Intelligent control method and system for obstacle avoidance of aerial work platform

Abstract

The application discloses an intelligent obstacle avoidance control method and system for a high-altitude operation platform, and relates to the field of high-altitude operation platforms.The method comprises the following steps: collecting sensing data of a working area of the high-altitude operation platform, and constructing a three-dimensional space map based on the sensing data; identifying dynamic and static obstacles in the three-dimensional space map, and dividing the working area into risk levels in combination with a topological potential field algorithm; according to the working area division result, planning an initial obstacle avoidance path by using a random tree algorithm, optimizing the initial obstacle avoidance path in combination with the structural parameters of the high-altitude operation platform, and generating a final obstacle avoidance path; decoupling the final obstacle avoidance path into control instructions for vertical lifting and horizontal displacement, executing corresponding obstacle avoidance behavior modes based on the control instructions, and monitoring the execution result to optimize the control instructions.The application realizes behavior modeling of dynamic obstacles by constructing a three-dimensional space map, and improves the safety and efficiency of path planning.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aerial work platforms, and in particular to an aerial work platform obstacle avoidance intelligent control method and system. BACKGROUND

[0002] An aerial work platform (AWP), also commonly known as an aerial work vehicle or aerial lift, is a device specifically designed for high-altitude work. This kind of device is widely used in construction, power maintenance, bridge inspection, advertising board maintenance, warehouse operation and other fields. Aerial work platform can provide a safe and stable platform for workers to complete various tasks in the air, such as building facade maintenance, equipment installation, cleaning work, etc. Due to the particularity of high-altitude work, the operating environment is usually complex and changeable, full of various obstacles and potential dangerous areas, which makes it a great challenge to ensure work safety.

[0003] Most traditional aerial work platforms rely on manual operation, and operators avoid obstacles and accidents by observing the surrounding environment. Although this method is effective, it still has obvious limitations. First of all, the attention of the operator is easily affected by fatigue, working time, complex environment or sudden conditions, which may ignore potential obstacles or risks, leading to safety accidents. In addition, due to the limitations of human eyes in judging distance, angle and space, the success rate and accuracy of manual obstacle avoidance are not high in narrow or crowded environments.

[0004] Currently, the obstacle avoidance technology of aerial work platform has made some progress, mainly relying on sensor detection and basic obstacle avoidance algorithm to realize automatic obstacle avoidance. Many devices are equipped with ultrasonic sensors, laser radars and other sensors to detect obstacles around. However, these sensors have certain limitations in complex environments, with limited detection range and accuracy, and are easily disturbed by environmental factors such as light, weather changes or surface reflection. In addition, traditional obstacle avoidance algorithms are usually based on pre-set working environment models, mainly optimized for static obstacles and relatively simple terrain. For dynamic obstacles and complex, constantly changing terrain, the adaptability and response ability of these algorithms are poor, and they often cannot provide comprehensive and efficient obstacle avoidance protection.

[0005] In view of the problems in the related art, no effective solution has been proposed so far. SUMMARY

[0006] In view of the problems in the related art, the present application proposes an aerial work platform obstacle avoidance intelligent control method and system to overcome the above technical problems existing in the prior art.

[0007] To this end, the specific technical solutions adopted by the present application are as follows:

[0008] According to one aspect of the present application, an intelligent control method for aerial work platforms is provided, which comprises:

[0009] S1, collecting perception data of the working area of the aerial work platform, and constructing a three-dimensional space map based on the perception data;

[0010] S2, identifying dynamic and static obstacles in the three-dimensional space map, and combining a topological potential field algorithm to divide the working area into risk levels;

[0011] S3, according to the division result of the working area, using a random tree algorithm to plan an initial obstacle avoidance path, combining the structural parameters of the aerial work platform to optimize the initial obstacle avoidance path, and generating a final obstacle avoidance path;

[0012] S4, decoupling the final obstacle avoidance path into control instructions for vertical lifting and horizontal displacement, executing the corresponding obstacle avoidance behavior mode based on the control instructions, and monitoring the execution result to optimize the control instructions;

[0013] Wherein, S2 comprises:

[0014] S21, based on the three-dimensional space map, using a deep learning algorithm to identify and classify dynamic and static obstacles in the working area, and obtaining the position information of the dynamic and static obstacles;

[0015] S22, using digital twin technology to construct a digital twin model, and combining a potential field analysis algorithm to predict the motion trajectory of the dynamic obstacle;

[0016] S23, combining the prediction result of the dynamic obstacle and the position information of the static obstacle, using a topological potential field algorithm to divide the working area into a passing area, a cautious area and a prohibited area to identify different dynamic risk levels.

[0017] Optionally, collecting perception data of the working area of the aerial work platform, and constructing a three-dimensional space map based on the perception data comprises:

[0018] S11, collecting perception data of the working area of the aerial work platform through multi-modal sensors, and performing space-time synchronization and coordinate alignment on the collected perception data;

[0019] S12, using Kalman filtering algorithm to fuse the perception data after space-time synchronization and coordinate alignment, and combining a density-based clustering algorithm to denoise the fused perception data;

[0020] S13, dividing the denoised perception data into voxel grids, extracting the point centroid in each voxel as a target point, and constructing a three-dimensional space map by splicing the target points.

[0021] Optionally, the digital twin model is constructed by using a digital twin technology, and a motion trajectory of the dynamic obstacle is predicted by combining a potential field analysis algorithm, which comprises:

[0022] S221, based on the position information of the dynamic obstacle, a digital twin model is constructed by using a digital twin technology, and the motion state of the dynamic obstacle is updated through the fused real-time perception data;

[0023] S222, time-frequency analysis is performed on the motion perception data in the digital twin model, motion cycle characteristics are extracted, and the interactive topological relationship between the dynamic obstacles is mined by combining a graph neural network to generate a spatiotemporal correlation risk matrix to quantify the dynamic trend and potential risk;

[0024] S223, the spatiotemporal correlation risk matrix is mapped to a potential field intensity distribution, a potential field change caused by the motion of the dynamic obstacle is calculated by using a fast multipole algorithm, and a kinematic model of the dynamic obstacle is established based on the potential field change result;

[0025] S224, the kinematic model is combined with a long short-term memory network to predict the motion trajectory of the obstacle, and the kinematic model is optimized through real-time perception data feedback.

