CN120953386B - Intelligent control method and system for obstacle avoidance of aerial working platform - Google Patents
Intelligent control method and system for obstacle avoidance of aerial working platformInfo
- Publication number
- CN120953386B CN120953386B CN202511465933.2A CN202511465933A CN120953386B CN 120953386 B CN120953386 B CN 120953386B CN 202511465933 A CN202511465933 A CN 202511465933A CN 120953386 B CN120953386 B CN 120953386B
- Authority
- CN
- China
- Prior art keywords
- dynamic
- obstacle
- obstacle avoidance
- motion
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an intelligent control method and system for obstacle avoidance of an aerial working platform, and relates to the field of aerial working platforms. The method comprises the steps of identifying dynamic and static barriers in a three-dimensional space map, carrying out risk grade division on a working area by combining a topological potential field algorithm, planning an initial obstacle avoidance path by utilizing a random tree algorithm according to a working area division result, optimizing the initial obstacle avoidance path by combining construction parameters of an aerial working platform to generate a final obstacle avoidance path, decoupling the final obstacle avoidance path into control instructions of vertical lifting and horizontal displacement, executing corresponding obstacle avoidance behavior modes based on the control instructions, and monitoring an execution result to optimize the control instructions. According to the invention, by constructing the three-dimensional space map, the behavior modeling of the dynamic obstacle is realized, and the safety and the efficiency of path planning are improved.
Description
Technical Field
The invention relates to the field of aerial work platforms, in particular to an intelligent obstacle avoidance control method and system for an aerial work platform.
Background
An aerial platform (Aerial Work Platform, AWP), also commonly referred to as an aerial vehicle or lift platform, is a device specifically designed for aerial operations. The equipment is widely applied to the fields of building construction, power maintenance, bridge detection, billboard maintenance, warehouse operation and the like. The aerial working platform can provide a safe and stable platform for operators so that the operators can finish various tasks such as building facade maintenance, equipment installation, cleaning work and the like at high altitude. Due to the specificity of overhead operations, the operating environment is often complex and variable, filling up various obstacles and potentially dangerous areas, which makes ensuring operational safety a great challenge.
Most of the traditional aerial work platforms rely on manual operation, and operators avoid obstacles and accidents through observation of the surrounding environment. This approach, while effective, has significant limitations. First, the attention of the operator is susceptible to fatigue, long working hours, complex environments or sudden conditions, and thus potential obstacles or risks may be ignored, resulting in the occurrence of safety accidents. In addition, because the human eyes have limited judgment on distance, angle and space, the success rate and accuracy of artificial obstacle avoidance are not high in narrow or crowded environments.
At present, an obstacle avoidance technology of an aerial working platform has advanced to a certain extent, and mainly relies on sensor detection and a basic obstacle avoidance algorithm to realize automatic obstacle avoidance. Many devices detect surrounding obstacles by equipping them with sensors such as ultrasonic sensors, lidar, etc. However, these sensors have certain limitations in complex environments, have limited detection range and accuracy, and are susceptible to interference from environmental factors such as light, weather changes, or surface reflection. In addition, conventional obstacle avoidance algorithms are generally based on a preset working environment model, and are mainly optimized for static obstacles and relatively simple terrains, and for dynamic obstacles and complex and constantly changing terrains, the adaptability and the coping ability of the algorithms are poor, and comprehensive and efficient obstacle avoidance protection cannot be provided.
For the problems in the related art, no effective solution has been proposed at present.
Disclosure of Invention
Aiming at the problems in the related art, the invention provides an intelligent control method and system for obstacle avoidance of an aerial working platform, which are used for overcoming the technical problems existing in the related art.
For this purpose, the invention adopts the following specific technical scheme:
According to one aspect of the invention, there is provided an intelligent control method for obstacle avoidance of an aerial working platform, the method comprising:
s1, acquiring sensing data of a working area of an aerial working platform, and constructing a three-dimensional space map based on the sensing data;
s2, identifying dynamic and static barriers in the three-dimensional space map, and classifying the risk level of the working area by combining a topological potential field algorithm;
S3, planning an initial obstacle avoidance path by utilizing a random tree algorithm according to a working area division result, and optimizing the initial obstacle avoidance path by combining construction parameters of an aerial work platform to generate a final obstacle avoidance path;
s4, decoupling the final obstacle avoidance path into control instructions of vertical lifting and horizontal displacement, executing a corresponding obstacle avoidance behavior mode based on the control instructions, and monitoring an execution result to optimize the control instructions;
Wherein S2 includes:
s21, identifying and classifying dynamic and static obstacles in a working area by using a deep learning algorithm based on a three-dimensional space map, and acquiring position information of the dynamic and static obstacles;
S22, constructing a digital twin model by using a digital twin technology, and predicting the motion trail of the dynamic obstacle by combining a potential field analysis algorithm;
S23, combining the prediction result of the dynamic obstacle and the position information of the static obstacle, and dividing the working area into a passing area, a cautious area and a forbidden area by using a topological potential field algorithm so as to identify different dynamic risk levels.
Optionally, collecting sensing data of a working area of the aerial work platform, and constructing the three-dimensional space map based on the sensing data includes:
S11, acquiring sensing data of a working area of the aerial working platform through a multi-mode sensor, and performing space-time synchronization and coordinate alignment on the acquired sensing data;
S12, fusion processing is carried out on the perception data after space-time synchronization and coordinate alignment by using a Kalman filtering algorithm, and denoising is carried out on the fused perception data by combining a density-based clustering algorithm;
S13, dividing the denoised perception data into voxel grids, extracting point centroids in each voxel as target points, and constructing a three-dimensional space map through target point stitching.
Optionally, constructing a digital twin model by using a digital twin technology, and predicting the motion trail of the dynamic obstacle by combining a potential field analysis algorithm comprises:
S221, constructing a digital twin model by using a digital twin technology based on the position information of the dynamic obstacle, and updating the motion state of the dynamic obstacle through the fused real-time sensing data;
S222, performing time-frequency analysis on motion perception data in the digital twin model, extracting motion cycle characteristics, and mining interaction topological relations among dynamic obstacles by combining a graph neural network to generate a space-time association risk matrix so as to quantify dynamic trends and potential risks;
S223, mapping the space-time associated risk matrix into potential field intensity distribution, calculating potential field change caused by dynamic obstacle movement by using a fast multipole sub-algorithm, and establishing a dynamic model of the dynamic obstacle based on a potential field change result;
S224, combining the kinematic model and the long-term and short-term memory network, predicting the motion trail of the obstacle, and feeding back and optimizing the kinematic model through real-time sensing data.
