A path intelligent navigation system for photovoltaic installation robots

Abstract

The application provides a path intelligent navigation system for a photovoltaic installation robot, and relates to the field of path navigation. The system comprises an environment perception module, a path planning module, a navigation control module, a communication unit and a data storage module. The environment perception module generates an environment map containing slope, material friction coefficient and obstacles through the fusion of laser radar, camera and ultrasonic sensor. The path planning module integrates the environment map, robot position and task state, uses an algorithm to generate an optimal path and dynamically adjusts energy consumption weight and energy quota. The navigation control module uses a fuzzy algorithm to control movement, splits emergency tasks through task mirror fractal technology and monitors in real time. The communication unit synchronizes data with a warehouse device and a mechanical arm system. The data storage module records historical parameters and optimizes algorithms through machine learning. The system balances energy consumption, time and safety targets through a multi-target game equalizer, corrects paths and inserts a pseudo-path optimization model through a dynamic coupling unit, and realizes intelligent navigation and adaptive path planning of the photovoltaic installation robot.

Description

Technical Field

[0001] The present invention relates to the field of path navigation, and in particular to a path intelligent navigation system for a photovoltaic installation robot. Background Art

[0002] As the global energy structure transformation accelerates, the scale and complexity of photovoltaic power station construction have increased significantly. Traditional manual installation methods are limited by the changing terrain, tight construction schedules and the risks of high-altitude operations, making it difficult to meet the needs of large-scale deployment. Automated installation robots have become an important development direction for the industry, but their autonomous navigation systems need to overcome core challenges such as dynamic environment perception, multi-target path optimization and task chain collaboration to achieve efficient and safe photovoltaic panel installation operations.

[0003] Existing photovoltaic robot navigation technologies mostly rely on a single sensor to build a static environment model and use a fixed-weight algorithm to plan the shortest path. Some solutions use GPS positioning combined with preset trajectories to perform tasks, but are prone to signal loss in obstructed areas. Some studies have attempted to introduce reinforcement learning to optimize paths, but the computational overhead is high and the real-time performance is insufficient. In addition, existing systems usually handle navigation and robotic arm control independently, and task execution relies on a serial process. Frequent manual intervention is required in the event of sudden obstacles or energy fluctuations, resulting in low overall robustness.

[0004] Current technology has the following limitations: during operation, the robot's endurance is insufficient, which severely limits its continuous operation capability and deployment range; in the event of sudden obstacles or mission interruptions, it relies on manual operation, resulting in high response delays and a sharp drop in task execution efficiency; in complex environments, it is difficult to flexibly coordinate different task objectives based on real-time status, resulting in rigid decision-making and easy to cause operational risks. Summary of the Invention

[0005] Technical problems solved

[0006] In response to the shortcomings of the existing technology, the present invention provides a path intelligent navigation system for photovoltaic installation robots to solve the problems raised in the above background technology, such as the robot's insufficient endurance during operation, which seriously limits its continuous operation capability and deployment range; it relies on manual operation in the event of sudden obstacles or task interruptions, with high response delays and a sharp drop in task execution efficiency; and it is difficult to flexibly coordinate different task objectives according to real-time status in complex environments, resulting in rigid decision-making and easy operational risks.

[0007] Technical Solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: a path intelligent navigation system for a photovoltaic installation robot, comprising:

[0009] The environmental perception module collects the three-dimensional spatial information of the installation environment, ground material characteristics and obstacle location data in real time through lidar, cameras and ultrasonic sensors, and obtains the dynamic state parameters of the robot in real time through the robot's management system, drive control system and positioning system. After multi-sensor data fusion processing, it generates an environmental map containing slope, material friction coefficient and obstacle distribution, and transmits the environmental map and dynamic state parameters to the path planning module;

[0010] The path planning module is based on the environmental map, the robot's current position, and the target position. It also integrates the inventory status of the photovoltaic panel storage device and the operation progress of the robotic arm control system obtained by the communication unit, and uses a path planning algorithm to generate an optimal navigation path. When an environmental mutation event or a task chain energy overrun event is detected, the dynamic coupling unit is automatically activated, and the path energy consumption weight is corrected through the environmental mutation adaptive rule chain. The energy quota is dynamically allocated through the task chain energy conservation network, and a multi-objective game equalizer is used to adjust the priorities of energy consumption, time, and safety goals.

[0011] The navigation control module controls the movement of the robot's walking mechanism according to the optimal navigation path, and uses task mirror fractal technology to split emergency tasks into micro-kernel tasks that are executed in parallel. It monitors the robot's speed and direction in real time, and pauses movement and triggers path replanning when encountering sudden obstacles.

[0012] The communication unit interacts with the photovoltaic panel storage device and the robotic arm control system in real time to synchronously update the inventory location and task execution status; the data storage module pre-stores a mapping table of ground material type and friction coefficient, and records environmental parameters, historical paths, energy consumption data, execution results and operation records to build a historical database, and optimizes the path planning algorithm through the data fed back by the path planning module.