[0026] Optionally, the motion perception data in the digital twin model is subjected to time-frequency analysis, motion cycle characteristics are extracted, and the interactive topological relationship between the dynamic obstacles is mined by combining a graph neural network to generate a spatiotemporal correlation risk matrix to quantify the dynamic trend and potential risk, which comprises:

[0027] S2221, the motion perception data in the digital twin model is subjected to time-frequency analysis by using a short-time Fourier transform, and the periodicity of the motion perception data is extracted to obtain the motion cycle characteristics of the dynamic obstacle;

[0028] S2222, according to the motion perception data of the dynamic obstacle, each dynamic obstacle is defined as a node, and the interaction relationship between the obstacles is defined as an edge to construct an interactive topological graph;

[0029] S2223, the interactive topological graph is learned by using a graph neural network to identify and analyze the topological structure and dynamic interaction mode between the dynamic obstacles, and to capture the spatiotemporal interaction information between the dynamic obstacles;

[0030] S2224, the potential risk between the dynamic obstacles is calculated by combining the motion cycle characteristics and the spatiotemporal interaction information, a spatiotemporal correlation risk matrix is constructed, and the dynamic change and potential risk of the obstacle are predicted by analyzing the change trend of the spatiotemporal correlation risk matrix.

[0031] Optionally, the spatiotemporal correlation risk matrix is mapped to a potential field intensity distribution, the potential field change caused by the movement of the dynamic obstacle is calculated using the fast multipole algorithm, and a kinematic model of the dynamic obstacle is established based on the potential field change result, including:

[0032] S2231, the spatiotemporal correlation risk matrix is converted into a potential field intensity distribution using a nonlinear mapping function, and the mapped potential field intensity values are assigned to the corresponding grid nodes to construct a potential field intensity distribution map;

[0033] S2232, the potential field intensity distribution map is spatially divided using an octree structure, and cluster analysis is performed according to the position information of the dynamic obstacle to form a tree-like hierarchical structure;

[0034] S2233, the tree-like hierarchical structure is combined with real-time motion perception data of the dynamic obstacle, the potential field change caused by the movement of the dynamic obstacle is calculated using the fast multipole algorithm, and the potential field intensity distribution is updated;

[0035] S2234, the motion characteristics of the dynamic obstacle are extracted from the updated potential field intensity distribution, and a kinematic model of the dynamic obstacle is established based on the motion characteristics.

[0036] Optionally, the formula of the kinematic model is:

[0037] ;

[0038] In the formula, r ( t +1) represents the position of the dynamic obstacle at time t +1; r ( t ) represents the position of the dynamic obstacle at time t ; t represents time; v ( t ) represents the velocity of the dynamic obstacle at time t ; Δ t represents the time step; F ( t ) represents the external force on the dynamic obstacle at time t caused by the potential field; m represents the mass of the dynamic obstacle.

[0039] Optionally, according to the working area division result, an initial obstacle avoidance path is planned using a random tree algorithm, the initial obstacle avoidance path is optimized in combination with the structural parameters of the aerial work platform, and a final obstacle avoidance path is generated, including:

[0040] S31, according to the region division result, a three-dimensional grid map is constructed, and a Gaussian mixture model is used to fit the obstacle distribution boundary to generate an obstacle avoidance constraint boundary;

[0041] S32, performing path search based on the target-biased random tree algorithm based on the obstacle avoidance constraint boundary to generate an initial obstacle avoidance path;

[0042] S33, obtaining the configuration parameters of the aerial work platform, defining an optimization objective set including path length, minimum safety distance, curvature smoothness, and energy consumption coefficient, and dynamically adjusting the weights of the optimization objectives according to the control task requirements by using the entropy weight algorithm;

[0043] S34, adjusting the initial obstacle avoidance path by using the particle swarm optimization algorithm according to the weight adjustment result to generate a final obstacle avoidance path.

[0044] Optionally, the final obstacle avoidance path is decoupled into vertical lifting and horizontal displacement control instructions, corresponding obstacle avoidance behavior modes are executed based on the control instructions, and the execution results are monitored to optimize the control instructions, including:

[0045] S41, based on the preset spatial coordinates, the final obstacle avoidance path is decoupled into vertical and horizontal coordinate sequences, and a polynomial fitting technique is used to generate a motion profile of the aerial work platform;

[0046] S42, analyzing the motion profile, generating vertical lifting and horizontal displacement control instructions by using a fuzzy control algorithm, and accelerating the generation of control instructions through a parallel computing framework;

[0047] S43, according to the generated control instructions, selecting and executing corresponding obstacle avoidance behavior modes, the obstacle avoidance behavior modes including an emergency braking model and a dynamic detour behavior mode;

[0048] S44, real-time monitoring the execution effect of the selected obstacle avoidance behavior mode, and optimizing the control instructions based on the monitoring results.

[0049] Optionally, analyzing the motion profile, generating vertical lifting and horizontal displacement control instructions by using a fuzzy control algorithm, and accelerating the generation of control instructions through a parallel computing framework includes:

[0050] S421, extracting vertical and horizontal feature parameters from the generated motion profile, establishing a fuzzy input space, the feature parameters including height deviation, speed, and curvature error;

[0051] S422, combining the fuzzy input space and the fuzzy control algorithm, constructing a fuzzy control model including an input layer, a membership degree calculation level, and a fuzzy reasoning layer, and introducing a rule reduction layer to optimize the fuzzy control model;

[0052] S423, processing the vertical and horizontal feature parameters by using the optimized fuzzy control model to generate vertical lifting control instructions, and combining the preset horizontal control rules to obtain horizontal displacement control instructions;

[0053] S424. Vertical lifting and horizontal displacement control commands are assigned to different processing threads, and multi-core processors and parallel computing frameworks are used to accelerate the generation of control commands.

[0054] According to another aspect of the present invention, an intelligent control system for obstacle avoidance of aerial work platforms is also provided, the system comprising:

[0055] The sensing data acquisition and spatial construction module is used to collect sensing data of the working area of ​​the aerial work platform and construct a three-dimensional spatial map based on the sensing data.

[0056] The obstacle recognition and region segmentation module is used to identify dynamic and static obstacles in the 3D spatial map and combine the topological potential field algorithm to classify the risk level of the work area.

[0057] The obstacle avoidance path generation module is used to plan the initial obstacle avoidance path based on the work area division results using a random tree algorithm, and optimize the initial obstacle avoidance path in combination with the construction parameters of the aerial work platform to generate the final obstacle avoidance path.