Optionally, performing time-frequency analysis on motion perception data in the digital twin model, extracting motion cycle characteristics, and mining interaction topological relations among dynamic obstacles in combination with a graph neural network to generate a space-time associated risk matrix, so as to quantify dynamic trends and potential risks, wherein the steps include:
S2221, performing time-frequency analysis on motion perception data in a digital twin model by utilizing short-time Fourier transform, and extracting periodic characteristics of the motion perception data to obtain motion periodic characteristics of a dynamic obstacle;
s2222, defining each dynamic obstacle as a node according to the motion perception data of the dynamic obstacle, defining the interaction relation between the obstacles as edges, and constructing an interaction topological graph;
s2223, learning an interaction topological graph by utilizing a graph neural network, identifying and analyzing a topological structure and a dynamic interaction mode between dynamic obstacles, and capturing space-time interaction information between the dynamic obstacles;
S2224, combining the motion cycle characteristics and the space-time interaction information, calculating potential risks among the dynamic obstacles, constructing a space-time associated risk matrix, and predicting the dynamic changes and the potential risks of the obstacles by analyzing the change trend of the space-time associated risk matrix.
Optionally, mapping the space-time associated risk matrix into a potential field intensity distribution, calculating a potential field change caused by motion of the dynamic obstacle by using a fast multipole sub-algorithm, and establishing a kinematic model of the dynamic obstacle based on a potential field change result comprises:
s2231, converting the space-time associated risk matrix into potential field intensity distribution by using a nonlinear mapping function, and distributing the mapped potential field intensity values to corresponding grid nodes to construct a potential field intensity distribution map;
S2232, performing space division on a potential field intensity distribution map by utilizing an octree structure, and performing cluster analysis according to the position information of the dynamic obstacle to form a tree-shaped hierarchical structure;
s2233, combining the tree-shaped hierarchical structure with real-time motion sensing data of the dynamic obstacle, calculating potential field change caused by motion of the dynamic obstacle by using a fast multipole sub-algorithm, and updating potential field intensity distribution;
s2234, extracting motion characteristics of the dynamic obstacle from the updated potential field intensity distribution, and establishing a kinematic model of the dynamic obstacle based on the motion characteristics.
Alternatively, the kinematic model is formulated as:
;
Where 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 speed of the dynamic obstacle at time t, Δt represents the time step, F (t) represents the external force caused by the potential field to which the dynamic obstacle is subjected at time t, and m represents the mass of the dynamic obstacle.
Optionally, according to the division result of the working area, planning an initial obstacle avoidance path by using a random tree algorithm, and optimizing the initial obstacle avoidance path in combination with the construction parameters of the aerial work platform, wherein generating a final obstacle avoidance path comprises:
s31, constructing a three-dimensional grid map according to the regional division result, fitting an obstacle distribution boundary by using a Gaussian mixture model, and generating an obstacle avoidance constraint boundary;
s32, based on the obstacle avoidance constraint boundary, performing path search by using a random tree algorithm based on target deflection, and generating an initial obstacle avoidance path;
S33, acquiring construction parameters of an aerial working platform, defining an optimization target set comprising path length, minimum safe distance, curvature smoothness and energy consumption coefficient, and dynamically adjusting the weight of the optimization target according to the control task requirement by utilizing an entropy weight algorithm;
And S34, adjusting the initial obstacle avoidance path by using a particle swarm optimization algorithm according to the weight adjustment result, and generating a final obstacle avoidance path.
Optionally, decoupling the final obstacle avoidance path into control instructions of vertical lift and horizontal displacement, executing a corresponding obstacle avoidance behavior mode based on the control instructions, and monitoring an execution result to optimize the control instructions includes:
S41, decoupling a final obstacle avoidance path into a vertical coordinate sequence and a horizontal coordinate sequence based on preset space coordinates, and generating a motion profile of an aerial working platform by using a polynomial fitting technology;
S42, analyzing the motion profile, generating control instructions of vertical lifting and horizontal displacement by using a fuzzy control algorithm, and accelerating the generation of the control instructions by parallel computing frames;
S43, selecting and executing a corresponding obstacle avoidance behavior mode according to the generated control instruction, wherein the obstacle avoidance behavior mode comprises an emergency braking model and a dynamic detour behavior mode;
S44, monitoring the execution effect of the selected obstacle avoidance behavior mode in real time, and optimizing the control instruction based on the monitoring result.
Optionally, analyzing the motion profile, generating control instructions of vertical lift and horizontal displacement by using a fuzzy control algorithm, and accelerating the generation of the control instructions by parallel computing frames includes:
s421, extracting vertical and horizontal characteristic parameters from the generated motion profile, and establishing a fuzzy input space, wherein the characteristic parameters comprise height deviation, speed and curvature error;
s422, constructing a fuzzy control model comprising an input layer, a membership calculation layer and a fuzzy reasoning layer by combining a fuzzy input space and a fuzzy control algorithm, and introducing a rule reduction layer to optimize the fuzzy control model;
s423, processing vertical and horizontal characteristic parameters by using the optimized fuzzy control model, generating a vertical lifting control instruction, and combining a preset horizontal control rule to obtain a horizontal displacement control instruction;
S424, distributing the vertical lifting and horizontal displacement control instructions to different processing threads, and accelerating the generation of the control instructions by utilizing the multi-core processor and the parallel computing framework.
According to another aspect of the present invention, there is also provided an intelligent control system for obstacle avoidance of an aerial work platform, the system comprising:
The sensing data acquisition and space construction module is used for acquiring sensing data of the working area of the aerial working platform and constructing a three-dimensional space map based on the sensing data;
The obstacle recognition and region division module is used for recognizing dynamic and static obstacles in the three-dimensional space map and carrying out risk classification on the working region by combining a topological potential field algorithm;
The obstacle avoidance path generation module is used for planning an initial obstacle avoidance path by utilizing a random tree algorithm according to the division result of the working area, and optimizing the initial obstacle avoidance path by combining the construction parameters of the aerial work platform to generate a final obstacle avoidance path;
The control instruction and execution optimization module is used for decoupling the final obstacle avoidance path into control instructions of vertical lifting and horizontal displacement, executing corresponding obstacle avoidance behavior modes based on the control instructions, and monitoring the execution results to optimize the control instructions.
The beneficial effects of the invention are as follows:
1. According to the invention, by constructing the three-dimensional space map, the geometric characteristics of the static obstacle can be captured, and the motion characteristics of the dynamic obstacle can be analyzed by utilizing time sequence perception data, so that the behavior modeling of the dynamic obstacle is realized, a comprehensive and accurate environment cognition basis is provided for path planning, and the safety and efficiency of path planning are improved.
2. According to the invention, the digital twin technology and the potential field analysis algorithm are combined to establish the motion prediction model of the dynamic obstacle, and the topological potential field algorithm is utilized to divide the dynamic risk area of the working space, so that the path planner can identify and pre-judge the change of the risk area in advance, thereby providing a precious time window for path adjustment and ensuring safer and more efficient path decision.
