Dynamic environment-oriented intelligent mobile robot navigation method

Abstract

The invention discloses an intelligent mobile robot navigation method for a dynamic environment, and relates to the technical field of computer processing, and the method comprises the following steps: S01, collecting the dynamic environment of a robot in real time through a laser radar, an RGB-D camera and a UWB module, building a robot real-time environment perception model, and carrying out the real-time environment perception of the robot; the robot real-time environment sensing model comprises parameterized description of metal reflection, dynamic obstacles and narrow terrains; s02, determining a plurality of key design parameters of navigation of the intelligent mobile robot; and S03, based on the plurality of key design parameters, constructing an environment perception accuracy objective function of the intelligent mobile robot and a real-time response and scene understanding depth objective function. According to the invention, high-precision environment perception, low-delay path re-planning and safe and efficient navigation are synchronously realized in a dynamic environment through metal reflection interference filtering, dynamic obstacle acceleration pre-judgment and narrow terrain adaptive passing strategies.

Description

Technical Field

[0001] This invention relates to the field of computer processing technology, and in particular to a navigation method for intelligent mobile robots in dynamic environments. Background Technology

[0002] With the widespread adoption and deepening application of robotic devices in daily life and industrial scenarios, the technological shortcomings of existing navigation systems are becoming increasingly prominent, posing a key bottleneck to improving robot efficiency and intelligence. Currently, the industry generally uses ranging radar, ultrasonic sensors, and satellite navigation equipment (such as GNSS) to build the basic architecture of robot navigation systems. However, these technical solutions reveal a series of significant defects in actual complex operating environments, especially in highly dynamic scenarios: In industrial environments with strong reflectivity (such as automotive welding workshops where metal reflectivity exceeds 80%), traditional laser SLAM (Simultaneous Localization and Mapping) technology is severely interfered with. Its positioning error significantly exceeds the critical threshold of 50mm, resulting in a substantial decrease in the robot's absolute positioning accuracy, directly affecting the execution quality of key processes such as precision assembly and welding. Especially when facing complex scenarios with limited space, such as narrow passages of 0.8 meters, traditional local path planning algorithms such as Dynamic Window (DWA) fail to effectively integrate semantic information of the scene (such as passage boundary attributes and accessibility judgments), relying solely on geometric obstacle avoidance, leading to rigid robot path planning and frequent decision-making errors. The actual robot pass rate was as low as 68%, which is far from meeting the stringent requirements of precision manufacturing environments for operational safety and traffic efficiency.

[0003] In summary, existing navigation solutions based on mainstream sensors suffer from serious deficiencies in the accuracy, real-time performance, and depth of scene understanding when dealing with interference from metallic reflections, sudden dynamic obstacles, and narrow, complex terrain. This not only directly increases operational risks and safety costs but also fundamentally limits the reliable deployment and effectiveness of robots in a wider range of complex scenarios. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention provides a navigation method for intelligent mobile robots in dynamic environments, the method comprising the following steps: S01. The robot's dynamic environment is collected in real time by using LiDAR, RGB-D camera and UWB module to establish a real-time environmental perception model for the robot. The real-time environmental perception model for the robot includes parametric descriptions of metal reflection, dynamic obstacles and narrow terrain. S02. Determine several key design parameters for the navigation of the intelligent mobile robot; S03. Based on the aforementioned key design parameters, construct the objective function for the accuracy of environmental perception of the intelligent mobile robot, and the objective function for the depth of real-time response and scene understanding; S04. Combining the objective function of environmental perception accuracy with the objective function of real-time response and scene understanding depth, an improved genetic algorithm is used for multi-objective optimization to achieve a Pato'e optimal solution set that balances the two functions; S05. Using the key design parameters corresponding to the Patolei optimal solution set, generate the robot's dynamic movement path.

[0005] Preferably, the robot's real-time environment perception model in step S01 includes: S11. When the lidar acquires the cloud reflectivity R of the metal reflection along the robot's predetermined path, when R ≥ R metal When this occurs, the reflection interference filtering algorithm is triggered to remove falsely detected obstacles in high-reflectivity areas, where R metal The reflectivity threshold; S12. Obtain the coordinates of the dynamic obstacle in the three-dimensional space of the robot's real-time environment perception model, and divide the multiple coordinates obtained over a continuous predetermined time interval into several data segments A according to a window. i The distance d between two adjacent coordinate points is obtained. Subtracting the two adjacent distances d, summing the results, and dividing by the number of coordinate points, yields data segment A. i The average velocity V under i平 Finally, change the V of the two adjacent windows. i平 Subtracting the values ​​gives the average difference, and then dividing the average difference by the total duration of the two adjacent windows gives [A]. i A i-1 The acceleration of i is then summed, and the velocity change rate a is obtained by multiplying i by 100%.

