Tower crane real-time obstacle avoidance path planning method, storage medium and tower crane controller

Abstract

The embodiment of the application provides a tower crane real-time obstacle avoidance path planning method, a storage medium and a tower crane, and belongs to the technical field of mechanical control. The method comprises the following steps: acquiring position data of obstacles and articles carried in a current working environment of a tower crane, and an article swing angle and an angular velocity relative to the tower crane; acquiring a next target position expected to be reached by the articles according to current path planning, and determining an operation to be performed on the tower crane according to the next target position; inputting the position data, the swing angle, the angular velocity and the operation into a tower crane obstacle avoidance model, and determining whether the operation meets a plurality of constraint condition parameters according to a model output result; if the determination result is not met, modifying the current operation and correcting the next target position expected to be reached by the articles until the determination result is met, so that a current planning path of the articles is corrected until a final target point of the planning path is reached. The method can realize real-time automatic obstacle avoidance and realize a truly "unmanned tower crane".

Description

Technical Field

[0001] This invention relates to the field of mechanical control technology, specifically to a path planning method, storage medium, and tower crane controller for real-time obstacle avoidance of tower cranes. Background Technology

[0002] Traditional tower crane operations require at least two workers: one observing from the ground and directing the other via walkie-talkie from the control room. Current technologies for unmanned tower cranes utilize multi-dimensional tower crane visualization monitoring systems to acquire real-time information and transmit data via radio frequency waves, enabling remote control operation from the ground and replacing the inherent risks of high-altitude work. However, the core of the entire operation still revolves around human intervention, and true unmanned operation is not yet fully realized. Summary of the Invention

[0003] The purpose of this invention is to provide a path planning method for real-time obstacle avoidance of tower cranes. This method can accurately and automatically control the operation of tower cranes by setting multiple reward and penalty functions and corresponding multiple constraint parameters, and the operator only needs to act as a supervisor.

[0004] To achieve the above objectives, embodiments of the present invention provide a path planning method for real-time obstacle avoidance of tower cranes, the method comprising:

[0005] Acquire the position data of obstacles and transported items in the current working environment of the tower crane, as well as the swing angle and angular velocity of the items relative to the tower crane;

[0006] Based on the current path planning, obtain the next target location that the item is expected to reach, and determine the operation to be performed on the tower crane accordingly;

[0007] The position data, swing angle, angular velocity, and operation are input into the tower crane obstacle avoidance model. Based on the output of the tower crane obstacle avoidance model, it is determined whether the operation meets multiple constraint parameters.

[0008] If the determination result is not compliant, the current operation to be performed on the tower crane is modified, and the next target position expected to be reached by the item is revised until the determination result is compliant. This process corrects the planned path of the item from its current position to the next target position until the final target point of the planned path is reached.

[0009] The tower crane obstacle avoidance model is a neural network model, which is obtained by training through a predefined reward and punishment function and a number of corresponding constraint parameters.

[0010] The predefined reward and penalty function includes at least two of a plurality of reward and penalty functions: accuracy reward and penalty function, obstacle avoidance reward and penalty function, and oscillation reward and penalty function.

[0011] Preferably, the predefined reward and penalty function is the weighted sum of the plurality of reward and penalty functions; the plurality of reward and penalty functions are set according to the plurality of constraint parameters; and the different weights of the plurality of reward and penalty functions are set according to the safety importance of the construction site.

[0012] Preferably, the accuracy reward / penalty function is used to determine the reward / penalty value of the operation based on the accuracy of the item reaching the next target location; the accuracy reward / penalty function is the sum of the accuracy status reward / penalty value and the accuracy behavior reward / penalty value of the operation at the current moment; the accuracy status reward / penalty value is calculated by the distance between the item's position and the next target location, a first threshold, and a second threshold, where the first threshold and the second threshold are the corresponding constraint parameters of the accuracy reward / penalty function.

[0013] Furthermore, the accuracy status reward / penalty value when the distance between the item and the next target location is less than the first threshold is greater than the accuracy status reward / penalty value when the distance between the item and the next target location is less than the second threshold.