[0013] Preferably, the environment perception module cooperates with a laser radar, a camera and an ultrasonic sensor; the laser radar is installed on the top of the robot with a fixed support, reflects laser pulses through a rotating mirror, covers a horizontal 360-degree and vertical 30-degree field of view angle with a scanning frequency of 20 times per second, and generates three-dimensional point cloud data; the point cloud data is distinguished between ground points and non-ground points through a density clustering algorithm, the terrain slope is calculated according to the height difference of the ground point cloud, and the slope value is determined by comparing the height change rate of adjacent point clouds with a preset terrain model; the camera uses dual-spectrum imaging technology to synchronously collect RGB images and infrared images, superimposes the two through an image registration algorithm, inputs a pre-trained convolutional neural network for ground material classification, matches the classification result with a pre-stored material friction coefficient mapping table in the data storage module, and outputs a real-time energy consumption influence factor; the energy consumption influence factor = reference friction coefficient correction value × (1 + slope correction coefficient) × normalized proportion factor; the energy consumption influence factor is used to quantify the comprehensive influence of ground material and slope on the motion energy consumption of the robot, and the calculation logic is: the material type identified by the camera is used to query the pre-stored friction coefficient, such as 0.6 for asphalt, combined with the slope data detected by the laser radar, such as 10% slope corresponding to friction coefficient correction value 0.6 × 1.1 = 0.66, and then the corrected friction coefficient is mapped to the energy consumption influence factor range of 0.5-2.0 through a normalization formula, for example, friction coefficient 0.66 corresponds to factor 1.15, which is finally used as a multiplier of the path energy consumption weight in the path planning module; the ultrasonic sensor emits 40kHz pulse signals at a frequency of 5Hz, calculates the distance data of obstacles within 1 meter in front of the robot by measuring the time difference of the echo, and corrects the measurement error between multiple sensors through a triangular positioning method.

[0014] Preferably, the pre-trained convolutional neural network is further extended for potential difference prediction model training, and the historical database is used to synchronously store photovoltaic panel infrared images, cable connection state labels and measured potential difference data for training the potential difference prediction model; the specific operation is: after completing the ground material classification, the camera extracts the shadow distribution features on the surface of the photovoltaic panel through the gray value variance of the infrared image, the formula is , wherein I pixel is the pixel gray value, and I avg is the average gray value of the area; and the terminal color, cable direction and cable connection state are identified through the RGB image, and then input into the pre-trained voltage difference prediction model to output the open-circuit potential difference estimate value between adjacent panels; during the movement of the robot along the optimal path, if the robot arm holds the photovoltaic panel and enters the range of 5 meters from the installed photovoltaic panel array, when the open-circuit potential difference estimate value between the target installation position and the adjacent panel exceeds 30V, an insulation avoidance instruction is triggered: the offset of the path in the normal direction of the array is recalculated, and the offset distance satisfies When ΔV is the voltage difference, the unit of the offset distance is meter, and it is ensured that the minimum distance between the robot arm's motion trajectory and the live parts is ≥0.3 meters and the angle between the end effector axis and the live parts is ≥45°.

[0015] Preferably, among the dynamic state parameters, the remaining battery power is transmitted by the robot's management system through the CAN bus at a frequency of 1Hz with an accuracy of ±1%; the maximum movement speed is dynamically calculated by the drive control system based on the motor performance curve and the real-time ground friction coefficient, and transmitted to the environmental perception module through shared memory; the turning radius is a preset parameter, fixed at 0.5 meters, stored in the configuration file of the drive control system, and read by the environmental perception module through the Modbus protocol; the current position is updated at a frequency of 10Hz by the GPS and odometer fusion positioning system, and the positioning error is less than ±0.05 meters; the target position is sent by the photovoltaic panel storage device through the communication unit, including the three-dimensional coordinates of the installation point and the priority Level identification; multi-sensor data is time-series aligned and noise filtered through a Kalman filter, and then input into a Bayesian network for probabilistic fusion to generate an environmental map containing a slope value of 0%-40% and a material friction coefficient of 0.2-0.8, and obstacle coordinates in a three-dimensional Cartesian coordinate system; the environmental map is encapsulated in JSON format and integrated with dynamic state parameters to form a comprehensive data packet, which is transmitted to the path planning module via the Gigabit Ethernet protocol with a transmission period of 100 milliseconds. Data integrity is verified before each transmission. If the verification fails, the sensor self-test program is triggered; the calibration parameters of the lidar and camera are regularly updated using the checkerboard calibration method to ensure that the data fusion accuracy error is less than 2%.