3. According to the invention, the vertical and horizontal decoupling control of the motion instruction is realized by adopting the model predictive control framework, and the closed-loop optimization system is constructed by combining the platform state feedback, so that the stability of the system is ensured, the flexibility and the energy efficiency ratio of the obstacle avoidance behavior are improved, the obstacle can be avoided more efficiently while the stability is maintained, and the energy use efficiency is optimized.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of an intelligent control method for obstacle avoidance of an aerial work platform according to an embodiment of the present invention;
Fig. 2 is a schematic block diagram of an intelligent obstacle avoidance control system for an overhead working platform according to an embodiment of the invention.
In the figure:
1. the system comprises a perception data acquisition and space construction module, a barrier identification and region division module, a barrier avoidance path generation module and a control instruction and execution optimization module.
Detailed Description
For the purpose of further illustrating the various embodiments, the present invention provides the accompanying drawings, which are a part of the disclosure of the present invention, and which are mainly used for illustrating the embodiments and for explaining the principles of the operation of the embodiments in conjunction with the description thereof, and with reference to these matters, it will be apparent to those skilled in the art to which the present invention pertains that other possible embodiments and advantages of the present invention may be practiced.
According to the embodiment of the invention, an intelligent obstacle avoidance control method and system for an aerial working platform are provided.
The invention is further described with reference to the accompanying drawings and the specific embodiments, as shown in fig. 1, an intelligent control method for obstacle avoidance of an aerial work platform according to an embodiment of the invention, the method includes:
s1, acquiring sensing data of a working area of an aerial working platform, and constructing a three-dimensional space map based on the sensing data.
Preferably, collecting sensing data of a working area of the aerial work platform, and constructing a three-dimensional space map based on the sensing data includes:
S11, acquiring sensing data of a working area of the aerial working platform through a multi-mode sensor, and performing space-time synchronization and coordinate alignment on the acquired sensing data;
S12, fusion processing is carried out on the perception data after space-time synchronization and coordinate alignment by using a Kalman filtering algorithm, and denoising is carried out on the fused perception data by combining a density-based clustering algorithm;
S13, dividing the denoised perception data into voxel grids, extracting point centroids in each voxel as target points, and constructing a three-dimensional space map through target point stitching.
It should be noted that, in a specific embodiment, collecting sensing data of a working area of an aerial working platform, and constructing a three-dimensional space map based on the sensing data includes:
Firstly, acquiring 10 ten thousand point cloud data per second through a laser radar, shooting 30 frames of images per second through a camera, recording 100 times of acceleration and angular speed sensing data per second through an IMU, then, using a GPS time synchronization algorithm to ensure that time stamps of all sensors are consistent, aligning a coordinate system of the laser radar and the camera to the coordinate system of the IMU through a calibration system, then, using a Kalman filtering algorithm to fuse the sensing data of the laser radar, the camera and the IMU, estimating the accurate position and the motion track of a platform, denoising the laser radar point cloud sensing data by using a DBSCAN algorithm, removing noise points, retaining about 80% of sensing data points, finally, dividing the denoised point cloud sensing data into voxel grids of 0.1 meter x 0.1 meter, calculating the centroid of each voxel as a target point to form about 50 ten thousand target points, and constructing a three-dimensional symbiotic map of a building site through the stitching of the target points, wherein the map precision reaches the centimeter level.
In addition, the perception data includes visual data, distance and location data, radar data, environmental information data, GPS data, sensor fusion data, and the like.
S2, identifying dynamic and static barriers in the three-dimensional space map, and classifying the risk of the working area by combining a topological potential field algorithm.
Wherein S2 includes:
S21, based on the three-dimensional space map, identifying and classifying the dynamic and static obstacles in the working area by using a deep learning algorithm, and acquiring the position information of the dynamic and static obstacles.
It should be noted that, the static obstacle is usually a fixed object, such as a building, a device, etc., and the characteristics of the obstacle are that the position is relatively fixed and the change is small, the objects can be identified and classified by a deep learning model obtained through pre-training, and the static obstacle is taken as a category during training, and the model learns the characteristics of the static obstacle according to the spatial distribution of the point cloud.
Dynamic obstacles include moving objects, such as pedestrians, vehicles, drones, etc., which are typically characterized by a continuously changing position over time. The point cloud sensing data can be input into a deep learning model according to time sequence, the characteristics of the dynamic obstacle can be captured by utilizing the time difference, and for the dynamic obstacle, the object with larger position change among a plurality of time frames can be identified by utilizing the comparison of multi-frame sensing data through motion detection.
S22, constructing a digital twin model by using a digital twin technology, and predicting the motion trail of the dynamic obstacle by combining a potential field analysis algorithm.
Preferably, constructing a digital twin model by using a digital twin technology, and predicting the motion trail of the dynamic obstacle by combining a potential field analysis algorithm comprises:
S221, constructing a digital twin model by using a digital twin technology based on the position information of the dynamic obstacle, and updating the motion state of the dynamic obstacle through the fused real-time sensing data.
S222, performing time-frequency analysis on motion perception data in the digital twin model, extracting motion cycle characteristics, and mining interaction topological relations among dynamic obstacles by combining a graph neural network to generate a space-time association risk matrix so as to quantify dynamic trends and potential risks.
Preferably, performing time-frequency analysis on motion perception data in the digital twin model, extracting motion cycle characteristics, and mining interaction topological relations among dynamic obstacles in combination with a graph neural network to generate a space-time associated risk matrix so as to quantify dynamic trends and potential risks, wherein the method comprises the following steps:
s2221, performing time-frequency analysis on the motion perception data in the digital twin model by utilizing short-time Fourier transform, and extracting periodic characteristics of the motion perception data to obtain the motion periodic characteristics of the dynamic obstacle.
S2222, defining each dynamic obstacle as a node according to the motion perception data of the dynamic obstacle, defining the interaction relation between the obstacles as edges, and constructing an interaction topological graph.
It should be noted that, in the specific embodiment, it is assumed that there are 4 dynamic obstacles A, B, C, D, and their position information and speed are recorded, which are calculated according to the calculation:
The distance between A and B is 0.5 m, the speed difference is 0.2m/s, the edge is constructed, the distance between A and C is 2m, the speed difference is 1m/s, the edge is not constructed, the distance between B and D is 1m, the speed difference is 0.3m/s, the edge is constructed, and the interactive topological graph is A-B and B-D.
S2223, learning the interaction topological graph by utilizing the graph neural network, identifying and analyzing the topological structure and the dynamic interaction mode among the dynamic obstacles, and capturing the space-time interaction information among the dynamic obstacles.