[0014] The accuracy reward / penalty value of the item when it is close to the target point is greater than the accuracy reward / penalty value of the item when it is far away from the target point.

[0015] Preferably, the obstacle avoidance reward and penalty function is used to determine the reward and penalty value of the operation based on the degree to which the item moves away from the obstacle during the tower crane's transport of the item; the obstacle avoidance reward and penalty function is the sum of the obstacle avoidance state reward and penalty value and the obstacle avoidance behavior reward and penalty value of the operation at the current moment; the obstacle avoidance state reward and penalty value is calculated by the distance between the item's position and the obstacle, a third threshold, and a fourth threshold, and the third threshold and the fourth threshold are the corresponding constraint condition parameters of the obstacle avoidance reward and penalty function.

[0016] Furthermore, the obstacle avoidance reward / penalty value when the distance between the item and the obstacle is less than the third threshold is greater than the obstacle avoidance reward / penalty value when the distance between the item and the obstacle is less than the fourth threshold.

[0017] The obstacle avoidance behavior reward / penalty value of the item as it moves away from the obstacle is greater than the obstacle avoidance behavior reward / penalty value of the item as it moves closer to the obstacle.

[0018] Preferably, the swing reward / penalty function is used to determine the swing reward / penalty value of the operation based on the degree of swing of the item relative to the tower crane during the transport of the item by the tower crane; the swing reward / penalty value is calculated by the angular velocity of the item relative to the tower crane, the swing angle, the fifth threshold, and the sixth threshold, and the fifth threshold and the sixth threshold are the corresponding constraint condition parameters of the swing reward / penalty function.

[0019] Furthermore, the swing reward / penalty value when the swing angle of the item is within the required swing range and the angular velocity of the item relative to the tower crane is within the required range is greater than the swing reward / penalty value when the swing angle of the item is within the required swing range but the angular velocity of the item relative to the tower crane is not within the required range.

[0020] On the other hand, the present invention provides a machine-readable storage medium storing instructions for causing a machine to execute: the tower crane obstacle avoidance model training method of the present application, or the tower crane real-time obstacle avoidance path planning method.

[0021] On the other hand, the present invention provides a tower crane controller, which includes: at least one processor, and at least one memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute: the tower crane obstacle avoidance model training method of the present application, or the tower crane real-time obstacle avoidance path planning method.

[0022] Through the above technical solution, when controlling the tower crane, the operation from the current position to the next target position, which is also the current path planning step, involves inputting the position data of obstacles and transported items in the current working environment of the tower crane, the swing angle of the items, the angular velocity of the items relative to the tower crane, and the operation to be performed on the tower crane into the tower crane obstacle avoidance training model. Based on the accuracy penalty value and / or obstacle avoidance penalty value and / or swing penalty value output by the tower crane obstacle avoidance training model, it is determined whether the total penalty value meets the safety requirements of the on-site operation, thereby determining whether the current path planning needs to be corrected until the final target point of the planned path is reached. In this process, the operation of the tower crane transporting items meets the accuracy requirements, the swing angle of the items during transport meets the operational safety obstacle avoidance requirements, and the angular velocity of the items relative to the tower crane meets the operational swing amplitude requirements. This enables the tower crane to automatically perform obstacle avoidance operations, eliminating the need for manual operation.

[0023] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0024] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0025] Figure 1 This is a flowchart of an embodiment of the path planning method for real-time obstacle avoidance of tower cranes according to this application; and

[0026] Figure 2This is a flowchart of an embodiment of the tower crane obstacle avoidance model training method of this application. Detailed Implementation

[0027] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0028] This invention provides a path planning method for real-time obstacle avoidance of tower cranes, which solves the algorithm for precise automatic control of tower cranes to avoid obstacles in real time when transporting goods. It should be noted that this application also provides an embodiment illustrating a method for obtaining a tower crane obstacle avoidance model, but this application is not limited thereto, and those skilled in the art can use other available tower crane obstacle avoidance models. In this embodiment, the tower crane obstacle avoidance model obtained from the embodiment of this application can be directly used. The current state of the transported goods is input into the tower crane obstacle avoidance model, and the planned path from the current position to the next target point can be determined / corrected based on the output of the tower crane obstacle avoidance model. The following is combined with... Figure 1 Explain its implementation process.