[0058] The control command and execution optimization module is used to decouple the final obstacle avoidance path into control commands for vertical lifting and horizontal displacement, execute the corresponding obstacle avoidance behavior mode based on the control commands, and monitor the execution results to optimize the control commands.

[0059] The beneficial effects of this invention are as follows:

[0060] 1. By constructing a three-dimensional spatial map, this invention can not only capture the geometric features of static obstacles, but also analyze the motion characteristics of dynamic obstacles using time-series sensing data, thereby realizing the behavior modeling of dynamic obstacles, providing a comprehensive and accurate environmental cognition basis for path planning, and improving the safety and efficiency of path planning.

[0061] 2. This invention combines digital twin technology with potential field analysis algorithms to establish a motion prediction model for dynamic obstacles, and uses a topological potential field algorithm to dynamically divide the workspace into risk areas, enabling the path planner to identify and predict changes in risk areas in advance, thereby providing a valuable time window for path adjustment and ensuring safer and more efficient path decision-making.

[0062] 3. This invention achieves vertical and horizontal decoupling control of motion commands by adopting a model predictive control framework, and constructs a closed-loop optimization system by combining platform state feedback. This not only ensures the stability of the system, but also improves the flexibility and energy efficiency of obstacle avoidance behavior. It can avoid obstacles more efficiently while maintaining stability and optimize energy use efficiency. Attached Figure Description

[0063] In order to make the technical solutions in the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative effort based on these drawings.

[0064] Figure 1 is a flow chart of an intelligent obstacle avoidance control method for a high-altitude operation platform according to an embodiment of the present application;

[0065] Figure 2 is a principle block diagram of an intelligent obstacle avoidance control system for a high-altitude operation platform according to an embodiment of the present application.

[0066] In the drawings:

[0067] 1, perception data acquisition and space construction module; 2, obstacle identification and region division module; 3, obstacle avoidance path generation module; 4, control instruction and execution optimization module. DETAILED DESCRIPTION

[0068] To further illustrate the embodiments, the present application provides drawings, which are part of the disclosure of the present application, mainly used to illustrate the embodiments, and can explain the operating principle of the embodiments in conjunction with the related description of the specification. With reference to these contents, those skilled in the art should understand other possible embodiments and advantages of the present application.

[0069] According to an embodiment of the present application, an intelligent obstacle avoidance control method and system for a high-altitude operation platform are provided.

[0070] The present application will be further described in conjunction with the drawings and specific embodiments, as shown in the drawings, according to the intelligent obstacle avoidance control method for a high-altitude operation platform according to an embodiment of the present application, the method comprises: Figure 1

[0071] S1, collecting perception data of a working area of a high-altitude operation platform, and constructing a three-dimensional space map based on the perception data.

[0072] Preferably, collecting perception data of a working area of a high-altitude operation platform, and constructing a three-dimensional space map based on the perception data comprises:

[0073] S11, collecting perception data of a working area of a high-altitude operation platform by a multi-modal sensor, and performing space-time synchronization and coordinate alignment on the collected perception data;

[0074] S12, performing fusion processing on the perception data after space-time synchronization and coordinate alignment by using Kalman filtering algorithm, and performing denoising on the fused perception data in combination with a density-based clustering algorithm; ​

[0075] S13, dividing the de-noised perception data into a voxel grid, extracting the point centroid in each voxel as a target point, and constructing a three-dimensional space map through splicing of the target points.

[0076] It should be noted that, in specific embodiments, collecting perception data of the aerial work platform working area and constructing a three-dimensional space map based on the perception data includes:

[0077] First, 100,000 point cloud data are collected per second by the laser radar, 30 frames of images are taken per second by the camera, and 100 acceleration and angular velocity perception data are recorded per second by the IMU; then, a GPS time synchronization algorithm is used to ensure that the timestamps of all sensors are consistent, and a calibration system is used to align the coordinate systems of the laser radar and the camera to the coordinate system of the IMU; subsequently, a Kalman filter algorithm is used to fuse the perception data of the laser radar, the camera and the IMU, estimate the accurate position and motion trajectory of the platform, apply a DBSCAN algorithm to de-noise the laser radar point cloud perception data, and retain about 80% of the perception data points after removing noise points; finally, the de-noised point cloud perception data is divided into a voxel grid of 0.1m x 0.1m x 0.1m, the centroid of each voxel is calculated as a target point, and about 500,000 target points are generated; through splicing of the target points, a three-dimensional space map of the construction site is constructed, and the map accuracy reaches centimeter level.

[0078] In addition, the perception data includes visual data, distance and position data, radar data, environmental information data, GPS data and sensor fusion data, etc.

[0079] S2, identifying dynamic and static obstacles in the three-dimensional space map, and dividing the working area into risk levels in combination with a topological potential field algorithm.

[0080] S2 includes:

[0081] S21, based on the three-dimensional space map, using a deep learning algorithm to identify and classify dynamic and static obstacles in the working area, and obtaining position information of the dynamic and static obstacles.

[0082] It should be noted that static obstacles are usually fixed objects such as buildings and equipment, and the characteristics of these obstacles are relatively fixed positions with small changes; a deep learning model obtained by pre-training can identify and classify these objects; during training, static obstacles are classified as a category, and the model learns the characteristics of static obstacles according to the spatial distribution of point clouds.

[0083] Dynamic obstacles include moving objects such as pedestrians, vehicles, drones, etc., and the features of these obstacles are usually constantly changing in time. Point cloud perception data can be input into a deep learning model in time sequence, and the features of dynamic obstacles can be captured by using the differences in time. For dynamic obstacles, motion detection can be used to identify objects with large position changes between multiple time frames by comparing multiple frames of perception data.

[0084] S22, a digital twin model is constructed using digital twin technology, and a potential field analysis algorithm is used to predict the motion trajectory of the dynamic obstacle.

[0085] Preferably, the digital twin model is constructed using digital twin technology, and the potential field analysis algorithm is used to predict the motion trajectory of the dynamic obstacle, which includes:

[0086] S221, based on the position information of the dynamic obstacle, a digital twin model is constructed using digital twin technology, and the motion state of the dynamic obstacle is updated through the fused real-time perception data.

[0087] S222, time-frequency analysis is performed on the motion perception data in the digital twin model, the motion cycle characteristics are extracted, and the interactive topological relationship between the dynamic obstacles is mined using a graph neural network to generate a spatiotemporal correlation risk matrix to quantify the dynamic trend and potential risk.