S2224, combining the motion cycle characteristics and the space-time interaction information, calculating potential risks among the dynamic obstacles, constructing a space-time associated risk matrix, and predicting the dynamic changes and the potential risks of the obstacles by analyzing the change trend of the space-time associated risk matrix.
It should be noted that, in a specific embodiment, performing time-frequency analysis on motion perception data in a digital twin model, extracting motion cycle characteristics, and mining interaction topological relations among dynamic obstacles in combination with a graph neural network to generate a space-time associated risk matrix, so as to quantify dynamic trends and potential risks, including:
1. Motion cycle feature extraction, namely performing short-time Fourier transform on motion perception data, and detecting a significant energy peak in a frequency domain based on a transformation process, wherein the AGV1 cycle is 0.83+/-0.05 s (corresponding to start-stop cycle), the AGV2 cycle is 1.21+/-0.03 s (corresponding to loading and unloading operation), and the personnel gait cycle is 1.12+/-0.1 s.
2. Establishing an adjacency matrix based on a minimum distance threshold value between barriers, and calculating edge weights by adopting a Gaussian kernel function;
;
Where 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, where σ=0.3m, and a time-varying undirected graph G (t) e R 5×5 is constructed.
3. And training the graph neural network, namely adopting a space-time graph convolutional network, setting a 3-layer graph convolutional layer (64 channels per layer), and adopting mask cross entropy to a loss function, wherein the training set/verification set is divided according to 8:2.
4. The risk matrix construction and analysis are carried out, wherein motion cycle characteristics (Fourier coefficients) and attention force weights are fused, a space-time correlation risk matrix G (t) E R 5×5 is constructed, R value sudden increase (collision risk increase) when the AGV1 interacts with personnel is obtained, risk propagation presents exponential decay characteristics, the influence radius is about 1.8m, and a risk event can be predicted in advance by 0.8-1.2s through matrix characteristic value analysis.
5. The prediction effect is that the real track is compared with the prediction risk area, the system successfully pre-warns 4 times of potential collision (pre-warning time window is 1.5-2.3 s), and the false positive rate is controlled at 3.1%. The risk matrix visualization result shows that the overlapping degree of the high-risk area and the actual movement track of the obstacle reaches 89.4%.
And S223, mapping the space-time associated risk matrix into potential field intensity distribution, calculating potential field change caused by the motion of the dynamic obstacle by using a fast multipolar sub-algorithm, and establishing a kinematic model of the dynamic obstacle based on a potential field change result.
Preferably, mapping the space-time associated risk matrix into a potential field intensity distribution, calculating a potential field change caused by motion of the dynamic obstacle by using a fast multipole sub-algorithm, and establishing a kinematic model of the dynamic obstacle based on a potential field change result comprises:
s2231, converting the space-time associated risk matrix into potential field intensity distribution by using a nonlinear mapping function, and distributing the mapped potential field intensity values to corresponding grid nodes to construct a potential field intensity distribution map;
S2232, performing space division on a potential field intensity distribution map by utilizing an octree structure, and performing cluster analysis according to the position information of the dynamic obstacle to form a tree-shaped hierarchical structure;
s2233, combining the tree-shaped hierarchical structure with real-time motion sensing data of the dynamic obstacle, calculating potential field change caused by motion of the dynamic obstacle by using a fast multipole sub-algorithm, and updating potential field intensity distribution;
s2234, extracting motion characteristics of the dynamic obstacle from the updated potential field intensity distribution, and establishing a kinematic model of the dynamic obstacle based on the motion characteristics.
Preferably, the kinematic model is formulated as:
;
Where 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 speed of the dynamic obstacle at time t, Δt represents the time step, F (t) represents the external force caused by the potential field to which the dynamic obstacle is subjected at time t, and m represents the mass of the dynamic obstacle.
It should be noted that, in a specific embodiment, mapping the space-time associated risk matrix into a potential field intensity distribution, calculating a potential field change caused by the motion of the dynamic obstacle by using a fast multipole sub-algorithm, and establishing a kinematic model of the dynamic obstacle based on a potential field change result includes:
1. Mapping potential field intensity, namely mapping a risk matrix R (t) into potential field intensity by adopting a hyperbolic tangent function;
;
Where phi (x, y) denotes the potential field strength at the coordinate point (x, y), tanh denotes the hyperbolic tangent function, R (x, y) denotes the risk matrix value at the coordinate point (x, y), lambda denotes the scaling factor, where lambda=0.8, and the experimentally measured barrier center potential field strength reaches 0.92 (normalized value), the edge region decays exponentially.
2. Octree spatial clustering, namely constructing an octree structure with depth of 5, carrying out dynamic clustering according to the position of an obstacle, controlling the maximum clustering error within 0.3m, and positively correlating the update frequency of tree nodes with the speed of the obstacle (updating period is 1.2s when v=0.5 m/s).
3. And (3) fast multipole updating, namely calculating potential field change by adopting an FMM algorithm, setting an expansion coefficient p=6, reducing the calculation complexity from O (N2) to O (N), reducing the time consumption of single updating from 48ms to 8ms in the traditional method, and ensuring that the prediction error of the potential field change is less than 5%.
4. The kinematic model is constructed by extracting the motion characteristics (such as speed, acceleration, displacement and the like) of each dynamic obstacle from the updated potential field intensity distribution, and building a kinematic model based on the characteristics, wherein the change of the potential field influences the acceleration or speed of the obstacle (for example, the motion strategy is adjusted through perceived change of the potential field), and then the change of the potential field intensity is taken into the kinematic model as the influence of external force.
S224, combining the kinematic model and the long-term and short-term memory network, predicting the motion trail of the obstacle, and feeding back and optimizing the kinematic model through real-time sensing data.
It should be noted that, combining the kinematic model and the long-short term memory network, predicting the motion trail of the obstacle, and optimizing the kinematic model through real-time sensing data feedback includes:
and when the AGV starts accelerating or changing direction in the running process, the speed and acceleration sensing data fed back in real time are input into the LSTM model for online training, and the model gradually optimizes the track prediction precision according to the real-time sensing data adjustment.
S23, combining the prediction result of the dynamic obstacle and the position information of the static obstacle, and dividing the working area into a passing area, a cautious area and a forbidden area by using a topological potential field algorithm so as to identify different dynamic risk levels.
It should be noted that, in a specific embodiment, mapping the space-time associated risk matrix into a potential field intensity distribution, calculating a potential field change caused by the motion of the dynamic obstacle by using a fast multipole sub-algorithm, and establishing a kinematic model of the dynamic obstacle based on a potential field change result includes:
1. And (3) predicting the dynamic track, namely analyzing the predicted result of the dynamic obstacle, and obtaining the future 3-second track of the forklift, wherein the average displacement error is 0.18m, and the maximum error is not more than 0.35m.
