array, and synchronously updates the optimal predecessor node record array. After the destination node is visited, the system traverses the predecessor node array in reverse from the destination node to the starting node to obtain the ordered sequence of state nodes contained in the path with the minimum total cost, and outputs the state node visit sequence.

[0049] Subunit 203: In the state node access sequence, identify the state nodes corresponding to the task type change, and filter the state nodes with task segmentation characteristics as a set of candidate task key frame state nodes, and divide them into continuous task segments.

[0050] The process of selecting state nodes that possess task segmentation characteristics includes the following steps:

[0051] Based on the task type of each state node in the state node access sequence, the location where the task type changes is identified, and the corresponding state node is marked as a candidate task segment boundary node.

[0052] For each candidate task segment boundary node, the task switching intensity is evaluated by combining the jump magnitude of the transition probability cost of the preceding and following state nodes, and state nodes with a change magnitude lower than the change threshold are filtered out.

[0053] Specifically, for each candidate boundary node, the difference between its transition probability cost and that of the previous and next state nodes is calculated, defined as the positive transition value and the negative transition value. The larger absolute value of these two values ​​is taken as the transition magnitude of the current node, which represents the degree of path cost mutation faced by the system when the task switches at the path cost level.

[0054] If the jump amplitude is less than the set change threshold, it indicates that although the task label of the state node has changed, the path cost has not changed significantly and cannot effectively support the physical execution of task switching. The system will then exclude it from the candidate set to improve the representativeness and switching rationality of subsequent key frame nodes.

[0055] The retained candidate task segment boundary nodes are aggregated into a set of candidate task keyframe state nodes, and continuous task segments between adjacent state nodes are divided according to their location.

[0056] Specifically, after completing the task switching intensity screening, the system arranges the remaining candidate task segment boundary nodes according to their positional order in the original access sequence, forming a set of candidate task keyframe state nodes. Subsequently, using two adjacent keyframe state nodes in this set as demarcation points, the state node access sequence is divided into multiple non-overlapping subsequences. Each subsequence constitutes a continuous task segment path, mapped to a specific task type.

[0057] This division method ensures structural connectivity and consistency within task segments, and that inter-segment switching nodes exhibit significant path characteristic jumps, which helps the path control system achieve optimized navigation within segments and scheduling for inter-segment switching.

[0058] For candidate task keyframe state nodes, identify the set of adjacent transition edges formed by the preceding and following state nodes, and extract transition edge pairs with cross-connection relationships.

[0059] Specifically, the local path structure of each candidate keyframe state node is further analyzed. In the two transition edges formed between the adjacent state nodes and itself, it is searched for whether there is a cross connection relationship in the spatial path direction. That is, it is determined whether the two transition edges have structural features such as different path directions, spatial intersection or voxel overlap.

[0060] If such structural features exist, the candidate keyframe node is considered to be located at a path intersection or structural abrupt change location, and has high path switching significance. Its edge pair is collected into the cross-connection edge set for subsequent spatial overlap calculation.

[0061] For each pair of transition edges, the ratio of the number of overlapping voxels to the total number of adjacent voxels in the spatial coordinate system is calculated and used as the spatial overlap score of the task switching point.

[0062] The specific operation is as follows: count the number of overlapping voxels A between the path voxel sets of two edges in 3D space, calculate the total number of voxels B in the union of the two sets, and then calculate the spatial overlap score S=A / B. This score reflects the degree of overlap of the current path switching point in space. The higher the overlap, the smoother the path structure transformation and the more stable the connection, making it suitable as a keyframe transition point; conversely, a lower overlap indicates a drastic path change or a jump, and should be carefully considered before including it in a keyframe.

[0063] f. Compare the spatial overlap score with the stability threshold, select state nodes with scores higher than the stability threshold as high-confidence task key frame state nodes, and form a key frame state node sequence.

[0064] Subunit 204: In the set of candidate task keyframe state nodes, the task switching intensity between adjacent state nodes is scored by combining the spatial overlap of the cross-connection transition edges, and state nodes with scores exceeding the stability threshold are selected and aggregated into the final task keyframe state node sequence.