[0088] Preferably, the time-frequency analysis is performed on the motion perception data in the digital twin model, the motion cycle characteristics are extracted, and the interactive topological relationship between the dynamic obstacles is mined using a graph neural network to generate a spatiotemporal correlation risk matrix to quantify the dynamic trend and potential risk, which includes:

[0089] S2221, short-time Fourier transform is used to perform time-frequency analysis on the motion perception data in the digital twin model, and the periodicity of the motion perception data is extracted to obtain the motion cycle characteristics of the dynamic obstacle.

[0090] S2222, according to the motion perception data of the dynamic obstacle, each dynamic obstacle is defined as a node, and the interaction relationship between the obstacles is defined as an edge, and an interactive topological graph is constructed.

[0091] It should be noted that in specific embodiments, it is assumed that there are four dynamic obstacles A, B, C, and D, and their position information and speed are recorded. According to the calculation:

[0092] The distance between A and B is 0.5 meters, and the speed difference is 0.2 m / s, so an edge is constructed. The distance between A and C is 2 meters, and the speed difference is 1 m / s, so no edge is constructed. The distance between B and D is 1 meter, and the speed difference is 0.3 m / s, so an edge is constructed. Therefore, the interactive topological graph is: A--B, B--D.

[0093] S2223, learn the interaction topology graph using a graph neural network, identify and analyze the topology structure and dynamic interaction mode between dynamic obstacles, and capture the spatio-temporal interaction information between dynamic obstacles.

[0094] S2224, combine the motion cycle characteristics and spatio-temporal interaction information to calculate the potential risk between dynamic obstacles, construct a spatio-temporal correlation risk matrix, and predict the dynamic changes and potential risks of obstacles by analyzing the trend of the spatio-temporal correlation risk matrix.

[0095] It should be noted that in specific embodiments, motion perception data in the digital twin model is subjected to time-frequency analysis to extract motion cycle characteristics, and the interaction topology relationship between dynamic obstacles is mined using a graph neural network to generate a spatio-temporal correlation risk matrix to quantify dynamic trends and potential risks, including:

[0096] 1. Motion cycle feature extraction: Short-time Fourier transform is performed on the motion perception data, and based on the transformation process, significant energy peaks are detected in the frequency domain, for example, AGV1 cycle 0.83±0.05s (corresponding to start-stop cycle), AGV2 cycle 1.21±0.03s (corresponding to loading and unloading operation), and personnel gait cycle 1.12±0.1s.

[0097] 2. Interaction topology graph construction: An adjacency matrix is established based on the minimum distance threshold between obstacles, and a Gaussian kernel function is used to calculate edge weights;

[0098] ;

[0099] wherein, w ij represents the edge weight between node i and node j . d ij represents the distance between node i and node j . σ represents a constant; wherein σ =0.3m, a time-varying undirected graph G(t)∈R 5×5 is constructed.

[0100] 3. Graph neural network training: A spatio-temporal graph convolution network is used, with 3 layers of graph convolution layers (64 channels per layer), a time convolution kernel length of 3, a loss function of mask cross-entropy, and a training set / validation set divided by 8:2.

[0101] 4. Risk matrix construction and analysis: Fusion of motion cycle characteristics (Fourier coefficients) and graph attention weights to construct a spatio-temporal correlation risk matrix G(t)∈R 5×5The results showed that when AGV1 interacts with personnel, the R value suddenly increases (the risk of collision increases), and the risk propagation exhibits an exponential decay characteristic with an impact radius of about 1.8m. Through matrix eigenvalue analysis, risk events can be predicted 0.8-1.2s in advance.

[0102] 5. Prediction Results: Comparing the actual trajectory with the predicted risk area, the system successfully issued warnings for 4 potential collisions (warning time window 1.5-2.3s), with a false positive rate controlled at 3.1%. The risk matrix visualization results show that the overlap between the high-risk area and the actual trajectory of the obstacle reaches 89.4%.

[0103] S223. Map the spatiotemporal correlation risk matrix to a potential field intensity distribution, use the fast multipole algorithm to calculate the potential field changes caused by the motion of dynamic obstacles, and establish a kinematic model of dynamic obstacles based on the potential field change results.

[0104] Preferably, the spatiotemporal correlation risk matrix is ​​mapped to a potential field intensity distribution, and the potential field changes caused by the motion of dynamic obstacles are calculated using the fast multipole algorithm. Based on the potential field change results, a kinematic model of the dynamic obstacle is established, including:

[0105] S2231. Use a nonlinear mapping function to convert the spatiotemporal correlation risk matrix into a potential field intensity distribution, and assign the mapped potential field intensity values ​​to the corresponding grid nodes to construct a potential field intensity distribution map.

[0106] S2232. Use an octree structure to spatially divide the potential field intensity distribution map, and perform cluster analysis based on the location information of dynamic obstacles to form a tree-like hierarchical structure.

[0107] S2233. Combining the tree-like hierarchical structure with real-time motion perception data of dynamic obstacles, the fast multipole algorithm is used to calculate the potential field changes caused by the motion of dynamic obstacles and update the potential field intensity distribution.

[0108] S2234. Extract the motion characteristics of the dynamic obstacle from the updated potential field intensity distribution, and establish a kinematic model of the dynamic obstacle based on the motion characteristics.

[0109] Preferably, the formula for the kinematic model is:

[0110] ;

[0111] In the formula, r ( t +1) indicates that the dynamic obstacle occurs in time. t +1 position; r ( t ) indicates the dynamic obstacle in time t Location; t Indicates time; v (t ) represents the velocity of the dynamic obstacle at time t ; Δ t represents the time step; F t ) represents the external force on the dynamic obstacle at time t due to the potential field change; m represents the mass of the dynamic obstacle.

[0112] It should be noted that in specific embodiments, the spatiotemporal correlation risk matrix is mapped to a potential field intensity distribution, the fast multipole method is used to calculate the potential field change caused by the movement of the dynamic obstacle, and the kinematic model of the dynamic obstacle is established based on the potential field change result, including:

[0113] 1. Potential field intensity mapping: the hyperbolic tangent function is used to map the risk matrix R ( t ) to the potential field intensity;

[0114] ;

[0115] In the formula, φ ( x,y ) represents the potential field intensity at coordinate point x,y ); tanh represents the hyperbolic tangent function; R ( x,y ) represents the risk matrix value at coordinate point x,y ); λ represents a scaling factor; wherein λ = 0.8; experimentally measured obstacle center potential field intensity reaches 0.92 (normalized value), and the edge region decays exponentially.