[0016] Preferably, the path planning module is based on the environmental map, the current position of the robot and the target position, and integrates the inventory status of the photovoltaic panel storage device and the operation progress of the robotic arm control system obtained by the communication unit, and uses the path planning algorithm to generate the optimal navigation path. The path planning module receives the environmental map from the environmental perception module, and the real-time positioning data of the robot is obtained by integrating GPS and odometer. The target position coordinates are the photovoltaic panel installation target point provided by the photovoltaic panel storage device, and the photovoltaic panel inventory status synchronized by the communication unit, such as the remaining number of photovoltaic panels and storage coordinates, and the task execution progress fed back by the robotic arm control system, such as the completion percentage of the current grasping action; the module first proposes The dynamic constraints of the robot are as follows: the maximum movement speed is determined by the rated torque of the robot's drive motor and the ground friction coefficient. The preset motor performance curve is obtained by the table lookup method, and the maximum allowable speed is calculated in combination with the real-time friction coefficient; the turning radius is pre-set to a fixed value of 0.5 meters based on the robot chassis structure and the differential steering model, and its compatibility with the current path curvature is verified by the kinematic equation; the remaining battery power is monitored in real time by the robot's management system, and the remaining energy is calculated by the voltage-current integration method and converted into the drivable distance, that is, every 1% of the power corresponds to 10 meters; the ground material friction coefficient extracts the value of the current position from the environmental map. If the robot is in the intersection area of ​​multiple materials, the weighted average is taken.

[0017] Preferably, based on the dynamic constraints, the path planning module calls the Dijkstra algorithm to generate an initial path: first, the historical path data in the data storage module is loaded, and the historical path segments that are similar to the slope distribution, obstacle density and material type of the current environment map are screened out through the similarity matching algorithm; the screened segments are spliced ​​into candidate paths according to topological connectivity, and the total energy consumption of each candidate path is calculated through the energy consumption evaluation model, and the formula is total energy consumption = path length × average friction coefficient + number of turns × turning energy consumption coefficient; if there are multiple candidate paths, the path with the lowest total energy consumption is selected as the initial shortest path; if there is no matching segment in the historical data, a graph network is constructed based on the rasterization model of the environment map, with the grid center as the node and the energy consumption of movement between nodes as the edge weight, and the weight is distance × friction coefficient. The Dijkstra algorithm is run to traverse all reachable nodes to generate the global shortest path from the starting point to the end point.

[0018] Preferably, after the initial path is generated, the path planning module monitors two types of events in real time: environmental mutation events are determined by the slope mutation rate and the obstacle distance change rate of the environment perception module, when the laser radar detects a sudden change in terrain slope or the ultrasonic sensor reports a sudden decrease in obstacle distance, it is determined that an environmental mutation event has occurred, such as a slope change of more than 5% / second or a sudden decrease in obstacle distance of more than 0.2 meters / second; and task chain energy overspending events are determined by comparing the difference between actual energy consumption and allocated budget, such as an overspending of more than 15%; when an event is detected, the dynamic coupling unit is immediately activated, which calls the environmental mutation adaptive rule chain, matches the pre-set external disturbance parameters according to the mutation type, the external disturbance parameters are factors that change in real time in the environment and have a direct impact on the robot path planning, including terrain slope mutation value, obstacle distance sudden decrease rate, ground material friction coefficient instantaneous change and sudden weather conditions such as rain, snow, sandstorm and haze; when the slope increases, the path weight is adjusted according to the rule of increasing the energy consumption weight by 2% for every 1% increase in slope, and the task chain energy conservation network is triggered to redistribute the energy quota of subsequent tasks; if the environmental mutation index exceeds the pre-set threshold, i.e. the slope > 30% or the obstacle distance < 0.3 meters, the current path is forcibly interrupted, the emergency path generation algorithm is called, the A* algorithm is used to replan the bypass path with safety as the highest priority, and the navigation control module is used to execute emergency stop or obstacle avoidance actions.

[0019] Preferably, the task chain energy conservation network monitors the energy consumption data of each microkernel task in real time, and the microkernel tasks are divided into three levels according to preset priorities: the first level is the critical task: photovoltaic panel installation task, the second level is the sub-critical task: photovoltaic panel transportation task, and the third level is the low-priority task: equipment self-inspection or non-essential path movement; when the actual energy consumption of a task exceeds the allocated budget, an energy overspending event is triggered, and the network first deducts the energy quota of subsequent tasks in proportion according to the task priority level. The specific rules are: the energy quota deduction ratio of the third-level task is 70% of the overspending amount, the second-level task deducts 30%, and the first-level task does not deduct; the deducted energy is distributed in reverse according to the priority of the overspending task, that is, the first-level task is supplemented first, and the second-level task is second; if the energy is still insufficient, a compensation path is generated, and the generation of the compensation path is based on the task priority level. The dependency graph of the tasks is analyzed by a topological sorting algorithm. For example, the transport task must be completed before the installation task. The execution order of the task chain is rearranged to ensure that the first-level task takes up the remaining energy first. The adjusted energy quota is encapsulated in JSON format by the communication unit, including the task ID, the adjusted energy value and the execution time window, and is synchronized to the robotic arm control system in real time. The robotic arm control system dynamically adjusts the grasping speed and installation pressure parameters according to the energy quota. For example, when the energy quota of the first-level task increases, the robotic arm grasping speed is reduced by 10% to match the high-precision installation requirements. If the energy consumption requirements of the critical task cannot be met after two consecutive adjustments, the path planning module is triggered to generate the shortest recharging path and all non-critical tasks are suspended until the battery power recovers above the threshold.