[0006] S13, Define the width threshold W narrow When the spacing between consecutive point clouds is less than W narrow At that time, a three-dimensional geometric constraint model of the channel is constructed, and the output of the three-dimensional geometric constraint model of the channel is an environmental feature vector E=[R metal Rate of change of velocity a, W narrow (x, y, z)).

[0007] Preferably, the determination of several key design parameters for intelligent mobile robot navigation in step S02 includes: S21. The robot's current coordinates are the starting point of the original coordinates in the three-dimensional space of the robot's real-time environmental perception model; S22. Based on the environmental feature vector E, determine the coordinate set of the dynamic obstacle, metal reflector, and narrow terrain in the three-dimensional space of the above real-time environmental perception model; S23. Based on the obtained speed change rate 'a' of the dynamic obstacle and the current travel speed of the robot, calculate the time it takes for the dynamic obstacle to reach the trajectory Q and the time it takes for the robot to reach the trajectory Q, in order to determine whether a negotiation has occurred. If a negotiation has occurred, reduce the time it takes to reach the trajectory Q by a predetermined time threshold [-5s, -7s] and calculate the robot's average moving speed.

[0008] Preferably, the objective function for environmental perception accuracy in step S03 is defined as: ; in, This refers to the positioning error of the metal reflection area. The dynamic obstacle position prediction error is determined by the average Euclidean distance between the actual trajectory and the predicted trajectory. Let a be the rate of change of velocity. This is due to measurement error in narrow terrain.

[0009] Preferably, the objective function for real-time response and scene understanding in step S03 is defined as: ; in, The time delay from environmental change to path replanning is measured using system clock stamps. For scene semantic understanding depth, ,and .

[0010] Preferably, step S04, which involves using an improved genetic algorithm for multi-objective optimization to achieve a balanced Patolei optimal solution set for the two functions, includes: S41. Determine the parameter search boundary for multi-objective optimization based on the physical constraints of sensor fusion weights, dynamic obstacle prediction time, narrow passage passing strategy coefficients, environmental update frequency, and path planning iteration number. S42. During the iteration process, individuals with the highest overall ranking in the Pareto front are retained as elite solutions, and a mutation rate strategy that dynamically decays with the iteration process is adopted. S43. Generate a Pareto optimal solution set that balances environmental perception accuracy, real-time response, and scene understanding depth through multi-generational genetic operations. S44. Based on the actual application scenario requirements, select optimized design parameters from the Pareto optimal solution set that are suitable for industrial scenarios or complex terrain scenarios.

[0011] Preferably, the dynamic movement path of the robot generated in step S05 includes: S51. Combine the sensor fusion weights Wsensor and dynamic obstacle prediction time T0 from the Pareto optimal solution set. p Narrow passage strategy coefficient k narrow Input to the route planner; S52, according to T p Predict the trajectory of dynamic obstacles and generate a boundary path of the avoidance buffer zone that bypasses the trajectory; S53, Channel width W based on real-time detection current sum coefficient k narrow Calculate the condition that r=k narrow ×R robot The arc path, where R robot The radius of the robot body; S54. When the ambient reflectance exceeds the threshold R metal At that time, W was adopted. sensor The adjusted sensor data generates a conservative path, reducing the moving speed and shortening the spacing between path points; S55. Path refresh is triggered at the environment update frequency of the Pareto solution set, and Bézier curves are used to connect path nodes to ensure turning continuity.

[0012] Preferably, the generation of the avoidance buffer boundary path in step S52 needs to satisfy the safety constraint that the real-time distance between the path point and the obstacle is ≥ v. rel ×T p , where v rel This represents the robot's speed relative to the obstacle.

[0013] The present invention has at least the following beneficial effects: 1. By establishing a real-time perception model that integrates metal reflectivity threshold, dynamic obstacle acceleration quantification model and narrow terrain geometric constraints, false obstacles in high reflectivity areas are effectively filtered out, obstacle velocity change rate and terrain features are accurately calculated, and the positioning distortion problem caused by metal reflection interference is solved.

[0014] 2. Based on the velocity change rate and time negotiation mechanism, the risk of intersection between dynamic obstacles and robot path is predicted 5 to 7 seconds in advance. Combined with the prediction time optimized by the improved genetic algorithm, a buffer path that meets the safety distance is generated, which fundamentally avoids the risk of sudden collision and improves the safety of dynamic scenes.