[0116] 2. Octree space clustering: an octree structure with a depth of 5 is constructed, dynamic clustering is performed according to the obstacle position, and the maximum clustering error is controlled within 0.3 m; the tree node update frequency is positively correlated with the obstacle speed (when v = 0.5 m / s, the update period is 1.2 s).

[0117] 3. Fast multipole update: the FMM algorithm is used to calculate the potential field change, the expansion coefficient p = 6 is set, the calculation complexity is reduced from O(N²) to O(N), the time consumption of a single update is reduced from 48 ms of the traditional method to 8 ms, and the potential field change prediction error is < 5%.

[0118] ​4. Kinematic model construction: Extract the motion characteristics (such as speed, acceleration, displacement, etc.) of each dynamic obstacle from the updated potential field intensity distribution, and establish a kinematic model based on these characteristics; where the change of the potential field affects the acceleration or speed of the obstacle (for example, adjust the motion strategy through the perceived potential field change), then the potential field intensity change is taken as an external force and included in the kinematic model.

[0119] S224, combine the kinematic model with the long short-term memory network to predict the motion trajectory of the obstacle, and optimize the kinematic model through real-time perception data feedback.

[0120] It should be noted that combining the kinematic model with the long short-term memory network to predict the motion trajectory of the obstacle, and optimizing the kinematic model through real-time perception data feedback includes:

[0121] Using historical perception data (such as motion perception data in the past 30 seconds) to train the long short-term memory network model, combining the kinematic model to preliminarily model the motion of the AGV; when the AGV starts to accelerate or change direction during operation, the real-time feedback of speed and acceleration perception data is input into the LSTM model for online training, and the model is adjusted according to the real-time perception data to gradually optimize the trajectory prediction accuracy.

[0122] S23, combine the prediction results of dynamic obstacles with the position information of static obstacles, and use the topological potential field algorithm to divide the working area into passing areas, cautious areas and prohibited areas to identify different dynamic risk levels.

[0123] It should be noted that in specific embodiments, the spatiotemporal correlation risk matrix is mapped to the potential field intensity distribution, the fast multipole method is used to calculate the potential field change caused by the motion of dynamic obstacles, and the kinematic model of dynamic obstacles is established based on the potential field change result.

[0124] 1. Dynamic trajectory prediction: analyze the prediction results of dynamic obstacles to obtain the future 3-second trajectory of the forklift, with an average displacement error of 0.18m and a maximum error of not more than 0.35m.

[0125] 2. Static obstacle centroid extraction: process point cloud perception data through DBSCAN clustering algorithm to extract the position of the shelf (clustering error <0.05m) and construct a static potential field.

[0126] ;

[0127] wherein, U static represents the total potential energy of the static potential field; N represents the number of static obstacles; i represents the index value; Q iIndicates the first i The charge of each obstacle (shelf Q=5, wall Q=10); P i Indicates the first i The location of the obstacle; X i Indicates the location of the reference point; || P i - Q i || indicates the first i The distance between the location of each obstacle and the reference point.

[0128] 3. Dynamic potential field calculation: Combining the predicted trajectory, the dynamic potential field is constructed using the velocity obstacle (VO) method;

[0129] ;

[0130] In the formula, U dynamic This represents the total potential energy of the dynamic potential field. M Indicates the number of dynamic obstacles; j Indicates the index value; γ Indicates the attenuation coefficient ( γ =0.8 is the attenuation coefficient). v j Indicates the first j The speed of the obstacle; τ This represents the safety margin in terms of time (τ=1.5s is the time margin). d min represents the minimum safe distance from the obstacle. d The potential field strength changes abruptly when min=0.5m.

[0131] 4. Risk Area Division: Overall Potential Field U = U static + U dynamic Set thresholds: for the passable area, U < 0.3; for the cautious area, 0.3 ≤ U < 0.7; for the prohibited area, U ≥ 0.7.

[0132] S3. Based on the work area division results, the initial obstacle avoidance path is planned using the random tree algorithm. The initial obstacle avoidance path is then optimized by combining the construction parameters of the aerial work platform to generate the final obstacle avoidance path.

[0133] Preferably, based on the work area division results, an initial obstacle avoidance path is planned using a random tree algorithm. This initial obstacle avoidance path is then optimized by combining it with the construction parameters of the aerial work platform, generating the final obstacle avoidance path, which includes:

[0134] S31, constructing a three-dimensional grid map according to the region division result, and fitting an obstacle distribution boundary by using a Gaussian mixture model to generate an obstacle avoidance constraint boundary.

[0135] It should be noted that the formula of the Gaussian mixture model is:

[0136] ;

[0137] In the formula, p x t , represents a probability density function at a position x ; K represents the number of Gaussian components in the Gaussian mixture model; k represents an index value; π k t , represents the weight of the k th Gaussian component at time t ; N represents a Gaussian distribution; x t represents a position at time t ; μ k t , represents the mean vector of the k th Gaussian component at time t ; k t , represents the covariance matrix of the k th Gaussian component at time t .

[0138] S32, based on the obstacle avoidance constraint boundary, performing path search by using a target bias-based random tree algorithm to generate an initial obstacle avoidance path.

[0139] S33, obtaining the structural parameters of the aerial work platform, defining an optimization objective set containing path length, minimum safety distance, curvature smoothness, and energy consumption coefficient, and dynamically adjusting the weights of the optimization objectives according to the control task requirements by using an entropy weight algorithm.

[0140] S34, according to the weight adjustment result, adjusting the initial obstacle avoidance path by using a particle swarm optimization algorithm to generate a final obstacle avoidance path.

[0141] It should be noted that in specific embodiments, according to the working region division result, the initial obstacle avoidance path is planned by using a random tree algorithm, and the initial obstacle avoidance path is optimized in combination with the structural parameters of the aerial work platform to generate a final obstacle avoidance path, which includes:

[0142] 1. Three-dimensional grid modeling and Gaussian mixture model fitting: constructing a voxel map, and fitting an obstacle boundary by using GMM.​​​​

[0143] 2. Target-biased path search: Set target bias probability and step size, generate initial path length.