[0020] Preferably, the multi-objective game equalizer takes energy consumption, time and safety as optimization targets, and constructs a multi-objective function containing three indicators; the energy consumption target is defined as the ratio of the total energy consumption of the path to the remaining battery power, the time target is the difference between the path execution time and the task deadline, and the safety target is an inverse proportional function of the minimum distance to the obstacle and the preset safety threshold; the multi-objective function is iteratively optimized by the NSGA-II algorithm, firstly generating an initial population containing 100 groups of path solutions, dividing the solution set into multiple levels using non-dominated sorting, and then screening out a uniformly distributed Pareto front solution set by comparing the congestion degree; each path solution in the solution set is marked with a specific score of energy consumption, time and safety, and sorted according to the non-dominated relationship; when the environmental perception module detects that the obstacle distance is lower than the safety threshold of 0.5 meters, the equalizer forces the weight of the safety target to be increased. The safety score is increased from the default 33% to over 70%, while the energy consumption and time weights are reduced, such as adjusting them to 15% respectively, and paths with a safety score higher than 80 are screened in the Pareto solution set. If the battery power is less than 30%, a safety priority decision is triggered. At this time, the safety weight is fixed at 70%, and the remaining 30% weight is dynamically allocated by energy consumption and time according to the urgency of the task. For example, the time weight of the critical task is increased to 20%, and the energy consumption weight is reduced to 10%. The finally selected path needs to be verified by the dynamic coupling unit to see whether it meets the robot's maximum speed, turning radius and other constraints. If the verification fails, the NSGA-II algorithm is re-called to generate a new solution set. The priority classification rules are as follows: in the default state: safety > time > energy consumption; in the low-battery state: safety > energy consumption > time, and the weight adjustment parameters are optimized by feedback from the task success rate and failure rate in the historical database.

[0021] Preferably, the navigation control module receives the optimal navigation path generated by the path planning module, and first decomposes the path into a discrete sequence of motion control instructions; the input variables of the fuzzy control algorithm include the obstacle distance measured in real time by the laser radar, with an accuracy of ±0.05 meters; the ground material friction coefficient is obtained from the data storage module by calling a mapping table after the camera identifies the material type, with a range of 0.2 to 0.8; and the target direction deviation is calculated by the difference between the current position of the robot and the coordinates of the path node, with a unit of degrees; the fuzzification process of the input variables uses a triangular membership function, for example, the obstacle distance is divided into three fuzzy sets: "near: 0-0.5 meters", "medium: 0.5-1.5 meters", and "far: more than 1.5 meters"; the ground friction coefficient is divided into three fuzzy sets: "low: 0.2-0.4", "medium: 0.4-0.6", and "high: 0.6-0.8"; and the target direction deviation is divided into three fuzzy sets: "small: within ±5°", "medium: ±5°-15°", and "large: more than ±15°"; the fuzzy rule base contains 27 rules, for example, if the obstacle distance is near, the friction coefficient is low, and the direction deviation is large, the output speed is low speed 0.2 m / s and the steering angle is sharp turn 15°; the fuzzy reasoning uses the Mamdani model, which is de-fuzzified by weighted average method, and finally outputs the accurate speed value between 0.1-1.5 m / s and the steering angle between ±30°; when the laser radar detects a sudden obstacle with a distance less than 0.3 meters, an emergency stop signal is triggered immediately, the motor torque of the robot walking mechanism is reduced to zero within 50 milliseconds, and the obstacle coordinates, including three-dimensional position and size, are uploaded to the path planning module through the communication unit to trigger path re-planning; in the emergency path execution stage, the task mirror fractal technology splits the original task, such as photovoltaic panel installation, into microkernel tasks according to functions, including grabbing, short-distance carrying, temporary obstacle avoidance, and state synchronization; the grabbing is performed by the robot arm from the storage device, which takes 2 seconds; the short-distance carrying moves to the target point along the emergency path, with a maximum distance of 1 meter; the temporary obstacle avoidance generates a local detour path to bypass the sudden obstacle; and the state synchronization updates the task progress to the main task chain; each microkernel task is managed by an independent real-time operating system thread, and the thread priority is assigned according to the task criticality, for example, the grabbing thread has the highest priority, followed by the short-distance carrying thread, and the temporary obstacle avoidance thread; resource mutual exclusion is realized through shared memory and semaphore mechanism when executing in parallel; the communication unit uses Modbus-TCP protocol to interact with the robot arm control system, sends the grabbing angle and installation timing instructions every 100 milliseconds, and receives the completion state feedback of the robot arm; when all microkernel tasks are completed, the state data of each thread is merged into the main task chain through hash verification, so that the overall efficiency loss is within the preset threshold range of 10%; if the data consistency check passes, the subsequent task is continued to be executed, otherwise the abnormal processing procedure is triggered and rolled back to the nearest recoverable state.