2. Extracting a static obstacle centroid, namely processing point cloud sensing data through a DBSCAN clustering algorithm, extracting the position of a goods shelf (the clustering error is less than 0.05 m), and constructing a static potential field;
;
where U static denotes the total potential of the static potential field, N denotes the number of static obstacles, i denotes the index value, Q i denotes the amount of charge of the i-th obstacle (shelf q=5, wall q=10), P i denotes the position of the i-th obstacle, X i denotes the position of the reference point, and i P i-Qi i denotes the distance between the i-th obstacle position and the reference point.
3. Calculating a dynamic potential field, namely constructing the dynamic potential field by adopting a speed barrier method (VO) in combination with a predicted track;
;
Where U dynamic denotes the total potential of the dynamic potential field, M denotes the number of dynamic obstacles, j denotes the index value, γ denotes the decay factor (γ=0.8 is the decay factor), v j denotes the speed of the jth obstacle, τ denotes the safety margin in time (τ=1.5 s is the time margin), dmin denotes the minimum safety distance from the obstacle, dmin=0.5M is the potential field strength abrupt change.
4. The risk area is divided into a total potential field U=U static+Udynamic, a threshold is set, wherein U is less than 0.3, a cautious area is less than or equal to 0.3 and less than or equal to 0.7, and a forbidden area is more than or equal to 0.7.
S3, planning an initial obstacle avoidance path by utilizing a random tree algorithm according to a working area division result, and optimizing the initial obstacle avoidance path by combining construction parameters of an aerial work platform to generate a final obstacle avoidance path.
Preferably, according to the division result of the working area, planning an initial obstacle avoidance path by using a random tree algorithm, optimizing the initial obstacle avoidance path in combination with construction parameters of an aerial work platform, and generating a final obstacle avoidance path comprises:
S31, constructing a three-dimensional grid map according to the regional division result, fitting an obstacle distribution boundary by using a Gaussian mixture model, and generating an obstacle avoidance constraint boundary.
It should be noted that, the formula of the gaussian mixture model is:
;
Where p (x t) denotes the probability density function at position x, K denotes the number of Gaussian components in the mixture Gaussian model, K denotes the index value, pi k (t) denotes the weight of the kth Gaussian component at time t, N denotes the Gaussian distribution, x t denotes the position of time t, μ k (t) denotes the mean vector of the kth Gaussian component at time t, and Σ k (t) denotes the covariance matrix of the kth Gaussian component at time t.
S32, based on the obstacle avoidance constraint boundary, performing path search by using a random tree algorithm based on target deflection, and generating an initial obstacle avoidance path.
S33, acquiring construction parameters of the aerial working platform, defining an optimization target set comprising path length, minimum safe distance, curvature smoothness and energy consumption coefficient, and dynamically adjusting the weight of the optimization target according to the control task requirement by utilizing an entropy weight algorithm.
And S34, adjusting the initial obstacle avoidance path by using a particle swarm optimization algorithm according to the weight adjustment result, and generating a final obstacle avoidance path.
It should be noted that, in a specific embodiment, according to a division result of a working area, an initial obstacle avoidance path is planned by using a random tree algorithm, and the initial obstacle avoidance path is optimized in combination with a construction parameter of an aerial working platform, and generating a final obstacle avoidance path includes:
1. And (3) fitting the three-dimensional grid modeling and the Gaussian mixture model, namely constructing a voxel map and fitting the obstacle boundary by adopting a GMM.
2. And (3) searching a path based on target deflection, namely setting target deflection probability and step length, and generating an initial path length.
3. Multi-objective optimization modeling, namely acquiring unmanned plane parameters, for example, the length of a horn is 0.8m, the safety radius constraint is 1.2m, the battery capacity is 6S12000mAh, and the energy consumption coefficient is 0.05Wh/m;
;
Wherein J represents the comprehensive objective function value, L represents the path length, dmin represents the minimum safety distance between the obstacle, kappa represents the path curvature, E represents the estimated energy consumption, and w 1、w2、w3、w4 represents the weight coefficient.
4. Dynamic weight adjustment and particle swarm optimization, namely calculating weight by adopting an entropy weight method;
;
Where w represents the weight and H represents the entropy.
For example, the particle swarm optimization parameter, the particle count is 50, the iteration is 100, and the inertia weight is 0.729. The optimized path pairs are shown in table 1.
Table 1 path contrast perception data table
S4, decoupling the final obstacle avoidance path into control instructions of vertical lifting and horizontal displacement, executing corresponding obstacle avoidance behavior modes based on the control instructions, and monitoring an execution result to optimize the control instructions.
Preferably, decoupling the final obstacle avoidance path into control instructions for vertical lift and horizontal displacement, executing a corresponding obstacle avoidance behavior pattern based on the control instructions, and monitoring the execution result to optimize the control instructions includes:
S41, decoupling the final obstacle avoidance path into a vertical coordinate sequence and a horizontal coordinate sequence based on preset space coordinates, and generating a motion profile of the aerial working platform by using a polynomial fitting technology.
It should be noted that the step of generating the motion profile of the aerial work platform by using the polynomial fitting technique is as follows:
1. polynomial fitting, namely, in the horizontal direction, adopting 5 times of polynomial fitting;
;
;
fitting by adopting 7 times of polynomials in the vertical direction;
;
Where x (t) represents a coordinate in a horizontal direction which varies with time t, y (t) represents a coordinate in a horizontal direction which varies with time t, z (t) represents a coordinate in a vertical direction which varies with time t, a k represents a horizontal coefficient, b k represents a horizontal coefficient, c k represents a vertical coefficient, and k represents an index value.
2. Motion profile generation by calculating motion parameters in each direction, wherein the velocity is v (t) =dx/dt, dy/dt, dz/dt, and the acceleration isAcceleration isMaximum speed 3m/s, maximum acceleration 2m/s 2, continuous jerk (jerk <0.5m/s 3) in the constraint.
S42, analyzing the motion profile, generating control instructions of vertical lifting and horizontal displacement by using a fuzzy control algorithm, and accelerating the generation of the control instructions by using a parallel computing framework.
Preferably, analyzing the motion profile, generating control instructions for vertical lift and horizontal displacement using a fuzzy control algorithm, and accelerating the generation of the control instructions by parallel computing frames includes:
s421, extracting vertical and horizontal characteristic parameters from the generated motion profile, and establishing a fuzzy input space, wherein the characteristic parameters comprise height deviation, speed and curvature error;
s422, constructing a fuzzy control model comprising an input layer, a membership calculation layer and a fuzzy reasoning layer by combining a fuzzy input space and a fuzzy control algorithm, and introducing a rule reduction layer to optimize the fuzzy control model;
s423, processing vertical and horizontal characteristic parameters by using the optimized fuzzy control model, generating a vertical lifting control instruction, and combining a preset horizontal control rule to obtain a horizontal displacement control instruction;
S424, distributing the vertical lifting and horizontal displacement control instructions to different processing threads, and accelerating the generation of the control instructions by utilizing the multi-core processor and the parallel computing framework.