[0144] 3. Multi-objective optimization modeling: Obtain UAV parameters; for example, arm length 0.8 m→ safety radius constraint 1.2 m, battery capacity 6S12000 mAh→ energy consumption coefficient 0.05 Wh / m; define objective function;

[0145] ;

[0146] In the formula, J represents the comprehensive target function value; L represents the path length; d min represents the minimum safety distance from the obstacle; κ represents the path curvature; E represents the estimated energy consumption; w 1 、w 2 、w 3 、w 4 all represent weight coefficients.

[0147] 4. Dynamic weight adjustment and particle swarm optimization: Calculate the weight using the entropy weight method;

[0148] ;

[0149] In the formula, w represents the weight; H represents the entropy value.

[0150] For example, particle swarm optimization parameters: particle number 50, iteration 100 times, inertia weight 0.729. The optimized path is shown in Table 1.

[0151] Table 1 Path Comparison Perception Data Table

[0152]

[0153] S4. Decouple the final obstacle avoidance path into vertical lift and horizontal displacement control instructions, execute the corresponding obstacle avoidance behavior mode based on the control instructions, and monitor the execution results to optimize the control instructions.

[0154] Preferably, decoupling the final obstacle avoidance path into vertical lift and horizontal displacement control instructions, executing the corresponding obstacle avoidance behavior mode based on the control instructions, and monitoring the execution results to optimize the control instructions comprises:

[0155] S41. Based on the preset spatial coordinates, decouple the final obstacle avoidance path into vertical and horizontal coordinate sequences, and use polynomial fitting technology to generate the motion profile of the aerial work platform.

[0156] It should be noted that the steps for generating the motion profile of the aerial work platform using polynomial fitting technology are as follows:

[0157] 1. Polynomial fitting: In the horizontal direction, a 5th-order polynomial is used for fitting;

[0158] ;

[0159] ;

[0160] Vertical direction: 7th order polynomial fitting was used;

[0161] ;

[0162] In the formula, x ( t ) represents the horizontal direction over time t Changing coordinates; y ( t ) represents the horizontal direction over time t Changing coordinates; z ( t ) indicates the change in the vertical direction over time t Changing coordinates; a k Indicates the level coefficient; b k Indicates the level coefficient; c k Indicates the vertical coefficient; k Indicates the index value.

[0163] 2. Motion Profile Generation: Calculate motion parameters in each direction, where velocity is... v ( t )= dx / dt ,d y / dt , dz / dt acceleration is accelerometer is The constraints include a maximum velocity of 3 m / s and a maximum acceleration of 2 m / s². 2 Continuous jerk (jerk < 0.5 m / s²) 3 ).

[0164] S42. Analyze the motion profile, use fuzzy control algorithm to generate control commands for vertical lifting and horizontal displacement, and accelerate the generation of control commands through a parallel computing framework.

[0165] Preferably, the motion profile is analyzed, and control commands for vertical lifting and horizontal displacement are generated using a fuzzy control algorithm. The generation of these control commands is accelerated using a parallel computing framework, including:

[0166] S421、From the generated motion profile, vertical and horizontal feature parameters are extracted to establish a fuzzy input space, including height deviation, speed and curvature error;

[0167] S422、The fuzzy input space is combined with the fuzzy control algorithm to construct a fuzzy control model containing an input layer, a membership calculation layer and a fuzzy reasoning layer, and a rule reduction layer is introduced to optimize the fuzzy control model;

[0168] S423, the optimized fuzzy control model is used to process vertical and horizontal feature parameters to generate vertical lifting control instructions, and combined with the preset horizontal control rules, horizontal displacement control instructions are obtained;

[0169] S424, the vertical lifting and horizontal displacement control instructions are distributed to different processing threads, and the multi-core processor and parallel computing framework are used to accelerate the generation of control instructions.

[0170] It should be noted that in specific embodiments, analyzing the motion profile, generating vertical lifting and horizontal displacement control instructions using a fuzzy control algorithm, and accelerating the generation of control instructions through a parallel computing framework include:

[0171] 1. Feature parameter extraction and fuzzy space establishment:

[0172] Extract motion profile features: height deviation Δ h (actual height-reference height); speed deviation Δ v (actual speed-preset speed); curvature error κ e (actual curvature-desired curvature);

[0173] Establish fuzzy input space: Δ h domain, [-3, 3]m; Δ v domain, [-1, 1]m / s; κ e domain, [-0.2, 0.2]; use a triangular membership function with an overlap rate of 0.3.

[0174] 2. Fuzzy control model construction and optimization: design 49 initial rules (7 input combinations x 7 output combinations); introduce a rule reduction layer, remove redundant rules based on PCA analysis, and use a genetic algorithm to optimize the rule base. After optimization, the number of rules is reduced to 19, and the calculation amount is reduced by 58%.

[0175] 3. Control instruction generation and parallel processing: vertical control instruction;

[0176] ;

[0177] wherein,u z represents the control instruction in the vertical direction; K p represents the proportional gain; Δ h represents the height deviation; K d represents the differential gain.

[0178] Vertical control instruction:

[0179] ;

[0180] wherein, u xy represents the control instruction in the horizontal direction; PID represents the proportional-integral-differential controller; Δ p represents the height deviation; Δ v represents the speed deviation; f Δ pdt represents the part in the integral term, representing the integral of the position deviation over time.

[0181] 4. Quantitative evaluation index: define the comprehensive control performance index:

[0182] ; obtain: average height tracking error <0.4m; speed response delay <80ms; CPU+GPU heterogeneous computing acceleration ratio 4.2x.

[0183] S43, according to the generated control instruction, select and execute the corresponding obstacle avoidance behavior mode, and the obstacle avoidance behavior mode includes an emergency braking model and a dynamic bypass behavior mode.

[0184] S44, real-time monitoring of the execution effect of the selected obstacle avoidance behavior mode, and optimizing the control instruction based on the monitoring result.

[0185] It needs to be pointed out that in specific embodiments, the final obstacle avoidance path is decoupled into control instructions of vertical lifting and horizontal displacement, the corresponding obstacle avoidance behavior mode is executed based on the control instruction, and the execution result is monitored to optimize the control instruction, which includes:

[0186] 1. Motion profile generation: decompose the final obstacle avoidance path into vertical coordinate sequence ( z i ) and horizontal coordinate sequence ( x i , y i ); polynomial fitting is performed on the vertical direction (7th order polynomial, fitting error <0.15m) and the horizontal direction (5th order polynomial, fitting error <0.2m) to generate the profile.