[0022] Preferably, the dynamic coupling unit includes a path disturbance generator, which generates three types of pseudo paths to optimize system decision-making; the first type of pseudo path is a high-energy consumption straight-through path, whose generation logic is based on the area with the lowest slope and least obstacles in the current environment map, forcing the robot to pass straight through at the maximum speed, ignoring the real-time energy consumption limit; the second type of pseudo path is a low-energy consumption detour path, which extends the path length and selects smooth areas with a ground friction coefficient below 0.3 to reduce energy consumption per unit distance; the third type of pseudo path is a random path containing virtual obstacles, which inserts simulated obstacle coordinates randomly in the map to test the robot's obstacle avoidance ability in complex environments; the insertion condition of the pseudo path is that the system detects that the topological structure of three consecutive path planning is more than 98% similar, specifically by calculating the similarity through the mean value of the Euclidean distance of path node coordinates, if the mean deviation is less than 2%, it is determined as a highly similar path.

[0023] Preferably, when executing the pseudo path, the navigation control module records the actual energy consumption data, including motor power, battery consumption rate and task time consumption, and monitors safety indicators in real time through laser radar and camera, such as the minimum distance from obstacles and the number of emergency stops; if the energy efficiency of a pseudo path improves by more than 10% compared to the historical optimal path, or the safety factor improves by more than 15% through the collision risk probability model calculation, it will be included in the formal path library, and the energy consumption weight parameter of the path planning module will be updated; otherwise, if the pseudo path execution process appears task interruption, energy consumption overruns or safety distance less than 0.2 meters, the path will be marked as a risk forbidden zone and prohibited from being called in subsequent planning; the data storage module corrects the virtual obstacle coordinate confidence in the environment map according to the pseudo path execution result, and synchronously optimizes the energy distribution parameters of the task chain energy conservation network, such as including high-energy consumption straight-through path success cases in energy budget adjustment rules, or limiting the path curvature of specific areas according to risk forbidden zone data; the whole process realizes system self-optimization through a closed-loop feedback mechanism, ensuring that the test and decision logic of the pseudo path is seamlessly embedded in the main task chain.

[0024] Preferably, the communication unit establishes a two-way communication link with the photovoltaic panel storage device and the robotic arm control system through the 5G network; receives the photovoltaic panel inventory position change information of the storage device every 200 milliseconds, and triggers the path planning module to regenerate the target path; at the same time, the real-time position and motion status of the robot are sent to the robotic arm control system, so that the robotic arm calculates the grasping angle and installation timing in advance; when the task chain energy conservation network detects that the actual energy consumption of a task exceeds the allocated energy budget, it is determined to be energy overspending, and the task chain energy conservation network is executed first to dynamically adjust the energy quota of subsequent tasks. When the task chain energy conservation network adjusts the energy quota for three consecutive times and still cannot meet the demand, it is determined to be an uncontrollable system state, and a manual intervention request is immediately sent to the monitoring terminal, and all dynamic adjustment operations are suspended, waiting for external instructions.

[0025] Preferably, the data storage module records the environmental parameters, historical paths, energy consumption data and operation logs of each path planning to build a historical database; the environmental parameters and energy consumption data in the historical database are analyzed by the random forest algorithm to identify high-frequency obstacle areas and abnormal energy consumption patterns, and high-risk areas with slopes greater than 25% or friction coefficients less than 0.3; the analysis results generate algorithm correction parameters, such as increasing the safety weight or limiting the path curvature, and feed them back to the path planning module; during the system idle period, the data cleaning process is started to delete redundant historical paths and back up the critical task log; the critical task log is a data set that records the complete process of photovoltaic panel installation operations to ensure the long-term stability of the system.

[0026] Beneficial effects

[0027] The present invention provides a path intelligent navigation system for a photovoltaic installation robot. It has the following beneficial effects:

[0028] The present invention generates high-precision environmental maps through multi-sensor data fusion, improving the accuracy of terrain slope, ground friction coefficient and obstacle identification; combines with dynamic coupling units to adjust path energy consumption weights and energy quotas in real time, effectively responding to environmental mutations and task chain energy fluctuations; uses task mirror fractal technology to split emergency tasks into parallel micro-kernel tasks to ensure uninterrupted key operations under sudden obstacles; dynamically balances energy consumption, time and safety priorities based on a multi-objective game equalizer, enhancing decision-making flexibility in complex scenarios; uses a path disturbance generator to insert pseudo-path optimization model parameters, continuously improves algorithm adaptability through historical data feedback, and significantly improves photovoltaic installation efficiency and system robustness.