[0015] 3. Real-time detection of width using a three-dimensional geometric constraint model of the channel, through the coefficient k narrow Dynamically adjusting the curvature of the arc path allows the robot to autonomously adjust its travel strategy based on terrain constraints, especially when the width is below a threshold W. narrowAchieve continuous passage with zero interruptions in complex channels.

[0016] 4. By optimizing the balance between environmental perception accuracy, real-time response, and scene understanding depth through multi-objective optimization, an improved genetic algorithm with dynamic decay mutation rate is adopted to simultaneously reduce path replanning delay and improve semantic understanding indicators at a 10-millisecond environmental update frequency, ensuring that the robot has both rapid response capability and high-order scene cognition capability in dynamic environments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in 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.

[0018] Figure 1 This is a flowchart of a navigation method for an intelligent mobile robot in a dynamic environment, as provided in Embodiment 1 of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0021] Example 1

[0022] This embodiment provides a navigation method for intelligent mobile robots in dynamic environments. The method includes the following steps: Figure 1 As shown: S01. Real-time acquisition of the robot's dynamic environment is achieved through LiDAR, RGB-D camera and UWB module, and a real-time environmental perception model of the robot is established. The real-time environmental perception model of the robot includes parametric descriptions of metal reflection, dynamic obstacles and narrow terrain. Specifically, the aforementioned LiDAR provides high-precision, high-resolution distance information (point cloud) and reflection intensity information (crucial). This is the core sensor for perceiving geometric structures (obstacle location, shape, and narrow passage dimensions) and identifying metallic reflection interference. The RGB-D camera provides color images and corresponding depth maps. This enhances object recognition (distinguishing dynamic obstacle types such as people and vehicles), provides texture information to aid semantic understanding, and supplements point cloud details at close range (especially in areas where the LiDAR may be obstructed or have low reflection). Secondly, the UWB module in this embodiment provides the robot's high-precision absolute position (especially in indoor / complex scenes where GPS is denied or signals are unstable). This is crucial for establishing a globally consistent environmental map, robot self-localization, and subsequent dynamic obstacle trajectory prediction calculations in the global coordinate system. This step, through sensor calibration (extrinsic and intrinsic parameters) and time synchronization, unifies data from different sensors (point cloud, depth map, image, position) into a single three-dimensional spatial coordinate system (typically the robot's body coordinate system or the global map coordinate system). Location information is fused using filtering algorithms (such as Kalman filtering or its variants), and perceptual information is fused using point cloud registration and image processing techniques to form a unified, dense environmental representation with semantic / reflective attributes.

[0023] Furthermore, the robot's real-time environmental perception model in the above steps includes: S11. When the lidar acquires the cloud reflectivity R of the metal reflection along the robot's predetermined path, when R ≥ R metal When this occurs, the reflection interference filtering algorithm is triggered to remove falsely detected obstacles in high-reflectivity areas, where R metal The reflectivity threshold; S12. Obtain the coordinates of the current dynamic obstacle in the 3D space of the robot's real-time environment perception model, and divide the multiple coordinates obtained over a continuous predetermined time interval into several data segments A according to a window. i The distance d between two adjacent coordinate points is obtained. Subtracting the two adjacent distances d, summing the results, and dividing by the number of coordinate points, yields data segment A. i The average velocity V under i平 Finally, change the V of the two adjacent windows. i平 Subtracting the values ​​gives the average difference, and then dividing the average difference by the total duration of the two adjacent windows gives [A]. i A i-1The acceleration of i is then summed, and the velocity change rate a is obtained by multiplying i by 100%.

[0024] S13, Define the width threshold W narrow When the spacing between consecutive point clouds is less than W narrow At that time, a three-dimensional geometric constraint model of the channel is constructed, and the output of the three-dimensional geometric constraint model of the channel is the environmental feature vector E=[R metal Rate of change of velocity a, W narrow (x, y, z)).

[0025] Specifically, step S11 handles metal reflection interference. Because highly reflective surfaces like metal reflect abnormally strong lidar signals, this can cause "false obstacles" (such as light spots on a wall being mistaken for obstacles) or distortion and positional drift of real metal objects in the point cloud. The lidar outputs the spatial coordinates (x, y, z) and reflectivity R of each point in real time. Assume an empirically / calibrated reflectivity threshold R is set. metal (For example, a reflectivity value much higher than that of ordinary walls and floors). Point-by-point analysis is performed on the point cloud scanned by the lidar. When the reflectivity R of a detected point or area is ≥ R... metal At that time, the reflection interference filtering algorithm is triggered.