[0029] Step S1: Obtain the position data of obstacles and transported items in the current working environment of the tower crane, as well as the swing angle and angular velocity of the items relative to the tower crane;

[0030] In this embodiment, the location data of the tower crane, obstacles, and transported items can be extracted from a real-time image of the entire tower crane's working environment. Specifically, the global camera mounted on the tower crane's working environment can be a depth camera, capable of encompassing the entire workspace and the tower crane within its field of view, ensuring that obstacles and transported items are within this field of view. Simultaneously, the camera mounted on the tower crane can be either a depth camera or a regular camera, used to capture the swing angle and angular velocity of the transported items relative to the tower crane. The state of the transported items is determined by comparing the images obtained from the global camera and the camera mounted on the tower crane.

[0031] In this embodiment, multiple real-time images of the hook's current status are acquired by a camera mounted on the tower crane. By comparing the state changes of the transported items in the real-time images acquired at different time points, the swing angle of the transported items and the angular velocity of the items relative to the tower crane are extracted.

[0032] Step S2: Based on the current path planning, obtain the next target location that the item is expected to reach, and determine the operation to be performed on the tower crane accordingly;

[0033] It should be noted that each time a new operation is performed, the data of the tower crane's current working environment needs to be retrieved to determine if the environment has changed, and thus decide whether the route needs to be replanned. This route planning can be either a global route or a local route.

[0034] Step S3: Input the position data, the swing angle, the angular velocity, and the operation into the tower crane obstacle avoidance model, and obtain the output results of the tower crane obstacle avoidance model;

[0035] In this embodiment, after inputting the status data of the transported goods and the operation to be performed on the tower crane into the tower crane obstacle avoidance model, the output result is an evaluation index of the transported goods reaching the target position, dynamically avoiding obstacles, and swing amplitude after performing the specified operation in the current state, so as to determine whether the operation to be performed meets the operation safety requirements.

[0036] Step S4: Determine whether the operation meets multiple constraint parameters based on the output of step S3. If the determination result is negative, modify the current operation to be performed on the tower crane to correct the next target position expected to be reached by the item, and return to step S2; if the determination result is positive, proceed to step S5.

[0037] In this step, if the crane obstacle avoidance model outputs a result indicating that the current operation on the crane meets operational safety requirements, then the operation determined by the current path planning step can be directly executed, meaning the current path planning step does not need to be modified. Conversely, if the crane obstacle avoidance model outputs a result indicating that the current operation on the crane does not meet operational safety requirements, then the current operation on the crane is modified, and the next target position expected to be reached by the item is corrected based on the modified operation, until the modified operation meets operational safety requirements. Simultaneously, the next target position expected to be reached by the item based on the compliant operation is obtained, thus enabling the correction of the planned path from the current node to the next target position.

[0038] Step S5: Determine / correct the planned path of the item from its current location to the next target location.

[0039] In this step, based on the determination result of step 4, the planned path of the item from the current location to the next target location is determined / corrected. Then, the tower crane executes the operation determined by the current path planning step, or executes the operation determined by the corrected current path planning step.

[0040] It should be noted that, in this embodiment, the tower crane obstacle avoidance model is a neural network model, and it is obtained by training through a predefined reward and penalty function and a number of corresponding constraint parameters; the predefined reward and penalty function includes at least two of a number of reward and penalty functions: accuracy reward and penalty function, obstacle avoidance reward and penalty function, and swing reward and penalty function.

[0041] Compared with the prior art, the technical advantages of this application are as follows:

[0042] (1) By using multiple reward and punishment functions, the operation of the tower crane transporting the goods is judged from the accuracy of operation, obstacle avoidance effect and swing amplitude. Therefore, the tower crane obstacle avoidance can automatically control the tower crane operation and achieve a truly unmanned tower crane.