[0029] The application synchronizes the photovoltaic panel inventory state and the mechanical arm control progress in real time through the communication unit, optimizes the synergy of path planning and task execution; the data storage module records environmental parameters and historical path data, identifies high-frequency risk areas and corrects energy consumption weights in combination with machine learning algorithms, and reduces abnormal energy consumption fluctuations; the navigation control module dynamically adjusts motion parameters using fuzzy algorithms to accurately match real-time environmental changes; the task chain energy conservation network dynamically allocates energy quotas according to priority, and prioritizes to ensure key task completion rate; the system continuously optimizes path planning strategies through a closed-loop feedback mechanism, reduces the need for manual intervention, and improves installation stability and long-term operation reliability in complex terrain. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the application will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the protection scope of the application.

[0031] Embodiment one

[0032] The embodiment of the application provides a path intelligent navigation system for a photovoltaic installation robot, comprising, in a flat installation scene of a photovoltaic power station, after the system is started, a top laser radar is used by an environment perception module to cover a horizontal 360-degree and vertical 30-degree field of view angle at a scanning frequency of 20 times per second, three-dimensional point cloud data is generated, ground points and non-ground points are distinguished through a density clustering algorithm, a terrain slope is calculated according to the height difference of the ground point cloud, and the slope value is determined as 2% through preset terrain model comparison; at the same time, a camera adopts a dual-spectrum imaging technology to synchronously collect RGB and infrared images, after image registration and superposition, a pre-trained convolutional neural network is input to classify the ground material, and after the ground material is identified as a concrete material, a mapping table is called from a data storage module to obtain a friction coefficient of 0.6; an ultrasonic sensor emits a 40 kHz pulse signal at a frequency of 5 Hz to detect that there is no obstacle within 1 meter in front of the robot; after multi-sensor data is time-aligned and noise-filtered through a Kalman filter, the data is input into a Bayesian network for probability fusion to generate an environment map containing a slope of 2%, a friction coefficient of 0.6 and no obstacle, and the environment map is packaged in a JSON format and transmitted to a path planning module through a gigabit Ethernet; after the path planning module receives the environment map, based on the current position of the robot and the installation coordinates provided by a photovoltaic panel storage device, in combination with the grabbing progress of 80% fed back by a mechanical arm control system, a dynamic constraint condition is extracted: the maximum motion speed is calculated as 1.2 m / s according to a motor performance curve and the friction coefficient of 0.6, the steering radius is fixed as 0.5 meters, and the corresponding drivable distance of the remaining battery capacity is 800 meters; a Dijkstra algorithm is called to load similar environment path segments in a historical database, such as a slope < 5%, a friction coefficient of 0.5-0.7 and no obstacle, a candidate path is spliced after the path segments, total energy consumption is calculated, and an initial path with the shortest path length and the least number of steering times is selected.

[0033] Then the navigation control module decomposes the path into discrete motion instructions, adopts a fuzzy control algorithm to match input variables such as long obstacle distance, high friction coefficient and small direction deviation, outputs a speed of 1.0 m / s and a steering angle of 0°, and controls the robot to move smoothly along the path; a communication unit synchronously updates the storage position of the storage device and the grabbing state of the mechanical arm every 200 milliseconds, the path planning module updates the target point to the next installation position after the mechanical arm completes grabbing; the data storage module records the current environment parameters and energy consumption data, and after analysis through a random forest algorithm, no abnormal pattern is found; the task chain energy conservation network monitors the energy consumption of each microkernel task, the system successfully completes the installation of 10 photovoltaic panels because there is no overspending event in a stable environment, the whole process takes 25 minutes, the total energy consumption is 18% of the battery capacity, and the historical database updates the task log of this time for subsequent optimization.

[0034] Embodiment two

[0035] The difference between this embodiment and the first embodiment lies in: the dynamic adjustment process for dealing with sudden obstacles and energy overruns in complex mountainous terrain; the lidar scanning of the environmental perception module detects that the terrain slope suddenly increases to 12%, and the ultrasonic sensor reports that an unmodeled rock obstacle appears 0.4 meters ahead, with a slope mutation rate exceeding 5% / second and a sudden deceleration rate of 0.3 meters / second from the obstacle distance, triggering an environmental mutation event; the path planning module immediately activates the dynamic coupling unit, calls the environmental mutation adaptive rule chain, matches the external interference parameters based on the sudden increase in slope by 12%, and increases the current path energy consumption weight from the default value of 50% to 74% according to the rule of increasing the energy consumption weight by 2% for every 1% increase in slope; the task chain energy conservation network detects that the actual energy consumption of the current handling task exceeds 20%, and deducts 70% of the energy quota for the third-level task equipment self-inspection according to the rules, and supplements it to the first-level installation task.