[0026] Step S12 quantifies the dynamic obstacle behavior, as simply detecting the obstacle's position is insufficient, especially in dynamic environments. Understanding its motion state (velocity, acceleration, and motion trend) is necessary to predict its future position and effectively avoid it. The intensity or trend of its motion is quantified by analyzing continuous time-series position data and calculating its velocity change rate 'a'. Dynamic obstacles (such as pedestrians and vehicles) are detected and tracked in real-time using fused perception data (primarily LiDAR / RGB-D point clouds), acquiring their coordinate point sequences in a unified 3D coordinate system. The continuously acquired coordinate points are divided into several data segments A according to fixed time windows (e.g., data points every 0.5 seconds). i (i=1,2,3...). For each data segment A i The processing is as follows: Calculate the distance d (Euclidean distance) between any two adjacent coordinate points within the segment. Then calculate the difference Δd between adjacent distances d (i.e., the change in distance from one point to the previous point). Sum all Δd values ​​within the segment and then divide by the number of coordinate points N in the segment. i This yields the average change in velocity over that time period. In fact, this result more accurately reflects the magnitude of the average acceleration over that time period (because Δd / Δt ≈ acceleration, implicitly including the time interval Δt). We can call it the average velocity V within the segment. i平 Calculate the values ​​of two adjacent data segments (e.g., A). i and Ai-1 V i平 and V i平-1 The difference ΔV between i平 =VV i平 -V i平-1 This difference ΔV i平 Divide by the total time span T of the two data segments window (For example, 1 second, if each segment is 0.5 seconds), obtain the average acceleration change 'a' between these two segments. {i,i-1} =ΔV i平 / T window This reflects whether the obstacle accelerated, decelerated, or moved at a constant speed between the two most recent time windows. Finally, a comprehensive rate of change of velocity 'a' is calculated: this is calculated by taking the current and several previous (e.g., the most recent 3-5) 'a' values. {i,i-1} Add them together, then divide by the number of {i, i-1} involved in the calculation, M, and then multiply by 100%. The value of a directly reflects the degree of drastic change in the recent motion state of the obstacle (a large absolute value indicates a rapid acceleration / deceleration and a fast change; a value close to 0 indicates a constant speed or a gradual change).

[0027] Furthermore, step S13 involves modeling the geometric constraints of narrow terrain, as narrow passages (such as doorways and aisles between shelves) pose a challenge to the robot's mobility. Accurate perception of the effective passage width and its three-dimensional geometry (e.g., whether it is tilted or has protrusions) is needed to determine whether passage is possible and how to plan a safe and smooth path. Based on the fused point cloud data (especially LiDAR), the point cloud distribution in front of and around the robot is analyzed. A width threshold W is defined. narrow This threshold is typically slightly larger than the robot's physical width (or diameter), with a safety margin. In the robot's forward direction, the distance between consecutive point clouds is scanned and analyzed (this can be understood as scanning "virtual tangents" across multiple height layers). When a consecutive point cloud spacing < W is detected... narrow When identifying a region as a potential narrow passage, a 3D geometric constraint model is constructed for that region: the point cloud is analyzed to fit the boundaries (which may be planes, surfaces, or point sets) on both sides of the passage. The minimum effective width of the passage at different heights and locations is calculated. The orientation of the passage (whether it is curved), ground flatness, and top height restrictions are analyzed. Ultimately, the core output of this model is the key parameter describing the narrow region: the minimum effective width value (usually the detected value less than W). narrow This involves parameterizing and quantizing complex geometric constraints, including the value of W (or its statistical value) and its spatial location information (x, y, z) (such as the coordinates of the channel entrance center point or a critical bottleneck point). The output is a clear W. narrowThe value tells the path planner "how narrow it is here," and the location information (x, y, z) tells the planner "where the bottleneck is." This directly serves the subsequent adaptive path generation in S53 (calculating the path while satisfying r=k). narrow *R robot The curved path allows the robot to precisely adjust its passage strategy (such as deceleration, posture adjustment, and selection of the optimal path point) according to the actual terrain, enabling it to safely and smoothly traverse narrow spaces and avoid getting stuck or colliding.