[0043] (2) The real-time obstacle avoidance path planning method can simultaneously satisfy multiple constraint parameters (corresponding to multiple reward and punishment functions), enabling multiple constraint parameters to achieve the expected effect at the same time.

[0044] In the above embodiment, the accuracy reward / penalty function R1(t) is used to evaluate whether the transported item has reached the target point after performing the specified operation in the current state.

[0045] Preferably, R1(t) can be set as follows:

[0046]

[0047]

[0048] R1(t)=γ 11 (t)+γ 12 (t)

[0049] in:

[0050] γ 11 (t) represents the accuracy state reward / penalty value of the current action at the current moment;

[0051] γ 12 (t) represents the reward / penalty value for the accuracy of the current action at the current moment;

[0052] R1(t) is the total accuracy bonus / penalty value for the current action at the current moment; d(t) is the distance between the location of the transported item and the target point at this moment.

[0053] d(t-1): The distance between the location of the item being transported in the previous moment and the target point;

[0054] d0: If the distance to the target point is less than d0, then the item is considered to have reached the target point;

[0055] d1: If the distance to the target point is less than d1, the item is considered to have reached the target point range (the distance between the item's location and the target point is within d1, which does not affect the use of the project, but the accuracy is lower than the d0 range).

[0056] It should be noted that the transportation effect is best when the distance from the target point is less than d0, and the smaller d0-d(t) is, the better the effect.

[0057] When the distance to the target point is less than d1, the delivery is successful but the accuracy is not high enough, and the smaller d1-d(t) is, the higher the accuracy.

[0058] When d(t) < d(t-1), it means that the item is gradually approaching the target point, and a reward of 1 is given;

[0059] When d(t) = d(t-1) and d(t) ≠ 0, it means that the item stops before reaching the target point, and a reward of -1 is given (actually a punishment, indicating that such behavior is not advisable).

[0060] When d(t) > d(t-1), it means that the item is gradually moving away from the target point. This behavior is incorrect, so a reward of -10 is given to train the network to avoid this behavior.

[0061] In other words, the accuracy state reward / penalty value when the distance between the item and the next target location is less than the first threshold d0 is greater than the accuracy state reward / penalty value when the distance between the item and the next target location is less than the second threshold d1; the accuracy behavior reward / penalty value when the item is close to the target point is greater than the accuracy behavior reward / penalty value when the item is far away from the target point.

[0062] In the above embodiments, the obstacle avoidance reward / penalty function R2(t) is used to evaluate whether the transported item dynamically avoids obstacles after performing a specified operation in the current state. Preferably, it can be set as follows:

[0063]

[0064]

[0065] R2(t)=γ 21 (t)+γ 22 (t)

[0066] in:

[0067] γ 21 (t) represents the obstacle avoidance reward / penalty value for the current action at the current moment;

[0068] γ 22 (t) represents the obstacle avoidance behavior reward / penalty value for the current action at the current moment;

[0069] l(t): The distance between the location of the transported item and the obstacle at this moment;

[0070] l(t-1): The distance between the location of the transported item and the obstacle in the previous moment;

[0071] l0: If the distance to the obstacle is less than l0, the object is considered to have hit the obstacle;

[0072] l1: If the distance to the obstacle is less than l1 but greater than l0, it is considered that although the object has not touched the obstacle, there is still a risk of it touching the obstacle within that range.

[0073] It should be noted that when the distance to the obstacle is less than l0, touching the obstacle should be avoided, and the smaller l0-l(t) is, the more serious the contact will be.

[0074] When the distance to the obstacle is less than l1 but greater than l0, the item does not touch the obstacle, but there is a risk of touching the obstacle, which should be avoided. The smaller l1-l(t) is, the greater the risk of touching the obstacle.

[0075] When l(t)≥l(t-1), the item is gradually moving away from the obstacle, and a reward value of 1 is given;

[0076] When l(t) < l(t-1), the item is gradually approaching the obstacle. This behavior is incorrect, so a reward of -1 is given to train the network to avoid this behavior.