[0187] 2. Fuzzy control instruction generation: Feature parameters, height deviation Δ h (Domain of discourse [-2,2]m), Horizontal position deviation Δ p (Domain of discourse [-3,3]m), velocity deviation Δ v (Domain of discourse [-1,1] m / s); Fuzzy rule base, 3-input × 3-output system, 27 rules in total, using triangular membership functions, overlap rate 0.4; Parallel computing acceleration: CUDA kernel function to implement fuzzy inference; GPU speedup 3.8x.

[0188] 3. Obstacle avoidance behavior pattern execution: |Δ h |>1.2m→Emergency braking (decelerate to 0.5m / s);|Δ p |>2m→Dynamic detour (triggers local replanning); others→normal tracking (PID control); behavior mode switching delay <50ms (measured value).

[0189] 4. Real-time monitoring and optimization:

[0190] Monitoring indicators: obstacle avoidance success rate (number of successful avoidances / total number of avoidances); control command response time; trajectory tracking error; optimization strategy: update fuzzy rule weights every 10 control cycles; use reinforcement learning to dynamically adjust PID parameters.

[0191] like Figure 2 As shown, according to another embodiment of the present invention, an intelligent obstacle avoidance control system for aerial work platforms is also provided, the system comprising:

[0192] The sensor data acquisition and spatial construction module 1, obstacle recognition and region division module 2, obstacle avoidance path generation module 3, and control command and execution optimization module 4 are connected in sequence.

[0193] The sensing data acquisition and spatial construction module 1 is used to collect sensing data of the working area of ​​the aerial work platform and construct a three-dimensional spatial map based on the sensing data.

[0194] The obstacle recognition and region division module 2 is used to identify dynamic and static obstacles in the three-dimensional spatial map and to classify the risk level of the work area by combining the topological potential field algorithm.

[0195] The obstacle avoidance path generation module 3 is used to plan the initial obstacle avoidance path based on the work area division results using a random tree algorithm, optimize the initial obstacle avoidance path in combination with the construction parameters of the aerial work platform, and generate the final obstacle avoidance path.

[0196] The control command and execution optimization module 4 is used to decouple the final obstacle avoidance path into control commands for vertical lifting and horizontal displacement, execute the corresponding obstacle avoidance behavior mode based on the control commands, and monitor the execution results to optimize the control commands.

[0197] In summary, by means of the above technical solutions of the present application, by constructing a three-dimensional space map, not only the geometric characteristics of static obstacles can be captured, but also the motion characteristics of dynamic obstacles can be analyzed by using time sequence perception data, so that the behavior modeling of dynamic obstacles is realized, a comprehensive and accurate environment cognition basis is provided for path planning, and the safety and efficiency of path planning are greatly improved. By combining digital twin technology and potential field analysis algorithm, a motion prediction model of dynamic obstacles is established, and a dynamic risk area division of the working space is carried out by using a topological potential field algorithm, so that the path planner can identify and predict the change of the risk area in advance, thereby providing a valuable time window for path adjustment, and ensuring safer and more efficient path decision. By using a model predictive control framework to realize vertical and horizontal decoupling control of motion instructions, and combining with platform state feedback to construct a closed-loop optimization system, not only the stability of the system is ensured, but also the flexibility and energy efficiency ratio of obstacle avoidance behavior are significantly improved, so that the obstacles can be avoided more agilely and efficiently while maintaining stability, and the energy use efficiency is optimized.

[0198] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An intelligent obstacle avoidance control method for a high-altitude work platform, characterized in that, The method comprises: S1, collecting perception data of the working area of the aerial work platform, and constructing a three-dimensional space map based on the perception data; S2, identifying dynamic and static obstacles in the three-dimensional space map, and combining a topological potential field algorithm to divide the working area into risk levels; S3, according to the working area division result, using a random tree algorithm to plan an initial obstacle avoidance path, combining the structural parameters of the aerial work platform to optimize the initial obstacle avoidance path, and generating a final obstacle avoidance path; S4, decoupling the final obstacle avoidance path into vertical lifting and horizontal displacement control instructions, executing corresponding obstacle avoidance behavior modes based on the control instructions, and monitoring the execution results to optimize the control instructions; Wherein, the S2 comprises: S21, based on the three-dimensional space map, using a deep learning algorithm to identify and classify dynamic and static obstacles in the working area, and obtaining position information of the dynamic and static obstacles; S22, using digital twin technology to construct a digital twin model, and combining a potential field analysis algorithm to predict the motion trajectory of the dynamic obstacle; S23, combining the prediction result of the dynamic obstacle and the position information of the static obstacle, using a topological potential field algorithm to divide the working area into a passing area, a cautious area and a prohibited area to identify different risk levels; The S22 comprises: S221, based on the position information of the dynamic obstacle, using digital twin technology to construct a digital twin model, and updating the motion state of the dynamic obstacle through fused real-time perception data; S222, performing time-frequency analysis on the motion perception data in the digital twin model, extracting motion cycle characteristics, and combining a graph neural network to mine the interaction topological relationship between dynamic obstacles, generating a spatio-temporal correlation risk matrix to quantify dynamic trends and potential risks; S223, mapping the spatio-temporal correlation risk matrix into a potential field intensity distribution, using a fast multipole algorithm to calculate the potential field change caused by the motion of the dynamic obstacle, and establishing a kinematic model of the dynamic obstacle based on the potential field change result; S224, combining the kinematic model and a long short-term memory network to predict the motion trajectory of the obstacle, and optimizing the kinematic model through real-time perception data feedback.

2. The intelligent obstacle avoidance control method for aerial work platforms according to claim 1, characterized in that, The collection of perception data of the working area of the aerial work platform, and the construction of a three-dimensional space map based on the perception data comprises: S11, collecting perception data of the working area of the aerial work platform through multi-modal sensors, and performing space-time synchronization and coordinate alignment on the collected perception data; S12, using Kalman filtering algorithm to fuse the perception data after space-time synchronization and coordinate alignment, and combining a density-based clustering algorithm to denoise the fused perception data; S13, dividing the denoised perception data into voxel grids, extracting the point centroid in each voxel as a target point, and constructing a three-dimensional space map by splicing the target points.

3. The intelligent obstacle avoidance control method for aerial work platforms of claim 2, wherein, The time-frequency analysis on the motion perception data in the digital twin model, the extraction of motion cycle characteristics, and the combination of a graph neural network to mine the interaction topological relationship between dynamic obstacles to generate a spatio-temporal correlation risk matrix to quantify dynamic trends and potential risks comprise: S2221, time-frequency analysis of motion perception data in the digital twin model is performed using short-time Fourier transform to extract the periodic characteristics of the motion perception data and obtain the motion period characteristics of the dynamic obstacles; S2222, each dynamic obstacle is defined as a node and the interaction relationship between the obstacles is defined as an edge based on the motion perception data of the dynamic obstacles to construct an interaction topology graph; S2223, the interaction topology graph is learned using a graph neural network to identify and analyze the topology structure and dynamic interaction mode between the dynamic obstacles and capture the spatio-temporal interaction information between the dynamic obstacles; S2224, the potential risks between the dynamic obstacles are calculated by combining the motion period characteristics and the spatio-temporal interaction information, a spatio-temporal correlation risk matrix is constructed, and the dynamic changes and potential risks of the obstacles are predicted by analyzing the change trend of the spatio-temporal correlation risk matrix.

4. The intelligent obstacle avoidance control method for aerial work platforms of claim 3, wherein, The mapping of the spatio-temporal correlation risk matrix to the potential field intensity distribution, the calculation of the potential field changes caused by the motion of the dynamic obstacles using the fast multipole algorithm, and the establishment of the kinematic model of the dynamic obstacles based on the potential field change results include: S2231, the spatio-temporal correlation risk matrix is converted into a potential field intensity distribution using a nonlinear mapping function, and the mapped potential field intensity values are assigned to the corresponding grid nodes to construct a potential field intensity distribution map; S2232, the potential field intensity distribution map is spatially divided using an octree structure, and clustering analysis is performed based on the position information of the dynamic obstacles to form a tree-like hierarchical structure; S2233, the tree-like hierarchical structure is combined with the real-time motion perception data of the dynamic obstacles, the fast multipole algorithm is used to calculate the potential field changes caused by the motion of the dynamic obstacles, and the potential field intensity distribution is updated; S2234, the motion characteristics of the dynamic obstacles are extracted from the updated potential field intensity distribution, and the kinematic model of the dynamic obstacles is established based on the motion characteristics.

5. The intelligent obstacle avoidance control method for aerial work platforms according to claim 4, characterized in that, The formula of the kinematic model is: ； wherein, r t +1) denotes the position of the dynamic obstacle at time t +1; r t ) denotes the position of the dynamic obstacle at time t t denotes time; v t ) denotes the velocity of the dynamic obstacle at time t t denotes the time step; F t ) denotes the external force on the dynamic obstacle at time t m denotes the mass of the dynamic obstacle.​​​​​​​ 6. The intelligent obstacle avoidance control method for aerial work platforms of claim 1, wherein, The initial obstacle avoidance path is planned using a random tree algorithm based on the work area division result, the initial obstacle avoidance path is optimized based on the structural parameters of the aerial work platform, and the final obstacle avoidance path is generated, which includes: S31, a three-dimensional grid map is constructed based on the region division result, and a Gaussian mixture model is used to fit the obstacle distribution boundary to generate an obstacle avoidance constraint boundary; S32, based on the obstacle avoidance constraint boundary, a random tree algorithm based on target bias is used for path search to generate an initial obstacle avoidance path; S33, the structural parameters of the aerial work platform are obtained, an optimization target set including path length, minimum safety distance, curvature smoothness and energy consumption coefficient is defined, and the weights of the optimization targets are dynamically adjusted according to the control task requirements using an entropy weight algorithm; S34, according to the weight adjustment result, the initial obstacle avoidance path is adjusted using a particle swarm optimization algorithm to generate a final obstacle avoidance path.

7. The intelligent obstacle avoidance control method for aerial work platforms of claim 1, wherein, The final obstacle avoidance path is decoupled into vertical lifting and horizontal displacement control instructions, the corresponding obstacle avoidance behavior mode is executed based on the control instructions, and the execution result is monitored to optimize the control instructions, which includes: S41, based on the preset spatial coordinates, the final obstacle avoidance path is decoupled into vertical and horizontal coordinate sequences, and a polynomial fitting technique is used to generate the motion profile of the aerial work platform; S42, analyze the motion profile, generate vertical lifting and horizontal displacement control instructions using a fuzzy control algorithm, and accelerate the generation of control instructions through a parallel computing framework; S43, according to the generated control instructions, select and execute the corresponding obstacle avoidance behavior mode, which includes an emergency braking model and a dynamic detour behavior mode; S44, real-time monitoring of the execution effect of the selected obstacle avoidance behavior mode, and optimizing the control instructions based on the monitoring results.

8. The intelligent obstacle avoidance control method for aerial work platforms of claim 7, wherein, The analysis of the motion profile, the generation of the vertical lifting and horizontal displacement control instructions using the fuzzy control algorithm, and the acceleration of the generation of the control instructions through the parallel computing framework include: S421, extract vertical and horizontal feature parameters from the generated motion profile, establish a fuzzy input space, and the feature parameters include height deviation, speed and curvature error; S422, combine the fuzzy input space and the fuzzy control algorithm to construct a fuzzy control model including an input layer, a membership degree calculation layer and a fuzzy reasoning layer, and introduce a rule reduction layer to optimize the fuzzy control model; S423, use the optimized fuzzy control model to process the vertical and horizontal feature parameters, generate vertical lifting control instructions, and combine the preset horizontal control rules to obtain horizontal displacement control instructions; S424, distribute the vertical lifting and horizontal displacement control instructions to different processing threads, and use multi-core processors and parallel computing frameworks to accelerate the generation of control instructions.

9. An intelligent obstacle avoidance control system for aerial work platforms, for implementing the intelligent obstacle avoidance control method for aerial work platforms according to any one of claims 1-8, characterized in that, The system includes: a perception data acquisition and space construction module for acquiring perception data of the working area of the aerial work platform and constructing a three-dimensional space map based on the perception data; an obstacle identification and region division module for identifying dynamic and static obstacles in the three-dimensional space map and dividing the working area into risk levels using a topological potential field algorithm; an obstacle avoidance path generation module for planning an initial obstacle avoidance path using a random tree algorithm based on the working area division results, optimizing the initial obstacle avoidance path based on the structure parameters of the aerial work platform, and generating a final obstacle avoidance path; a control instruction and execution optimization module for decoupling the final obstacle avoidance path into vertical lifting and horizontal displacement control instructions, executing the corresponding obstacle avoidance behavior mode based on the control instructions, and monitoring the execution results to optimize the control instructions.

Images