[0028] S02. Determine several key design parameters for the navigation of the intelligent mobile robot. The steps are as follows: S21. The robot's current coordinates are the starting point of the original coordinates in the three-dimensional space of the robot's real-time environmental perception model. S22. Based on the environmental feature vector E, determine the set of coordinates of the current dynamic obstacles, metallic reflectors, and narrow terrain in the three-dimensional space of the above real-time environmental perception model; S23. Based on the obtained velocity change rate 'a' of the dynamic obstacle and the current robot speed, calculate the time it takes for the dynamic obstacle to reach the path trajectory Q and the time it takes for the robot to reach the path trajectory Q to determine whether a negotiation has occurred. If a negotiation has occurred, reduce the time to reach the path trajectory Q by a predetermined time threshold [-5s, -7s] and calculate the robot's average moving speed.

[0029] Specifically, the robot's current pose (position + orientation) is used as the origin (0,0,0) in three-dimensional space, and the LiDAR / UWB positioning data is transformed to this coordinate system. For example, the UWB outputs global coordinates (Xg,Yg,Zg), which are converted into local coordinates (0,0,0) with the robot's current position as the origin through a coordinate transformation matrix, ensuring that the positions of all obstacles and path trajectories are calculated based on this benchmark. Then, the abstract environmental feature vector E=[R metal Rate of change of velocity a, W narrow [x, y, z] is resolved into a specific set of spatial coordinates. The three-dimensional coordinates of the obstacle corresponding to the velocity change rate 'a' in E are extracted to predict the dynamic obstacle's trajectory. Based on the reflectivity threshold R... metal Locate the coordinates of the high-reflectivity area to identify the metallic reflective object.

[0030] The dynamic obstacle negotiation time and speed control predicts path conflicts between the robot and obstacles through spatiotemporal negotiation analysis and dynamically adjusts the robot speed based on the risk level. Trajectory intersection prediction: Suppose the robot is currently moving along path Q (e.g., a straight line from (0,0,0) to (10,0,0)). Extrapolate its trajectory based on the obstacle coordinates and the rate of change of velocity a (e.g., a worker moves from (2.3,-1.1,0) at 1.5m / s in the positive X-axis direction). Calculate the intersection point P of the two trajectories (e.g., point (3.5,0,0) on Q). Time when the dynamic obstacle reaches point P: T_obs = distance (P, current position of obstacle) / current speed of obstacle (e.g., distance from worker to point P is 1.2m ÷ 1.5m / s = 0.8s) Time when the robot reaches point P: T_rob = distance (P, current position of robot) / current speed of robot (e.g., distance from robot to point P is 3.5m ÷ 1.2m / s ≈ 2.92s) Risk judgment and time buffer: If |T obs -T rob | < Time tolerance threshold (e.g., 1.5s) → classified as negotiation risk. High-risk handling strategy: When obstacle a > 0.6 (rapid acceleration), correct the robot's arrival time to: T rob =T rob -ΔT (where ΔT∈[-7s,-5s], selected by linear interpolation of the value of a) (Example: a=0.75 → take ΔT=-6.5s → T rob =2.92-6.5=-3.58s). A negative value means the robot needs to pass point P 3.58 seconds before the obstacle arrives. Based on the new time constraint T_rob', the required average speed is calculated backwards: V new = Distance(P, Robot's current position) / |T rob (Example: 3.5m ÷ 3.58s ≈ 0.98m / s, which is 18% slower than the original speed of 1.2m / s).

[0031] S03. Based on multiple key design parameters, construct the objective function for the environmental perception accuracy of the intelligent mobile robot, as well as the objective function for real-time response and scene understanding depth. Specifically: The objective function for environmental perception accuracy (which quantifies metal reflection positioning error, dynamic obstacle prediction error, and terrain measurement error into a unified perception accuracy index) is defined as follows: ; in, This refers to the positioning error of the metal reflection area. The dynamic obstacle position prediction error is determined by the average Euclidean distance between the actual trajectory and the predicted trajectory. Let a be the rate of change of velocity. This is due to measurement error in narrow terrain.

[0032] The objective function for real-time response and scene understanding (the time from environmental change to generating a new path) is defined as follows: ; in, The time delay from environmental change to path replanning is measured using system clock stamps. For scene semantic understanding depth, ,and .

[0033] S04. Combining the objective functions of environmental perception accuracy and real-time response and scene understanding depth, an improved genetic algorithm is used for multi-objective optimization to achieve a Pato's optimal solution set that balances the two functions. This includes the following steps: S41. Determine the parameter search boundary for multi-objective optimization based on the physical constraints of sensor fusion weights, dynamic obstacle prediction time, narrow passage passing strategy coefficients, environmental update frequency, and path planning iteration number. S42. During the iteration process, individuals with the highest overall ranking in the Pareto front are retained as elite solutions, and a mutation rate strategy that dynamically decays with the iteration process is adopted. S43. Generate a Pareto optimal solution set that balances environmental perception accuracy, real-time response, and scene understanding depth through multi-generational genetic operations. S44. Based on the actual application scenario requirements, select optimized design parameters from the Pareto optimal solution set that are suitable for industrial scenarios or complex terrain scenarios.

[0034] Specifically, the aforementioned improvement in environmental perception accuracy using an enhanced genetic algorithm... With real-time response capability The process involves finding the Pareto optimal equilibrium point and generating a set of optimal parameter configurations to adapt to different dynamic environments, providing a scientific basis for path planning. Based on the robot's physical limitations (such as minimum turning radius) and sensor capabilities, reasonable adjustment ranges are set for these "knobs." For example, k... narrow It cannot be less than 0.6, otherwise the robot simply won't be able to turn; T p The prediction time cannot exceed 8 seconds, as predictions that are too far in the future are meaningless. This ensures that the solutions found later are practically usable. Specifically, setting parameters within a reasonable range (e.g., obstacle prediction time 1-8 seconds) ensures that the solution set conforms to physical constraints. The optimal solution is retained during algorithm iterations, and the intensity of random perturbations is reduced as the algorithm progresses, accelerating convergence. This transforms experience-based navigation decisions into a mathematical optimization problem. Through dynamic parameter configuration, the robot acquires the ability to "recognize environmental risks and autonomously adjust its behavior," fundamentally solving the industry challenge of balancing safety and efficiency in dynamic environments.

[0035] S05. Using the key design parameters corresponding to the Patolei optimal solution set, generate the robot's dynamic movement path, which specifically includes the following steps: S51. Combine the sensor fusion weights Wsensor and dynamic obstacle prediction time T0 from the Pareto optimal solution set. p Narrow passage strategy coefficient k narrow Input to the route planner; S52, according to T p Predict the movement trajectory of dynamic obstacles and generate the boundary path of the avoidance buffer zone for the detour trajectory; S53, Channel width W based on real-time detection current sum coefficient k narrow Calculate the condition that r=k narrow ×R robot The arc path, where R robot The radius of the robot body; S54. When the ambient reflectance exceeds the threshold R metal At that time, W was adopted. sensor The adjusted sensor data generates a conservative path, reducing the moving speed and shortening the spacing between path points; S55. Path refresh is triggered at the environment update frequency of the Pareto solution set, and Bézier curves are used to connect path nodes to ensure turning continuity.

[0036] In step S52 above, the generation of the avoidance buffer boundary path must satisfy the safety constraint that the real-time distance between the path point and the obstacle is ≥ v. rel ×T p , where v rel This represents the robot's speed relative to the obstacle.

[0037] Specifically, Sensor fusion weight W sensor : Control the trust level of LiDAR / RGB-D data (e.g., W) sensor When the value is 0.8, the weight of the lidar accounts for 80%.

[0038] Dynamic obstacle prediction time T p : Determine the avoidance distance (safe buffer length = v) rel ×T p ).

[0039] Narrow channel coefficient k narrow Calculate the minimum turning radius r=k narrow ×R robot .

[0040] Predicting the future T of obstacles pThe safe distance D is extended outward from the center line of a movement path (such as a forklift traveling in a straight line from point A to point B). safe =v rel ×T p T p The prediction time selected for the current scene. Generate a buffer boundary path surrounding the centerline; the robot path must be strictly limited to outside the buffer. When encountering curved paths in narrow terrain, the real-time channel width W can be used as a reference. current Dynamically adjust the turning radius to calculate the maximum turning radius the robot can achieve. ; When the reflectivity R>R is detected metal (e.g., R=0.95 in the welded area), W is used. sensor When generating paths using weighted fusion data (such as reducing the weight of LiDAR to 70% and increasing the weight of UWB positioning): shorten the spacing between path points, reduce the speed by 50% to extend the environmental response time, and move away from highly reflective objects.

[0041] This embodiment effectively filters out falsely detected obstacles in high-reflectivity areas by establishing a real-time perception model that integrates a metal reflectivity threshold, a dynamic obstacle acceleration quantification model, and the geometric constraints of narrow terrain. It accurately calculates the obstacle velocity change rate and terrain features, thus solving the positioning distortion problem caused by metal reflection interference. Based on the velocity change rate and time negotiation mechanism, the risk of intersection between dynamic obstacles and the robot path is predicted 5-7 seconds in advance. Combined with an improved genetic algorithm to optimize the prediction time, a safe avoidance buffer path is generated, fundamentally avoiding the risk of sudden collisions and improving the safety of dynamic scenes. Furthermore, the width is detected in real time using a three-dimensional geometric constraint model of the channel, and the coefficient k is used to... narrow Dynamically adjusting the curvature of the arc path allows the robot to autonomously adjust its travel strategy based on terrain constraints, especially when the width is below a threshold W. narrow Achieve continuous passage with zero interruptions in complex channels.

[0042] Secondly, by optimizing the balance between environmental perception accuracy, real-time response, and scene understanding depth through multi-objective optimization, an improved genetic algorithm with dynamic decay mutation rate is adopted to simultaneously reduce path replanning delay and improve semantic understanding indicators at a 10-millisecond environmental update frequency, ensuring that the robot has both rapid response capability and high-order scene cognition capability in dynamic environments.

[0043] Example 2

[0044] This invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the following steps: The robot's dynamic environment is collected in real time by LiDAR, RGB-D camera and UWB module to establish a real-time environmental perception model for the robot. The real-time environmental perception model for the robot includes parametric descriptions of metal reflection, dynamic obstacles and narrow terrain. Determine several key design parameters for navigation in intelligent mobile robots; Based on several key design parameters, we construct an objective function for the accuracy of environmental perception and an objective function for the depth of real-time response and scene understanding of the intelligent mobile robot. By combining the objective functions of environmental perception accuracy and real-time response and scene understanding depth, an improved genetic algorithm is used for multi-objective optimization to achieve a Pato'e optimal solution set that balances the two functions. By utilizing the key design parameters corresponding to the Patolei optimal solution set, the dynamic movement path of the robot is generated.

[0045] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0046] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0047] Example 3

[0048] This invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program segment, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the following steps: The robot's dynamic environment is collected in real time by LiDAR, RGB-D camera and UWB module to establish a real-time environmental perception model for the robot. The real-time environmental perception model for the robot includes parametric descriptions of metal reflection, dynamic obstacles and narrow terrain. Determine several key design parameters for navigation in intelligent mobile robots; Based on several key design parameters, we construct an objective function for the accuracy of environmental perception and an objective function for the depth of real-time response and scene understanding of the intelligent mobile robot. By combining the objective functions of environmental perception accuracy and real-time response and scene understanding depth, an improved genetic algorithm is used for multi-objective optimization to achieve a Pato'e optimal solution set that balances the two functions. By utilizing the key design parameters corresponding to the Patolei optimal solution set, the dynamic movement path of the robot is generated.

[0049] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A dynamic environment oriented intelligent mobile robot navigation method, characterized by, The method comprises the following steps: S01, collecting the dynamic environment of the robot in real time through a laser radar, an RGB-D camera and a UWB module, establishing a real-time environment perception model of the robot, the real-time environment perception model of the robot comprising a description of metal reflection, dynamic obstacles and narrow terrain and parameters; S02, determining a plurality of key design parameters of intelligent mobile robot navigation; S03, based on a plurality of the key design parameters, constructing an environment perception accuracy objective function of the intelligent mobile robot, and a real-time response and scene understanding depth objective function; S04, comprehensively combining the environment perception accuracy objective function and the real-time response and scene understanding depth objective function, performing multi-objective optimization by using an improved genetic algorithm, and achieving a Pareto optimal solution set balancing the two functions; S05, generating a dynamic moving path of the robot by using the key design parameters corresponding to the Pareto optimal solution set. 2.The dynamic environment oriented intelligent mobile robot navigation method according to claim 1, wherein, The real-time environment perception model of the robot in the step S01 comprises: S11、when the laser radar acquires the cloud reflectivity R of the metal reflection on the predetermined form route of the robot, when R≥R metal , trigger the reflection interference filtering algorithm to eliminate the false detection obstacles in the high reflection area, wherein R metal is the reflectivity threshold; S12. Obtain the coordinates of the dynamic obstacle in the three-dimensional space of the robot's real-time environment perception model, and divide the multiple coordinates obtained over a continuous predetermined time interval into several data segments A according to a window. i The distance d between two adjacent coordinate points is obtained. Subtracting the two adjacent distances d, summing the results, and dividing by the number of coordinate points, yields data segment A. i The average velocity V under i平 Finally, change the V of the two adjacent windows. i平 Subtracting the values ​​gives the average difference, and then dividing the average difference by the total duration of the two adjacent windows gives [A]. i A i-1 The acceleration of ] is then added together, and the velocity change rate a is obtained by multiplying i by 100%. S13, define width threshold W narrow When the continuous point cloud spacing < W narrow , construct a channel three-dimensional geometric constraint model, and the channel three-dimensional geometric constraint model outputs an environmental feature vector E = [R metal , velocity change rate a, W narrow , (x, y, z)]. 3.The dynamic environment oriented intelligent mobile robot navigation method according to claim 1, wherein, The plurality of key design parameters of intelligent mobile robot navigation in the step S02 comprise: S21, taking the current coordinates of the robot as the starting point of the original coordinates in the three-dimensional space of the real-time environment perception model of the robot; S22, determining the coordinates set of the current dynamic obstacles, metal reflectors and narrow terrain in the three-dimensional space of the real-time environment perception model based on the environment feature vector E; S23, based on the speed change rate a of the dynamic obstacle obtained, and based on the current driving speed of the robot, calculating the time of the dynamic obstacle reaching the route track Q and the time of the robot reaching the route track Q to determine whether an intersection occurs, and if an intersection occurs, reducing a predetermined time threshold [-5s, -7s] from the time of reaching the route track Q to calculate the average moving speed of the robot. 4.The dynamic environment oriented intelligent mobile robot navigation method according to claim 1, wherein, The environment perception accuracy objective function in the step S03 is defined as: ； wherein, is a positioning error of the metal reflective region, is a dynamic obstacle position prediction error determined by the Euclidean distance mean of the actual trajectory and the predicted trajectory, is a speed change rate a, is a narrow terrain width measurement error. 5.The dynamic environment oriented intelligent mobile robot navigation method according to claim 1, wherein, The real-time response and scene understanding depth objective function in the step S03 is defined as: ； wherein, is the time delay for environmental change to path replanning, measured by system clock ticks, is the scene semantic understanding depth, , and . 6.The dynamic environment oriented intelligent mobile robot navigation method according to claim 1, wherein, The multi-objective optimization by using the improved genetic algorithm in the step S04 to achieve a Pareto optimal solution set balancing the two functions comprises: S41, determining the parameter search boundary of multi-objective optimization according to the physical constraint range of the sensor fusion weight, the dynamic obstacle prediction time, the narrow channel passing strategy coefficient, the environment update frequency and the path planning iteration number; S42, reserving the individuals with a comprehensive ranking in the Pareto front as elite solutions in the iteration process, and adopting a mutation rate strategy that dynamically decays with the iteration process; S43, generating a Pareto optimal solution set balancing the environment perception accuracy, the real-time response and the scene understanding depth through multi-generation genetic operation; S44, selecting the optimization design parameters suitable for the industrial scene or the complex terrain scene from the Pareto optimal solution set according to the actual application scene requirements. 7.The intelligent mobile robot navigation method for dynamic environment, according to claim 1, wherein, The generation of the dynamic moving path of the robot in the step S05 comprises: S51, the sensor fusion weight Wsensor in the pareto optimal solution set, the dynamic obstacle prediction time T p , the narrow passage passing strategy coefficient k narrow input to the path planner; S52, according to T p predict a motion trajectory of a dynamic obstacle, generate a detour buffer boundary path around the trajectory; S53, the channel width W based on real-time detection current and a coefficient k narrow , calculate an arc path satisfying r=k narrow x R robot , where R robot is the robot body radius; S54, when the environmental reflectivity exceeds a threshold R metal W sensor The adjusted sensor data generates a conservative path, reduces the moving speed and shortens the path point spacing; S55, triggering path refreshing with the environment update frequency of the Pareto solution set, connecting the path nodes by using a Bezier curve to ensure the continuity of turning. 8.The dynamic environment oriented intelligent mobile robot navigation method according to claim 7, wherein, The generation of the avoidance buffer boundary path in the step S52 needs to meet the safety constraint condition that the real-time distance between the path point and the obstacle is greater than v rel × T p , wherein v rel is the speed of the robot relative to the obstacle. 9.A non-transitory computer-readable storage medium having stored therein at least one instruction or at least one piece of program, characterized in that, The at least one instruction or the at least one program is loaded and executed by the processor to implement the intelligent mobile robot navigation method for dynamic environment as claimed in any one of claims 1-8.

10. An electronic device, comprising: The processor and the memory are included, and the memory has at least one instruction or at least one program stored therein, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the intelligent mobile robot navigation method for dynamic environment as claimed in any one of claims 1-8.

Images