[0065] First, for each candidate task segment boundary node, the system identifies the set of transition edge pairs formed by the transition edges between the node and the preceding and following nodes. Then, for each transition edge pair, the system extracts the voxel blocks covered in the 3D spatial coordinate system and calculates the spatial overlap between the voxel sets corresponding to the two transition edges, i.e., the ratio of the number of overlapping voxels to the total number of voxel unions. This ratio reflects the degree of spatial transition between adjacent path segments.

[0066] If the spatial overlap score is higher than the set stability threshold, the task switching point is considered to have strong structural stability and task representativeness. The system marks the corresponding state node as a high-confidence keyframe state node, and summarizes all such nodes to form the final task keyframe state node sequence, which serves as the core anchor point for path segmentation control.

[0067] The graph structure update module 300 is used to dynamically adjust the set of state nodes and edges in the state transition probability graph after obstacle detection or environmental changes, and reconstruct the keyframe sequence to adapt to the new structure.

[0068] The set of state nodes and edges in the dynamically adjusted state transition probability graph includes:

[0069] Step 1: After obtaining information on environmental changes or obstacle detection results, locate the affected area in the set of state units and mark the set of affected state nodes.

[0070] In actual operation, firefighting robots operate in highly dynamic environments, which may be significantly affected by collapses, the spread of dense smoke, or the appearance of new obstacles, thus impacting the original path planning structure. To enable timely perception of these changes, the system is equipped with an environmental perception module based on multi-source data fusion. This module dynamically analyzes the temporal changes in structures within the firefighting scenario based on visual information, LiDAR point clouds, and infrared thermography images.

[0071] After detecting newly emerging obstacle or passable areas, the system first searches for the corresponding spatial coordinates in the original set of 3D state units and quickly locates the affected set of state units using a rasterized indexing mechanism. Subsequently, based on the one-to-one mapping relationship between state nodes and state units, the system synchronously marks the state nodes corresponding to all affected state units, forming a set of affected state nodes.

[0072] Step 2: Perform a validity check on the set of affected state nodes. If the state unit where the state node is located is occluded or not accessible, remove it from the set of affected state nodes and simultaneously remove the associated transition edges.

[0073] Specifically, changes in the working environment may cause some previously accessible areas to become inaccessible, such as physical path closures caused by falling objects, excessively high smoke density, or exceeding temperature limits. Therefore, it is necessary to evaluate the validity of each state node in the set of affected state nodes. The system performs a triple evaluation of the current connectivity, visibility level, and heat load value of the state unit. If any dimension fails to meet the set accessibility conditions, the state unit is determined to be in an invalid state, and the corresponding state node is removed. Furthermore, to maintain the topological consistency of the state transition probability graph, the system traverses the current set of state transition edges, finds all transition edges that are connected to the aforementioned invalid state nodes, and removes these edges from the graph structure to avoid illegal jumps or empty path access issues in path planning.

[0074] Step 3: For regions where new accessible paths emerge due to structural obstruction, regenerate state nodes based on the connectivity of state units, and supplement transfer edges according to the spatial topological relationship with surrounding nodes.

[0075] It should be noted that in some cases, after obstacles are removed, smoke is cleared, or moving targets avoid the original path, accessible paths will be exposed again in the operation scenario. In order to make full use of this new open path, the system continuously detects the newly formed accessible paths in the state unit set based on the dynamic connectivity analysis mechanism.

[0076] Once a state unit in a certain area is found to meet the conditions for unobstructed space and is accessible to robots, the system will rebuild the state node at that unit location and assign corresponding task adaptation attributes and carrier adaptability identifiers.

[0077] Subsequently, the system checks whether the Euclidean distance between the newly generated state node and its neighboring state nodes is less than a jump threshold, and verifies whether a structurally connected path exists. If the above conditions are met, a transition edge is added. This supplementation operation not only ensures the integrity and dynamic reachability of the graph structure, but also improves path diversity, providing richer candidate paths for keyframe reconstruction.