[0036] At the same time, the multi-objective game equalizer forced the safety weight to be increased from 33% to 70% because the obstacle distance was lower than the safety threshold of 0.5 meters. The time and energy consumption weights were both adjusted to 15%. The Pareto front solution set was generated through the NSGA-II algorithm, and a detour path with a safety score of 85 was screened out. After receiving the new path, the navigation control module used the fuzzy control algorithm to adjust the speed to 0.5m / s and the steering angle to 10°, and triggered the task mirror fractal technology to split the original installation task into three micro-kernel tasks: grasping, short-distance transportation, and temporary obstacle avoidance. The grasping thread was completed by the robotic arm at a 10% reduced speed to complete high-precision board picking. Short-distance transportation moved 0.8 meters along the emergency path to avoid rocks. Temporary obstacle avoidance generated a local detour path and synchronized the state through shared memory.

[0037] The communication unit then uploads the obstacle coordinates to the path planning module in real time. The latter calls the A* algorithm to generate a random path containing virtual obstacles to test the obstacle avoidance capability. The path perturbation generator of the dynamic coupling unit detects that the similarity of three consecutive path plannings reaches 98.5%, and inserts a high-energy direct pseudo path. The robot attempts to pass through the low-slope area at a maximum speed of 1.2m / s. The actual energy efficiency is improved by 12% compared with the historical path, so it is included in the formal path library. The data storage module corrects the confidence level of the rock coordinates in the environmental map and optimizes the energy allocation parameters of the task chain. Finally, the system completes the key installation task with 28% battery power remaining, and the time is increased to 35 minutes, but the safety distance is always maintained above 0.5 meters. The historical database records this dynamic adjustment process to enhance the adaptive ability of subsequent complex environments.

[0038] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A path intelligent navigation system for a photovoltaic installation robot, characterized in that: include: The environmental perception module uses lidar, cameras, and ultrasonic sensors to collect three-dimensional spatial information of the installation environment, ground material characteristics, and obstacle location data in real time. It also obtains the robot's dynamic state parameters in real time through the robot's management system, drive control system, and positioning system. After multi-sensor data fusion processing, it generates an environmental map that includes slope, material friction coefficient, and obstacle distribution. At the same time, it obtains the inventory status of the photovoltaic panel storage device and the operation progress of the robotic arm control system through the communication unit, and transmits the environmental map and dynamic state parameters to the path planning module. The path planning module is based on the environmental map, the current position of the robot and the target position, and integrates the inventory status of the photovoltaic panel storage device and the operation progress of the robotic arm control system. It uses the path planning algorithm to generate the optimal navigation path. When an environmental mutation event is detected, the dynamic coupling unit is automatically activated. Through the environmental mutation adaptive rule chain, the path energy consumption weight is corrected in real time according to the external interference parameters corresponding to the environmental mutation event and the friction coefficient of the ground material. If the environmental mutation index exceeds the threshold, the current task is interrupted and an emergency path is generated. When a task chain energy overspending event is detected, the dynamic coupling unit is automatically activated. The energy quota of subsequent tasks is dynamically adjusted through the task chain energy conservation network, the energy budget of low-priority tasks is proportionally deducted, and a compensation path for adjusting the order of task execution is generated. The priority of energy consumption, time and safety goals is adjusted using a multi-objective game equalizer. The navigation control module controls the movement of the robot's walking mechanism according to the optimal navigation path, and uses task mirror fractal technology to split emergency tasks into micro-kernel tasks that are executed in parallel. It monitors the robot's speed and direction in real time, and pauses movement and triggers path replanning when encountering sudden obstacles. The communication unit interacts with the photovoltaic panel storage device and the robotic arm control system in real time to synchronously update the inventory location and task execution status; The data storage module pre-stores a mapping table of ground material types and friction coefficients, and records environmental parameters, historical paths, energy consumption data, execution results and operation records to build a historical database, and optimizes the path planning algorithm through the data fed back by the path planning module.