It should be noted that, in a specific embodiment, analyzing a motion profile, generating control instructions of vertical lift and horizontal displacement by using a fuzzy control algorithm, and accelerating the generation of the control instructions by parallel computing frames includes:
1. feature parameter extraction and fuzzy space establishment:
Extracting motion profile characteristics, namely height deviation deltah (actual height-reference height), speed deviation deltav (actual speed-preset speed), curvature error kappa e (actual curvature-expected curvature);
establishing fuzzy input space, namely a Deltah domain, [ -3,3] m, a Deltav domain, [ -1,1] m/s, a kappa e domain, [ -0.2,0.2], and adopting a triangular membership function, wherein the overlapping rate is 0.3.
2. The fuzzy control model is constructed and optimized by designing 49 initial rules (7 input combinations are multiplied by 7 output combinations), introducing a rule reduction layer, removing redundant rules based on PCA analysis, optimizing a rule base by adopting a genetic algorithm, reducing the number of optimized rules to 19, and reducing the calculated amount by 58%.
3. Control instruction generation and parallel processing, namely, vertical control instruction;
;
Where u z denotes a control command in the vertical direction, K p denotes a proportional gain, Δh denotes a height deviation, and K d denotes a differential gain.
Vertical control instruction:
;
Where u xy denotes a control command in the horizontal direction, PID denotes a PID controller, Δp denotes a height deviation, Δv denotes a speed deviation, fΔ pdt denotes a part of an integral term, and denotes the integral of a position deviation over time.
4. Quantitative evaluation index defining comprehensive control performance index:
The average height tracking error is less than 0.4m, the speed response delay is less than 80ms, and the CPU+GPU heterogeneous computing acceleration ratio is 4.2x.
S43, selecting and executing a corresponding obstacle avoidance behavior mode according to the generated control instruction, wherein the obstacle avoidance behavior mode comprises an emergency braking model and a dynamic detour behavior mode.
S44, monitoring the execution effect of the selected obstacle avoidance behavior mode in real time, and optimizing the control instruction based on the monitoring result.
It should be noted that, in a specific embodiment, decoupling the final obstacle avoidance path into control instructions of vertical lifting and horizontal displacement, executing a corresponding obstacle avoidance behavior mode based on the control instructions, and monitoring the execution result to optimize the control instructions includes:
1. And generating a motion profile, namely decomposing a final obstacle avoidance path into a vertical coordinate sequence (z i) and a horizontal coordinate sequence (x i,yi), and performing polynomial fitting on a vertical direction (a polynomial of 7 times, a fitting error of <0.15 m) and a horizontal direction (a polynomial of 5 times, a fitting error of <0.2 m) to generate the profile.
2. The fuzzy control instruction is generated by the characteristic parameter, the height deviation delta h (domain [ -2,2] m), the horizontal position deviation delta p (domain [ -3,3] m), the speed deviation delta v (domain [ -1,1] m/s), the fuzzy rule base, the 3 input x 3 output system, 27 rules, the triangle membership function, the overlapping rate 0.4, the parallel computing acceleration, the CUDA kernel function, the GPU acceleration ratio 3.8x.
3. The obstacle avoidance behavior mode is executed by adopting the following steps of I delta h I >1.2m, emergency braking (decelerating to 0.5 m/s), I delta p I >2m, dynamic detouring (triggering local re-planning), the rest, normal tracking (PID control) and behavior mode switching delay of <50ms (measured value).
4. Real-time monitoring and optimizing:
The monitoring indexes comprise obstacle avoidance success rate (successful avoidance times/total avoidance times), control instruction response time, track tracking error, optimization strategy, and PID parameter dynamic adjustment by reinforcement learning, wherein the fuzzy rule weight is updated every 10 control periods.
As shown in fig. 2, according to another embodiment of the present invention, there is also provided an intelligent control system for obstacle avoidance of an aerial work platform, the system comprising:
The sensing data acquisition and space construction module 1, the obstacle recognition and area division module 2, the obstacle avoidance path generation module 3 and the control instruction and execution optimization module 4 are sequentially connected.
The sensing data acquisition and space construction module 1 is used for acquiring sensing data of a working area of the aerial working platform and constructing a three-dimensional space map based on the sensing data;
The obstacle recognition and region division module 2 is used for recognizing dynamic and static obstacles in the three-dimensional space map and dividing the risk level of the working region by combining a topological potential field algorithm;
The obstacle avoidance path generation module 3 is used for planning an initial obstacle avoidance path by utilizing a random tree algorithm according to the division result of the working area, and optimizing the initial obstacle avoidance path by combining with the construction parameters of the aerial work platform to generate a final obstacle avoidance path;
the control instruction and execution optimization module 4 is configured to decouple the final obstacle avoidance path into a control instruction for vertical lifting and horizontal displacement, execute a corresponding obstacle avoidance behavior mode based on the control instruction, and monitor an execution result to optimize the control instruction.
In summary, by means of the technical scheme, by constructing the three-dimensional space map, the geometric characteristics of the static obstacle can be captured, and the motion characteristics of the dynamic obstacle can be analyzed by utilizing time sequence perception data, so that the behavior modeling of the dynamic obstacle 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 a digital twin technology and a potential field analysis algorithm, a motion prediction model of a dynamic obstacle is established, and a topological potential field algorithm is utilized to divide a dynamic risk area of a working space, so that a path planner can recognize and pre-judge the change of the risk area in advance, thereby providing a precious time window for path adjustment and ensuring safer and more efficient path decision. The vertical and horizontal decoupling control of the motion command is realized by adopting the model predictive control framework, and the closed-loop optimization system is constructed by combining the platform state feedback, so that the stability of the system is ensured, the flexibility and the energy efficiency ratio of the obstacle avoidance behavior are obviously improved, the obstacle can be avoided more agilely and efficiently while the stability is maintained, and the energy use efficiency is optimized.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.