[0078] Step 4: Re-execute the path planning operation on the adjusted state transition probability graph, and add, delete, and correct the sequence of state nodes in the task key frame to adapt to the updated state transition logic.

[0079] Specifically, the system uses the start and end points of each task in the current task segment as anchor points, re-executes the path planning operation based on transition probability cost, and generates a new state node access sequence. Based on this, the system extracts task switching points again, updates the set of task keyframe state nodes, and performs synchronous correction operations on the original keyframe sequence, including deleting invalid keyframes, inserting new structural parts, and fine-tuning the task boundaries of adjacent nodes. This ensures that the reconstructed keyframe state node sequence matches the current graph structure, guaranteeing the logical consistency and stability of subsequent path task segmentation control.

[0080] The path correction module 400 is used to perform path backtracking and local reconstruction based on the pointer array after the state transition probability graph structure changes or the path is interrupted, and to update the key frame sequence and restore the path continuity.

[0081] Path backtracking and local reconstruction based on pointer arrays, with secondary updates to the keyframe sequence including:

[0082] When the path structure in the state transition probability graph is interrupted or changed, determine the range of affected subsequences in the original state node access sequence and identify the start and end state nodes.

[0083] It should be noted that during the execution of a firefighting robot's mission, the structure of the state transition probability graph is not static. Dynamic events such as sudden disasters, the appearance of obstacles, structural collapse, or path failure may cause some state nodes to fail or transition edges to be interrupted, resulting in some nodes on the original path losing effective connectivity. In order to maintain the integrity of the path structure and the continuity of mission execution, it is necessary to quickly locate the affected area and identify the damaged path segments after detecting changes in the graph structure.

[0084] Specifically, the system retrieves the previously recorded state node access sequences one by one, identifies currently unreachable state nodes through a node marking mechanism, and backtracks sequentially forward and backward to adjacent reachable state nodes. This determines the start and end nodes of the affected path subsequence, thereby delineating the local scope of the path interruption. This operation avoids reconstructing the entire path, thus reducing computational overhead, improving the response efficiency of task recovery, and ensuring that the navigation structure established between the original keyframes remains as stable as possible.

[0085] Secondly, based on the forward pointer array, backtracking from the terminating node of the affected subsequence, the shortest path subsequence to the starting point is reconstructed.

[0086] The core technical feature of this forward pointer array is its ability to trace the shortest cost path during path planning and to support dynamically adjusted backtracking operations. After determining the start and end nodes of the affected subsequence, the system uses the end node as the backtracking starting point and retrieves the predecessor node by accessing the nodes pointed to by its forward pointers level by level until it reaches the starting node position before the path breaks.

[0087] The path backtracking process does not require recalculating the overall cost matrix of the state transition probability graph, which significantly reduces the complexity of graph traversal and improves the path repair speed. The resulting new path subsequence not only satisfies the principle of minimizing transition probability cost, but also has structural connectivity and navigation feasibility, providing basic structural support for subsequent local reconstruction and keyframe recovery.

[0088] In the backtracking path, by combining the unaffected task keyframe state nodes in the original state node access sequence, local connections and task segment reorganization are performed to restore the continuity of the path structure.

[0089] It should be noted that during the navigation process of the firefighting robot, the task keyframe status nodes serve a dual purpose of path segmentation and task scheduling. Once the path is interrupted, if the keyframe sequence is not corrected in time, it will directly affect the logical control of the navigation plan and the effectiveness of segmented execution. Therefore, after generating the shortest path subsequence through path backtracking, the system aligns and verifies the start and end nodes of this subsequence with the unaffected keyframe status nodes in the original access sequence to identify the mapping relationship between the path breakpoint and the keyframe sequence.

[0090] If a broken path is found to span multiple task segments, the system will reconstruct the boundary nodes of the task segments and reorganize the task segment structure based on the continuity of task types and the node connectivity structure. During this process, the system inserts necessary intermediate state nodes through local join operations based on the structural relationship between the backtracking path and the original keyframes, ensuring that each task segment has a complete navigation path. This operation effectively restores the navigation continuity between keyframes, guaranteeing that each task segment still has an executable path even under changes in the graph structure.

[0091] Finally, the backtracked and repaired state node access sequence is output, and the task key frame state node sequence is updated a second time to support subsequent task segmentation navigation control operations.

[0092] Specifically, the system searches for state nodes in the backtracking path where the task type has changed, and scores these nodes based on the magnitude of the transition probability cost mutation and the spatial overlap of the structural intersection edges. Nodes with high scores are selected as new keyframes and inserted into the original keyframe sequence. Simultaneously, for nodes where path breaks cause keyframe failures, the system deletes them and adjusts the task segment boundaries of adjacent keyframes, completing a secondary update of the keyframe sequence.

[0093] The updated keyframe state node sequence provides the navigation control unit with the latest and most stable task segmentation basis, thus maintaining the accuracy of the fire robot's path control and the stability of task scheduling under drastic environmental changes, demonstrating the key role of this module in path repair and the system's resilience support capabilities.

[0094] This embodiment also provides a computer device applicable to the autonomous navigation and obstacle avoidance system of a fire-fighting robot, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the autonomous navigation and obstacle avoidance system of the fire-fighting robot as proposed in the above embodiment.

[0095] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0096] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the autonomous navigation and obstacle avoidance system for the fire-fighting robot as proposed in the above embodiments.

[0097] In summary, this invention effectively ensures the accuracy and feasibility of path planning by constructing a set of three-dimensional state units and eliminating impassable areas. It identifies task switching points and automatically divides path segments according to task type, enhancing the logical clarity and control precision of navigation tasks. Especially in scenarios of path interruption or sudden environmental changes, this system can achieve local path reconstruction and continuity restoration through graph structure updates and path correction modules, avoiding the delays and resource waste caused by global replanning in traditional systems. This invention also integrates spatial overlap calculation and jump detection methods, effectively improving the reliability of key frame node identification and the stability of task switching. In conclusion, this invention significantly improves the navigation adaptability and task execution continuity of firefighting robots in dynamic, high-risk environments, providing strong technical support for intelligent emergency operations in high-risk environments.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An autonomous navigation and obstacle avoidance system for a firefighting robot, characterized in that: include: The state graph construction and obstacle perception coding module is used to divide the work area and construct the state transition probability graph; The keyframe extraction module is used to calculate the optimal state path in the state transition probability graph based on the transition probability cost, and extract the task type switching point as the keyframe state node to realize the path task segmentation control. The graph structure update module is used to dynamically adjust the set of state nodes and edges in the state transition probability graph after obstacle detection or environmental changes, and reconstruct the keyframe sequence to adapt to the new structure. The path correction module is used to perform path backtracking and local reconstruction based on the pointer array after the state transition probability graph structure changes or the path is interrupted, and to update the key frame sequence and restore the path continuity. The construction of the state transition probability graph includes: Based on the building structure diagram of the fire scene, the measured laser point cloud data and the preset grid size, a set of state units in three-dimensional space is constructed, and the inaccessible areas and occluded elements are removed from each spatial unit. Based on the task type set and carrier set, a set of state nodes is generated by using a pre-built task-carrier adaptation table to filter state units that meet carrier constraints. In the set of state nodes, if any pair of state nodes satisfies that the Euclidean distance between the state nodes is less than the jump threshold and that the state units they belong to have a structural connectivity relationship, then a transition edge is added to the set of transition edges, and a state transition probability graph is constructed.

2. The autonomous navigation and obstacle avoidance system for the firefighting robot as described in claim 1, characterized in that: The keyframe extraction module includes: Based on the set of state nodes, each state node is classified and labeled according to the set of task types and the set of carriers, and state nodes representing the start and end points of the task are selected. In the state transition probability graph, the path planning operation is performed by selecting the state nodes at the starting point and the ending point as the two ends, and combining the transition probability cost of the transition edge, to obtain an ordered sequence of state node visits. In the state node access sequence, identify the state nodes corresponding to the task type change, and filter the state nodes with task segmentation features as the candidate task keyframe state node set, and divide them into continuous task segments. In the candidate task keyframe state node set, for each candidate task segment boundary node, identify the set of transition edge pairs formed by the transition edges between the previous node and the next node; for each transition edge pair, calculate the spatial overlap degree of the voxel sets corresponding to the two transition edges, where the spatial overlap degree is the ratio of the number of overlapping voxels to the total number of voxel unions; filter state nodes whose spatial overlap degree scores exceed the stability threshold and aggregate them into the final task keyframe state node sequence.

3. The autonomous navigation and obstacle avoidance system for the firefighting robot as described in claim 2, characterized in that: The path planning operation includes: In the state transition probability graph, the selected start state node and end state node are identified and recorded as boundary nodes for path planning. For all accessible paths between boundary nodes, construct a graph structure with state nodes as vertices and transition probability costs as edge weights, and use cost accumulation to calculate path costs; For each passable path, the negative logarithmic probability costs based on the transition edges are summed, and the state node access sequence corresponding to the path with the minimum total cost is recorded.

4. The autonomous navigation and obstacle avoidance system for the firefighting robot as described in claim 3, characterized in that: The filtering of state nodes with task segmentation characteristics includes: Based on the task type of each state node in the state node access sequence, the location where the task type changes is identified, and the corresponding state node is marked as a candidate task segment boundary node. For each candidate task segment boundary node, the task switching intensity is evaluated by combining the jump magnitude of the transition probability cost of the preceding and following state nodes, and state nodes with a change magnitude lower than the change threshold are filtered out. The remaining candidate task segment boundary nodes are aggregated into a set of candidate task keyframe state nodes, and continuous task segments between adjacent state nodes are divided according to their location. For candidate task keyframe state nodes, identify the set of adjacent transition edges formed by the preceding and following state nodes, and extract the transition edge pairs with cross-connection relationships. For each pair of transition edges, the ratio of the number of overlapping voxels to the total number of adjacent voxels in the spatial coordinate system is calculated and used as the spatial overlap score of the task switching point. The spatial overlap score is compared with the stability threshold, and state nodes with scores higher than the stability threshold are selected as high-confidence task key frame state nodes, forming a key frame state node sequence.

5. The autonomous navigation and obstacle avoidance system for the firefighting robot as described in claim 1, characterized in that: The set of state nodes and edges in the dynamically adjusted state transition probability graph includes: After acquiring information on environmental changes or obstacle detection results, the affected areas in the set of state units are located, and the set of affected state nodes is marked. Perform a validity check on the set of affected state nodes. If the state unit where the state node is located is occluded or not accessible, remove it from the set of affected state nodes and simultaneously remove the associated transition edges. For regions where new accessible paths emerge due to structural obstruction, state nodes are regenerated based on the connectivity of state units, and transfer edges are added according to the spatial topological relationship with surrounding nodes. For the adjusted state transition probability graph, the path planning operation is re-executed, and the sequence of state nodes in the task key frame is added, deleted, or corrected to adapt to the state transition logic after the structure update.

6. The autonomous navigation and obstacle avoidance system for the firefighting robot as described in claim 1, characterized in that: The path backtracking and local reconstruction based on pointer arrays, and the secondary update of the keyframe sequence, include: When the path structure in the state transition probability graph is interrupted or changed, determine the range of the affected subsequence in the original state node access sequence and identify the start and end state nodes. Based on the forward pointer array, backtrack from the terminal node of the affected subsequence to reconstruct the shortest path subsequence to the starting point; In the backtracking path, by combining the unaffected task key frame state nodes in the original state node access sequence, local connections and task segment reorganization are performed to restore the coherence of the path structure. Output the backtracked and repaired state node access sequence, and perform a secondary update on the task key frame state node sequence to support subsequent task segmentation navigation control operations.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the autonomous navigation and obstacle avoidance system of the fire-fighting robot according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the autonomous navigation and obstacle avoidance system of the fire-fighting robot according to any one of claims 1 to 6.

Images