[0077] In other words, the obstacle avoidance reward / penalty value when the distance between the item and the obstacle is less than l0 is greater than the obstacle avoidance reward / penalty value when the distance between the item and the obstacle is less than l1. That is, the obstacle avoidance reward / penalty value when the item touches the obstacle is less than the obstacle avoidance reward / penalty value when the item is at risk of touching the obstacle. The obstacle avoidance behavior reward / penalty value when the item gradually moves away from the obstacle is greater than the obstacle avoidance behavior reward / penalty value when the item gradually moves closer to the obstacle. In the above embodiment, the swing reward / penalty function R3(t) is used to evaluate whether the swing amplitude of the transported item meets the requirements after performing a specified operation in the current state. Preferably, it can be set as follows:

[0078]

[0079] in:

[0080] ω(t): The angular velocity of the object relative to the crane at this moment;

[0081] ω0: If the angular velocity ω(t) of the object relative to the tower crane is less than the sixth threshold ω0 and is within the swing range, then it can be considered that the object is not swinging at this time;

[0082] θ(t): The angle of the object's swing at this moment;

[0083] θ0: If the swing angle θ(t) of the object is less than the fifth threshold θ0, then it is within the required swing range and meets the requirements.

[0084] It should be noted that when θ(t) < θ0 and ω(t) ≤ ω0, it means that the item hardly swings, which is close to the ideal effect, and the reward value is set to 100.

[0085] When θ(t) < θ0 and ω(t) > ω0, it means that although the object swings, it is within the required range. The effect is directly proportional to θ0 - θ(t) and inversely proportional to ω(t) - ω0.

[0086] At other times, if the item is not within the required range, set a reward value of -10 to train the network to avoid this behavior.

[0087] In other words, the swing penalty value when the swing angle of the item is within the required swing range and the angular velocity of the item relative to the tower crane is within the required range is greater than the swing penalty value when the swing angle of the item is within the required swing range but the angular velocity of the item relative to the tower crane is not within the required range.

[0088] This invention also provides an embodiment of a tower crane obstacle avoidance model training method. This training method uses tower crane operation data as a learning sample set to train a deep neural network model. During training, it comprehensively considers whether the relative position of the goods being transported by the tower crane to the destination and obstacles meets the operational accuracy requirements, whether the swing angle during goods transport meets the operational safety obstacle avoidance requirements, and whether the angular velocity of the goods relative to the tower crane meets the operational swing amplitude requirements. Accuracy reward and penalty functions, obstacle avoidance reward and penalty functions, and swing reward and penalty functions are set to obtain a tower crane obstacle avoidance model that meets the above multiple constraint parameters. This enables precise and automatic control of the tower crane to perform obstacle avoidance operations through the tower crane obstacle avoidance training model. The specific implementation process is described below in conjunction with... Figure 2 Please provide an explanation.

[0089] Step 1: Create an initial deep neural network model and initialize the experience pool and noise distribution;

[0090] In this embodiment, the deep neural network model employs an off-policy learning method, separating the action policy and the evaluation policy. The action policy (Actor) is a stochastic policy, ensuring sufficient exploration, while the evaluation policy (Critic) is a deterministic policy, reducing the amount of data that needs to be sampled and improving algorithm efficiency. Simultaneously, a dual-network approach using the current and target networks, along with an experience pool, is employed to calculate the target value. Furthermore, the experience pool is used to calculate the target Q-value, effectively breaking down the correlation between data points and preventing instability in the reinforcement learning algorithm caused by function approximation using deep neural networks.

[0091] It should be noted that, in addition to the action strategy and evaluation strategy, the current network and the target network, four more networks need to be created: the Actor current network, the Critic current network, the Actor target network, and the Critic target network.

[0092] In this embodiment, the data for initializing the experience pool comes from actual operation data. Specifically, by modeling the tower crane and its surrounding environment in a three-dimensional coordinate system, (depth) cameras are installed on the tower crane and in the tower crane's working environment, and the coordinates of objects entering the work space, such as unforeseen obstacles, are transmitted to the processor in real time to obtain actual operation data.

[0093] In this embodiment, the method of acquiring a global real-time image of the tower crane's working environment through a global depth camera, as described in the previous embodiment, can also be used, and the location data of obstacles and transported items can be extracted from the real-time image. This will not be elaborated further here.

[0094] Step 2: Obtain the tower crane's operational data as a learning sample set. The learning sample data in the learning sample set includes: the state S of the items being transported by the tower crane. t In the state S t Action A after adding noise t Execute the aforementioned action A t The resulting reward / penalty value R t and the next state S t+1 ;

[0095] Specifically, based on the actual operation of the tower crane, in state S t The following action A after adding noise will be executed. t The reward R obtained after performing this action t Next step, state S t+1 These data are stored in an experience pool as learning samples. The action refers to the operation performed on the tower crane. The state includes the relative position of the object with respect to the destination and obstacles, the swing angle of the object, and the angular velocity of the object relative to the tower crane. The next state is to execute action A. t The state.

[0096] It should be noted that the learning samples can be artificially defined operation samples, specifically: based on the known current state S t An action A is defined artificially. t To reach the next state S t+1 (The actual measured state), and a reward or penalty value R is given based on the effect of this action. t Repeat the above steps until you have obtained a sufficient set of sample data.

[0097] Step 3: Randomly sample a preset number of learning samples from the learning sample set obtained in Step 2, and train them according to a predefined reward and penalty function to obtain a tower crane obstacle avoidance model that meets multiple constraint parameters;

[0098] The plurality of reward and penalty functions include at least two of the following: an accuracy reward and penalty function, an obstacle avoidance reward and penalty function, and an oscillation reward and penalty function, and the reward and penalty functions are set in accordance with the plurality of constraint parameters.

[0099] Specifically, when the number of samples reaches N, the number of samples required for training the neural network, n samples are randomly sampled. The Critic network is updated based on the calculated loss function, the Actor network is updated based on the policy gradient, and both target networks are updated using a soft update method. After a pre-set number of training episodes, the model is saved. Accuracy reward and penalty functions, obstacle avoidance reward and penalty functions, and swing reward and penalty functions are set for operational accuracy requirements, obstacle avoidance safety requirements, and swing amplitude requirements, respectively. This allows for a comprehensive and automatic evaluation of whether the operation of the items transported by the tower crane meets the requirements, and controls the tower crane operation accordingly without manual intervention.

[0100] This invention provides a storage medium storing a program that, when executed by a processor, implements the tower crane obstacle avoidance model training method or the tower crane real-time obstacle avoidance path planning method.

[0101] This invention provides a processor for running a program, wherein the program executes the tower crane obstacle avoidance model training method or the tower crane real-time obstacle avoidance path planning method.

[0102] This invention provides a tower crane controller, which includes: at least one processor, at least one memory connected to the processor, and a bus; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the tower crane obstacle avoidance model training method or the tower crane real-time obstacle avoidance path planning method.

[0103] The processor contains a kernel, which retrieves the corresponding program unit from memory. One or more kernels can be configured; by adjusting kernel parameters, the operation of the tower crane can be automatically controlled, safely and smoothly transporting goods to their destination.

[0104] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0105] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. The processor executes the tower crane obstacle avoidance model training method or the tower crane real-time obstacle avoidance path planning method. The device described herein can be a server, PC, PAD, mobile phone, etc.

[0106] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing the tower crane obstacle avoidance model training method or the tower crane real-time obstacle avoidance path planning method.

[0107] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0108] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0109] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0110] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0111] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0112] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0113] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0114] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0115] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for real-time obstacle avoidance path planning of a tower crane, comprising: obtaining position data of an obstacle and a carried object in a current working environment of the tower crane, and a swing angle and an angular velocity of the object relative to the tower crane; obtaining a next target position expected to be reached by the object according to a current path planning, and determining an operation to be performed on the tower crane based on the next target position; inputting the position data, the swing angle, the angular velocity and the operation into a tower crane obstacle avoidance model, and determining whether the operation meets a plurality of constraint condition parameters based on a result output by the tower crane obstacle avoidance model; and if the determination result is not met, modifying the operation to be performed on the tower crane, and revising the next target position expected to be reached by the object, until the determination result is met, thereby revising a planned path of the object from a current position to the next target position until a final target point of the planned path is reached. The tower crane obstacle avoidance model is a neural network model, and is obtained by training a predefined reward and punishment function and a plurality of corresponding constraint condition parameters. The predefined reward and punishment function includes at least two of a plurality of reward and punishment functions: an accuracy reward and punishment function, an obstacle avoidance reward and punishment function, and a swing reward and punishment function. 2.The method of claim 1, wherein: the predefined reward and punishment function is a weight sum of the plurality of reward and punishment functions; the plurality of reward and punishment functions are set according to the plurality of constraint condition parameters; and different weights of the plurality of reward and punishment functions are set according to safety importance of a construction site. 3.The method of claim 1, wherein: the accuracy reward and punishment function is used to determine a reward and punishment value of the operation according to accuracy of the object reaching the next target position; the accuracy reward and punishment function is a sum of an accuracy state reward and punishment value and an accuracy behavior reward and punishment value of the operation at a current time; the accuracy state reward and punishment value is calculated by a distance between the object and the next target position, a first threshold value, and a second threshold value, the first threshold value and the second threshold value being constraint condition parameters corresponding to the accuracy reward and punishment function. 4.The method of claim 3, wherein: the accuracy state reward and punishment value when the distance between the object and the next target position is less than the first threshold value is greater than the accuracy state reward and punishment value when the distance between the object and the next target position is less than the second threshold value; the accuracy behavior reward and punishment value when the object is close to the target point is greater than the accuracy behavior reward and punishment value when the object is far from the target point. 5.The method of claim 1, wherein: the obstacle avoidance reward and punishment function is used to determine a reward and punishment value of the operation according to a degree of the object moving away from the obstacle during the tower crane carrying the object; the obstacle avoidance reward and punishment function is a sum of an obstacle avoidance state reward and punishment value and an obstacle avoidance behavior reward and punishment value of the operation at a current time. ​ The obstacle avoidance state reward and punishment value is calculated by the distance between the article and the obstacle, a third threshold value, and a fourth threshold value, and the third threshold value and the fourth threshold value are constraint condition parameters corresponding to the obstacle avoidance reward and punishment function.

6. The tower crane real-time obstacle avoidance path planning method according to claim 5, characterized in that, the obstacle avoidance state reward and punishment value when the distance between the article and the obstacle is less than the third threshold value, and the obstacle avoidance state reward and punishment value when the distance between the article and the obstacle is greater than the fourth threshold value; the obstacle avoidance degree behavior reward and punishment value when the article is gradually moving away from the obstacle, and the obstacle avoidance degree behavior reward and punishment value when the article is gradually moving closer to the obstacle.

7. The tower crane real-time obstacle avoidance path planning method according to claim 1, characterized in that, the swing reward and punishment function is used to determine the swing reward and punishment value of the operation according to the swing degree of the article relative to the tower crane during the article transportation of the tower crane; the swing reward and punishment value is calculated by the angular velocity of the article relative to the tower crane, a swing angle, a fifth threshold value, and a sixth threshold value, and the fifth threshold value and the sixth threshold value are constraint condition parameters corresponding to the swing reward and punishment function.

8. The tower crane real-time obstacle avoidance path planning method according to claim 1, characterized in that, the swing reward and punishment value when the swing angle of the article is within a required swing range and the angular velocity of the article relative to the tower crane is within a required range, and the swing reward and punishment value when the swing angle of the article is within the required swing range but the angular velocity of the article relative to the tower crane is not within the required range.

9. A storage medium, characterized by The storage medium includes a stored program, wherein when the program runs, the device where the storage medium is located is controlled to perform the tower crane real-time obstacle avoidance path planning method according to any one of claims 1-8.

10. A tower crane controller, characterized in that including: at least one processor and at least one memory connected to the processor; wherein the processor, the memory complete mutual communication through the bus; the processor is used to call the program instruction in the memory, to execute: the tower crane real-time obstacle avoidance path planning method according to any one of claims 1-8.

Images