Claims (9)
1. An intelligent control method for obstacle avoidance of an aerial working platform is characterized by comprising the following steps:
s1, acquiring sensing data of a working area of an aerial working platform, and constructing a three-dimensional space map based on the sensing data;
s2, identifying dynamic and static barriers in the three-dimensional space map, and classifying the risk level of the working area by combining a topological potential field algorithm;
S3, planning an initial obstacle avoidance path by utilizing a random tree algorithm according to a working area division result, and optimizing the initial obstacle avoidance path by combining construction parameters of an aerial work platform to generate a final obstacle avoidance path;
s4, decoupling the final obstacle avoidance path into control instructions of vertical lifting and horizontal displacement, executing a corresponding obstacle avoidance behavior mode based on the control instructions, and monitoring an execution result to optimize the control instructions;
Wherein, the S2 includes:
s21, identifying and classifying dynamic and static obstacles in a working area by using a deep learning algorithm based on a three-dimensional space map, and acquiring position information of the dynamic and static obstacles;
S22, constructing a digital twin model by using a digital twin technology, and predicting the motion trail of the dynamic obstacle by combining a potential field analysis algorithm;
S23, dividing a working area into a passing area, a cautious area and a forbidden area by using a topological potential field algorithm by combining a prediction result of a dynamic obstacle and position information of a static obstacle so as to identify different risk levels;
The S22 includes:
S221, constructing a digital twin model by using a digital twin technology based on the position information of the dynamic obstacle, and updating the motion state of the dynamic obstacle through the fused real-time sensing data;
S222, performing time-frequency analysis on motion perception data in the digital twin model, extracting motion cycle characteristics, and mining interaction topological relations among dynamic obstacles by combining a graph neural network to generate a space-time association risk matrix so as to quantify dynamic trends and potential risks;
S223, mapping the space-time associated risk matrix into potential field intensity distribution, calculating potential field change caused by dynamic obstacle movement by using a fast multipole sub-algorithm, and establishing a dynamic model of the dynamic obstacle based on a potential field change result;
S224, combining the kinematic model and the long-term and short-term memory network, predicting the motion trail of the obstacle, and feeding back and optimizing the kinematic model through real-time sensing data.
2. The intelligent control method for obstacle avoidance of an aerial work platform according to claim 1, wherein the steps of collecting the perceived data of the working area of the aerial work platform and constructing a three-dimensional space map based on the perceived data comprise:
S11, acquiring sensing data of a working area of the aerial working platform through a multi-mode sensor, and performing space-time synchronization and coordinate alignment on the acquired sensing data;
S12, fusion processing is carried out on the perception data after space-time synchronization and coordinate alignment by using a Kalman filtering algorithm, and denoising is carried out on the fused perception data by combining a density-based clustering algorithm;
S13, dividing the denoised perception data into voxel grids, extracting point centroids in each voxel as target points, and constructing a three-dimensional space map through target point stitching.
3. The intelligent control method for obstacle avoidance of an aerial working platform according to claim 2, wherein the performing time-frequency analysis on the motion perception data in the digital twin model, extracting motion cycle characteristics, and mining the interactive topological relation among the dynamic obstacles by combining the graph neural network to generate a space-time associated risk matrix to quantify the dynamic trend and the potential risk comprises:
S2221, performing time-frequency analysis on motion perception data in a digital twin model by utilizing short-time Fourier transform, and extracting periodic characteristics of the motion perception data to obtain motion periodic characteristics of a dynamic obstacle;
s2222, defining each dynamic obstacle as a node according to the motion perception data of the dynamic obstacle, defining the interaction relation between the obstacles as edges, and constructing an interaction topological graph;
s2223, learning an interaction topological graph by utilizing a graph neural network, identifying and analyzing a topological structure and a dynamic interaction mode between dynamic obstacles, and capturing space-time interaction information between the dynamic obstacles;
S2224, combining the motion cycle characteristics and the space-time interaction information, calculating potential risks among the dynamic obstacles, constructing a space-time associated risk matrix, and predicting the dynamic changes and the potential risks of the obstacles by analyzing the change trend of the space-time associated risk matrix.
4. The intelligent control method for obstacle avoidance of an overhead working platform according to claim 3, wherein mapping the space-time associated risk matrix into potential field intensity distribution, calculating potential field changes caused by motion of the dynamic obstacle by using a fast multipole sub-algorithm, and establishing a kinematic model of the dynamic obstacle based on the potential field change results comprises:
s2231, converting the space-time associated risk matrix into potential field intensity distribution by using a nonlinear mapping function, and distributing the mapped potential field intensity values to corresponding grid nodes to construct a potential field intensity distribution map;
S2232, performing space division on a potential field intensity distribution map by utilizing an octree structure, and performing cluster analysis according to the position information of the dynamic obstacle to form a tree-shaped hierarchical structure;
s2233, combining the tree-shaped hierarchical structure with real-time motion sensing data of the dynamic obstacle, calculating potential field change caused by motion of the dynamic obstacle by using a fast multipole sub-algorithm, and updating potential field intensity distribution;
s2234, extracting motion characteristics of the dynamic obstacle from the updated potential field intensity distribution, and establishing a kinematic model of the dynamic obstacle based on the motion characteristics.
5. The intelligent control method for obstacle avoidance of an aerial work platform according to claim 4, wherein the kinematic model has a formula as follows:
;
Where 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 speed of the dynamic obstacle at time t, Δt represents the time step, F (t) represents the external force caused by the potential field to which the dynamic obstacle is subjected at time t, and m represents the mass of the dynamic obstacle.
6. The intelligent control method for obstacle avoidance of an aerial work platform according to claim 1, wherein the planning an initial obstacle avoidance path by using a random tree algorithm according to the division result of the working area, optimizing the initial obstacle avoidance path in combination with the construction parameters of the aerial work platform, and generating a final obstacle avoidance path comprises:
s31, constructing a three-dimensional grid map according to the regional division result, fitting an obstacle distribution boundary by using a Gaussian mixture model, and generating an obstacle avoidance constraint boundary;
s32, based on the obstacle avoidance constraint boundary, performing path search by using a random tree algorithm based on target deflection, and generating an initial obstacle avoidance path;
S33, acquiring construction parameters of an aerial working platform, defining an optimization target set comprising path length, minimum safe distance, curvature smoothness and energy consumption coefficient, and dynamically adjusting the weight of the optimization target according to the control task requirement by utilizing an entropy weight algorithm;
And S34, adjusting the initial obstacle avoidance path by using a particle swarm optimization algorithm according to the weight adjustment result, and generating a final obstacle avoidance path.
7. The intelligent control method for obstacle avoidance of an aerial work platform according to claim 1, wherein decoupling the final obstacle avoidance path into control instructions for vertical lift and horizontal displacement, executing a corresponding obstacle avoidance behavior pattern based on the control instructions, and monitoring the execution results to optimize the control instructions comprises:
S41, decoupling a final obstacle avoidance path into a vertical coordinate sequence and a horizontal coordinate sequence based on preset space coordinates, and generating a motion profile of an aerial working platform by using a polynomial fitting technology;
S42, analyzing the motion profile, generating control instructions of vertical lifting and horizontal displacement by using a fuzzy control algorithm, and accelerating the generation of the control instructions by parallel computing frames;
s43, selecting and executing a corresponding obstacle avoidance behavior mode according to the generated control instruction, wherein the obstacle avoidance behavior mode comprises an emergency braking model and a dynamic detour behavior mode;
S44, monitoring the execution effect of the selected obstacle avoidance behavior mode in real time, and optimizing the control instruction based on the monitoring result.