2. The path intelligent navigation system for a photovoltaic installation robot according to claim 1, characterized in that: The environmental perception module generates three-dimensional point cloud data through the laser radar scanning to identify the terrain slope and obstacle shape; The camera identifies the type of ground material through images, and calculates the real-time energy consumption impact factor in combination with the friction coefficient mapping table pre-stored in the data storage module, which is used to correct the energy consumption weight of the path planning module; the ultrasonic sensor detects the distance information of close-range obstacles; The environmental perception module communicates with the robot's management system and drive control system through the CAN bus protocol, and obtains the robot's dynamic state parameters in real time, including the maximum movement speed, turning radius, remaining battery power, current position, target position and turning radius setting value; obtains the robot's current position through the GPS and odometer fusion positioning system, and receives the target position coordinates provided by the photovoltaic panel storage device through the communication unit; the environmental perception module fuses the multi-sensor data to generate an environmental map containing slope, material friction coefficient and obstacle distribution. The obstacle position, shape and dynamic change data recorded in the environmental map constitute an obstacle distribution model, and the environmental map and the dynamic state parameters are transmitted to the path planning module.

3. The path intelligent navigation system for a photovoltaic installation robot according to claim 2, characterized in that: The path planning module uses the robot's maximum movement speed, turning radius, remaining battery power and ground material friction coefficient provided by the environmental perception module as dynamic constraints, and uses the Dijkstra algorithm based on the dynamic constraints in combination with the historical path data of the data storage module to generate an initial shortest path; when an environmental mutation event is detected, the environmental mutation adaptive rule chain in the dynamic coupling unit is activated, and the path energy consumption weight is corrected in real time according to the external interference parameters and ground material friction coefficient corresponding to the environmental mutation event. If the environmental mutation index exceeds the threshold, the current task is interrupted and an emergency path is generated and executed through the navigation control module; when the energy consumption of a task in the task chain exceeds the budget, the task chain energy conservation network dynamically adjusts the energy quota of subsequent tasks, deducts the energy budget of low-priority tasks in proportion, and generates a compensation path for adjusting the order of task execution to ensure that critical tasks are completed first; the task chain is a sequence composed of multiple related tasks in photovoltaic installation operations.

4. The path intelligent navigation system for a photovoltaic installation robot according to claim 2, characterized in that: The dynamic coupling unit includes: a multi-objective game equalizer that calculates the Pareto front solution set of multi-objective optimization of energy consumption, time, and safety goals in real time. When the obstacle distance detected by the environmental perception module is lower than the safety threshold, the safety factor is determined to be lower than the preset threshold, and the safety weight is forcibly increased to above 70%. When the battery power is lower than 30%, a safety-first path selection decision is triggered, and a high-energy consumption path is actively selected to prioritize the completion of critical tasks. The dynamic coupling unit also includes a path perturbation generator that is used to insert pseudo paths to optimize the parameters of the obstacle distribution model. Based on the topological structure comparison between historical path data and the current path, when the similarity of three consecutive path plans exceeds 98%, the pseudo path is inserted into the path. The pseudo path includes pseudo high-energy consumption nodes or virtual obstacles generated based on historical data and the obstacle distribution model. After the pseudo path is executed, the obstacle distribution model in the environment map and the energy allocation parameters of the task chain energy conservation network are updated using the pseudo path execution results recorded by the data storage module.

5. The path intelligent navigation system for a photovoltaic installation robot according to claim 3, characterized in that: The navigation control module uses a fuzzy control algorithm to dynamically adjust the robot's motion parameters, and performs fuzzy rule matching based on the obstacle distance and ground material friction coefficient in the real-time environmental data provided by the environmental perception module and the target direction of the optimal navigation path generated by the path planning module to calculate the optimal solution for speed and steering angle; when the lidar detects a sudden obstacle, the robot's motion is immediately suspended and the obstacle position is fed back to the path planning module; during the emergency path execution process, the task mirror fractal technology is used to split the original task into multiple micro-kernel tasks executed in parallel, and after the micro-kernel tasks are completed, they are automatically merged into the main task chain to ensure uninterrupted key operations.

6. The path intelligent navigation system for a photovoltaic installation robot according to claim 4, characterized in that: The path disturbance generator generates three types of pseudo paths, including a high-energy-consumption direct path, a low-energy-consumption detour path, and a random path containing virtual obstacles; after executing the pseudo paths, actual energy consumption and safety data are collected. If the energy efficiency and safety factor of one of the pseudo paths are better than the corresponding thresholds of the historical optimal path, it will be included in the formal path library; otherwise, it will be marked as a risk-restricted area; and the ground friction model of the environmental mutation adaptive rule chain and the multi-objective weight parameters of the multi-objective game equalizer are updated through the feedback link between the data storage module and the path planning module.

7. The path intelligent navigation system for a photovoltaic installation robot according to claim 5, characterized in that: During the emergency path planning stage, the task mirror fractal technology decomposes the original task into multiple independent micro-kernel tasks. Each micro-kernel task includes photovoltaic panel grabbing, short-distance transportation and temporary obstacle avoidance instructions. The temporary obstacle avoidance instructions are dynamically generated based on the emergency path generated by the path planning module. The communication unit coordinates the parallel execution of multiple micro-kernel tasks, and automatically merges them into the main task chain after the micro-kernel tasks are completed, so that the overall efficiency loss is within the preset threshold range.