8. The intelligent control method for obstacle avoidance of an aerial work platform according to claim 7, wherein the analyzing the motion profile, generating control instructions for vertical lift and horizontal displacement by using a fuzzy control algorithm, and accelerating the generation of the control instructions by calculating the frame in parallel comprises:
S421, extracting vertical and horizontal characteristic parameters from the generated motion profile, and establishing a fuzzy input space, wherein the characteristic parameters comprise height deviation, speed and curvature error;
s422, constructing a fuzzy control model comprising an input layer, a membership calculation layer and a fuzzy reasoning layer by combining a fuzzy input space and a fuzzy control algorithm, and introducing a rule reduction layer to optimize the fuzzy control model;
s423, processing vertical and horizontal characteristic parameters by using the optimized fuzzy control model, generating a vertical lifting control instruction, and combining a preset horizontal control rule to obtain a horizontal displacement control instruction;
S424, distributing the vertical lifting and horizontal displacement control instructions to different processing threads, and accelerating the generation of the control instructions by utilizing the multi-core processor and the parallel computing framework.
9. An intelligent control system for obstacle avoidance of an aerial work platform, for implementing the intelligent control method for obstacle avoidance of an aerial work platform according to any one of claims 1 to 8, comprising:
The sensing data acquisition and space construction module is used for acquiring sensing data of the working area of the aerial working platform and constructing a three-dimensional space map based on the sensing data;
The obstacle recognition and region division module is used for recognizing dynamic and static obstacles in the three-dimensional space map and carrying out risk classification on the working region by combining a topological potential field algorithm;
The obstacle avoidance path generation module is used for planning an initial obstacle avoidance path by utilizing a random tree algorithm according to the division result of the working area, and optimizing the initial obstacle avoidance path by combining the construction parameters of the aerial work platform to generate a final obstacle avoidance path;
The control instruction and execution optimization module is used for decoupling the final obstacle avoidance path into control instructions of vertical lifting and horizontal displacement, executing corresponding obstacle avoidance behavior modes based on the control instructions, and monitoring the execution results to optimize the control instructions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511465933.2A CN120953386B (en) | 2025-10-14 | 2025-10-14 | Intelligent control method and system for obstacle avoidance of aerial working platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511465933.2A CN120953386B (en) | 2025-10-14 | 2025-10-14 | Intelligent control method and system for obstacle avoidance of aerial working platform |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN120953386A CN120953386A (en) | 2025-11-14 |
| CN120953386B true CN120953386B (en) | 2025-12-12 |
Family
ID=97613336
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202511465933.2A Active CN120953386B (en) | 2025-10-14 | 2025-10-14 | Intelligent control method and system for obstacle avoidance of aerial working platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN120953386B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN120560279A (en) * | 2025-07-31 | 2025-08-29 | 宁德师范学院 | Dynamic obstacle avoidance system in unmanned boat path planning |
| CN120595816A (en) * | 2025-08-07 | 2025-09-05 | 江苏正贸仓储设备制造有限公司 | Industrial robot walking control system based on obstacle recognition |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10948919B2 (en) * | 2017-09-11 | 2021-03-16 | Baidu Usa Llc | Dynamic programming and gradient descent based decision and planning for autonomous driving vehicles |
| CN120742953A (en) * | 2025-08-28 | 2025-10-03 | 广州基医云计算有限公司 | AI-assisted unmanned aerial vehicle low-altitude flight obstacle avoidance method and system |
-
2025
- 2025-10-14 CN CN202511465933.2A patent/CN120953386B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN120560279A (en) * | 2025-07-31 | 2025-08-29 | 宁德师范学院 | Dynamic obstacle avoidance system in unmanned boat path planning |
| CN120595816A (en) * | 2025-08-07 | 2025-09-05 | 江苏正贸仓储设备制造有限公司 | Industrial robot walking control system based on obstacle recognition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN120953386A (en) | 2025-11-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114384920B (en) | Dynamic obstacle avoidance method based on real-time construction of local grid map | |
| CN107861508B (en) | A method and device for local motion planning of a mobile robot | |
| Chung et al. | The detection and following of human legs through inductive approaches for a mobile robot with a single laser range finder | |
| CN118760204A (en) | Path planning method for intelligent inspection of drones at construction sites | |
| CN117519215A (en) | Multi-AGV driving control method, device, equipment and storage medium | |
| CN119645060A (en) | A collaborative inspection control method for drone groups used for power station inspection | |
| CN120445193A (en) | All-terrain autonomous navigation and task planning method and system for photovoltaic inspection robots | |
| CN105204511B (en) | A kind of decision-making technique of object autonomous | |
| CN120595851A (en) | UAV flight attitude adjustment method and system based on reinforcement learning | |
| CN120215514A (en) | Reinforced learning unmanned forklift obstacle avoidance scheduling method and system for dynamic obstacle | |
| CN120183188A (en) | Intelligent tracking and positioning system for hazardous chemicals transport vehicles at container terminals | |
| CN120116222A (en) | A method for dynamic environment adaptation and behavior optimization of humanoid robots based on artificial intelligence | |
| CN119919821A (en) | Safety analysis method and system for complex driving scenarios of warehouse robots | |
| CN118419788A (en) | Intelligent control system for approaching protection and anti-collision protection of monorail crane personnel | |
| Zheng et al. | Research on the safety of AGV path planning based on D* Algorithm | |
| CN120953386B (en) | Intelligent control method and system for obstacle avoidance of aerial working platform | |
| Mitroudas et al. | Multi-criteria decision making for autonomous uav landing | |
| CN120428724A (en) | A visual acquisition and avoidance control method for AGV robots based on deep thinking | |
| CN119440080A (en) | Drone low altitude obstacle avoidance system | |
| Zhang et al. | Path Planning for Mobile Robot Based on RGB-D SLAM and Pedestrian Trajectory Prediction | |
| Liu et al. | A robot obstacle avoidance approach with lidar and rgb camera data combined | |
| Tian et al. | Design of Intelligent Control System for Autonomous Transport Vehicle Based on Deep Learning | |
| CN120972982B (en) | AGV automatic driving control method and controller based on positioning sensing | |
| CN120722934B (en) | Unmanned aerial vehicle safety route planning method in urban environment based on deep learning | |
| Silva et al. | Improving Energy Performance of Camera Lidar Fusion by Intermittent Human Detection for Social Navigation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |