Motion track prediction method, device and equipment based on IP-R algorithm

Abstract

The invention discloses a motion trail prediction method based on an IP-R algorithm, and the method comprises the steps: introducing an improved particle swarm optimization algorithm for optimizing an optimal neural network parameter, initializing a particle swarm, selecting an average fitness change rate, and carrying out the prediction of time sequence data based on the prediction capability of an RBF neural network for time sequence data; constructing a particle swarm optimization function for adaptively adjusting an inertia weight factor; based on a particle swarm optimization function, optimizing parameters of the radial basis function neural network to obtain an IP-R model; collecting motion data of the moving object in the coking environment, and training the IP-R model to improve the prediction precision of the IP-R model on the motion trail of the moving object in the coking environment; and adopting the trained IP-R model to predict the motion track of the moving object in the coking environment to obtain a prediction result of the target object so as to improve the track prediction precision of the moving object in the coking environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of motion trajectory prediction, and in particular relates to a motion trajectory prediction method, device and equipment based on an IP-R algorithm. Background Art

[0002] In industrial environments, the degree of unmanned and intelligent operation is still low, and in many cases, it still relies heavily on manual labor, which can easily lead to safety accidents. Among them, artificial intelligence technology is less used in carbon deposition environments. Carbon deposition environments refer to industrial environments with high carbon dust content on site, which usually have environmental characteristics such as high dust concentration and unstructured environment.

[0003] During the movement of the coke pusher, the temperature and dust concentration at the site are high, and harmful gases are generated. Without the assistance of intelligent technology or equipment, the driver has to rely entirely on the naked eye to observe whether there are people in the movement area. This places high demands on the driver. This feature also determines that the personnel cannot maintain stable work for a long time, and the safety of the work site cannot be well guaranteed.

[0004] In this macro-environmental context, it is particularly important to provide an intelligent prediction method for trajectories in a carbon deposition environment, which can be used for real-time monitoring of complex industrial environments and timely prediction of the motion trajectories of mobile targets to provide protection for industrial safety production. Summary of the invention

[0005] The purpose of the present invention is to provide a motion trajectory prediction method based on the IP-R algorithm to improve the prediction accuracy of the motion trajectory of a moving object in a coking environment.

[0006] To solve the above problems, the technical solution of the present invention is: A motion trajectory prediction method based on IP-R algorithm, comprising: Initialize the radial basis neural network and determine the structure of the radial basis neural network; Initialize the particle population, select the average fitness change rate, and construct a particle swarm optimization function that adaptively adjusts the inertia weight factor; Based on the particle swarm optimization function, the parameters of the radial basis neural network are optimized to obtain the IP-R model; Collect motion data of moving objects in the coking environment and train the IP-R model to improve the prediction accuracy of the IP-R model for the motion trajectory of moving objects in the coking environment; The trained IP-R model is used to predict the motion trajectory of moving objects in the coking environment to obtain the prediction results of the target objects.

[0007] According to an embodiment of the present invention, the selecting of the average fitness conversion rate and constructing the particle swarm optimization function for adaptively adjusting the inertia weight factor further includes: Based on the fitness function of the PSO algorithm, calculate the average fitness change rate , the formula is: in, is the average fitness of the particle at time t, is the difference in the average fitness of the particle at two consecutive moments; Based on the average fitness change rate, a particle swarm optimization function for adaptively adjusting the inertia weight factor is constructed to optimize the parameter combination in the radial basis neural network. The formula is: Where t is the current iteration number, is the maximum number of iterations, is the maximum inertia weight factor, is the minimum inertia weight factor.

[0008] According to an embodiment of the present invention, the initializing particle population further comprises: Initialize the position and velocity of the particle population, each particle represents a set of radial basis neural network parameters; Calculate the fitness of each particle according to the individual optimal position of each particle and the global optimal position; and calculate the average fitness based on the fitness of each particle.

[0009] According to an embodiment of the present invention, optimizing the parameters of the radial basis neural network based on the particle swarm optimization function further includes: Based on the particle swarm optimization function, an update formula is constructed to update the speed and position of each particle until the preset requirements are met; the update formula is: in, For particles of The speed of the dimension; and is the learning factor; It is a particle of Dimension individual optimal position; It is a particle of Dimensional global optimal position; and are two independent random numbers, ranging from 0 to 1; It is a particle of The location of the dimension; is the inertia weight factor.

[0010] According to an embodiment of the present invention, the IP-R model includes an input layer, a hidden layer and an output layer; The input layer includes n neurons representing n input variables, and there are Q training samples, x q ,n represents the nth variable of the qth sample, then for the qth sample, the neural network input is: x q =[x q ,1 ,x q ,2 ,...,x q ,N ] n=1 ,2 ,...,N The hidden layer uses a Gaussian function as the activation function of the hidden layer, and the formula is: in, is an n-dimensional input sample vector; It is The center vector of the hidden layer node determines the position of the basis function; It is The expansion coefficient of the basis function of the hidden layer nodes; The output layer performs linear weighted summation on the output of the hidden layer, and the formula is: In the formula, is the weight from the hidden layer to the output layer.

[0011] According to one embodiment of the present invention, motion data of a moving object in a coking environment is collected, the coordinates of the moving object are used as feature vectors, and the collected motion data are divided into a training set and a test set. The IP-R model is trained based on the training set, and the prediction accuracy of the IP-R model is evaluated based on the test set.

[0012] A motion trajectory prediction device based on IP-R algorithm, comprising: RBF module, used to initialize the radial basis neural network and determine the structure of the radial basis neural network; IPSO module is used to initialize the particle population, select the average fitness change rate, and construct a particle swarm optimization function that adaptively adjusts the inertia weight factor; The IP-R module is used to optimize the parameters of the radial basis neural network based on the particle swarm optimization function to obtain the IP-R model; A training module is used to collect motion data of moving objects in a coking environment and train the IP-R model to improve the prediction accuracy of the IP-R model for the motion trajectory of moving objects in the coking environment; The prediction module is used to predict the motion trajectory of moving objects in the coking environment using the trained IP-R model to obtain the prediction result of the target object.

[0013] According to an embodiment of the present invention, the IPSO module is configured as follows: Based on the fitness function of the PSO algorithm, calculate the average fitness change rate , the formula is: in, is the average fitness of the particle at time t, is the difference in the average fitness of the particle at two consecutive moments; Based on the average fitness change rate, a particle swarm optimization function for adaptively adjusting the inertia weight factor is constructed to optimize the parameter combination in the radial basis neural network. The formula is: Where t is the current iteration number, is the maximum number of iterations, is the maximum inertia weight factor, is the minimum inertia weight factor.

[0014] A motion trajectory prediction device based on IP-R algorithm, comprising: A memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via a line; The at least one processor calls the instruction in the memory so that the motion trajectory prediction device based on the IP-R algorithm executes the motion trajectory prediction method based on the IP-R algorithm as described in any one of claims 1 to 6.

[0015] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the motion trajectory prediction method based on the IP-R algorithm as described in any one of claims 1 to 6 is implemented.

[0016] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art: The motion trajectory prediction method based on the IP-R algorithm in one embodiment of the present invention, based on the prediction ability of the RBF neural network for time series data, introduces an improved particle swarm optimization algorithm for optimizing the optimal neural network parameters, initializes the particle population, selects the average fitness change rate, and constructs a particle swarm optimization function for adaptively adjusting the inertia weight factor; based on the particle swarm optimization function, the parameters of the radial basis neural network are optimized to obtain the IP-R model; the motion data of the moving object in the coking environment are collected, and the IP-R model is trained to improve the prediction accuracy of the IP-R model for the motion trajectory of the moving object in the coking environment; the trained IP-R model is used to predict the motion trajectory of the moving object in the coking environment, and the prediction result of the target object is obtained to improve the trajectory prediction accuracy of the moving object in the coking environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of a motion trajectory prediction method based on IP-R algorithm in one embodiment of the present invention; Figure 2 RBF neural network structure diagram in one embodiment of the present invention; Figure 3 It is a PSO particle swarm algorithm optimization flow chart in one embodiment of the present invention; Figure 4 A schematic diagram of trajectory prediction of different models in one embodiment of the present invention; Figure 5 It is a block diagram of a motion trajectory prediction device based on IP-R algorithm in one embodiment of the present invention; Figure 6 Schematic diagram of a motion trajectory prediction device based on the IP-R algorithm in one embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following is a further detailed description of a motion trajectory prediction method based on the IP-R algorithm proposed by the present invention in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description and claims.

[0019] Embodiment 1 This embodiment provides a motion trajectory prediction method based on the IP-R algorithm, which is used for trajectory prediction of moving objects (such as pedestrians and bicycles) in a coking environment to provide protection for industrial safety production. Figure 1 , the motion trajectory prediction method based on the IP-R algorithm includes: S1: Initialize the radial basis neural network and determine the structure of the radial basis neural network; S2: Initialize the particle population, select the average fitness change rate, and construct a particle swarm optimization function that adaptively adjusts the inertia weight factor; S3: Based on the particle swarm optimization function, the parameters of the radial basis neural network are optimized to obtain the IP-R model; S4: Collect motion data of moving objects in the coking environment and train the IP-R model to improve the prediction accuracy of the IP-R model for the motion trajectory of moving objects in the coking environment; use the trained IP-R model to predict the motion trajectory of moving objects in the coking environment to obtain the prediction result of the target object.

[0020] In step S1, the radial basis neural network is initialized and the structure of the radial basis neural network is determined. The position coordinates of moving objects (such as pedestrians, bicycles, and electric bicycles) in the coking environment that change over time are selected as the input variables of the radial basis neural network, and the position coordinates at the target time are used as the output variables. The number of neurons N in the hidden layer of the neural network is determined, and the transfer function in the neurons is determined at the same time. The Gaussian function with the radial basis function expansion speed control factor is used to establish the initial radial basis neural network architecture.

[0021] Please see Figure 2 , the radial basis neural network includes an input layer, a hidden layer and an output layer, wherein the input layer receives input data, the hidden layer uses radial basis functions to perform nonlinear transformation on the data, and the output layer linearly combines and outputs the output of the hidden layer. In the figure, Xi and Yi are the position coordinates at time i, thus constituting the input layer of the RBF; the Gaussian function used in the hidden layer is used as the activation function, and the Gaussian function maps the input data to a high-dimensional feature space, making the samples easier to distinguish in the high-dimensional space; Xi+1 and Yi+1 are the position coordinates at time i+1, respectively, and constitute the output layer. The expression of the activation function is as follows: in, is an n-dimensional input sample vector (n=2 in this model), It is The center vector of the hidden layer node determines the position of the basis function; It is The expansion coefficient or width of the basis function of the hidden node determines the degree of response of the basis function to the input sample. When it is larger, the response range of the basis function is wider, otherwise it is narrower.

[0022] The output layer uses linear optimization, and its main task is to adjust the linear weights and calculate the linear weighted sum of the hidden layer unit outputs. The expression is as follows: In the formula is the weight from the hidden layer to the output layer.

[0023] In step S2, the particle population is initialized, the average fitness change rate is selected, and a particle swarm optimization function for adaptively adjusting the inertia weight factor is constructed.

[0024] In the process of RBF neural network learning, it is necessary to continuously and synchronously correct the center and variance of the basis function. As well as the weights from the hidden layer to the output layer, until the set error target is reached. The adjustment of different parameters may affect each other, resulting in unstable results or failure to achieve the optimal result. It is necessary to find a better combination of parameters through repeated trials and adjustments. Therefore, the introduction of the PSO optimization algorithm can help the RBF neural network perform a global search in the parameter space, and can further speed up the parameter search and improve the training efficiency of the RBF neural network.

[0025] Particle Swarm Optimization (PSO) is an optimization algorithm based on swarm intelligence. It simulates the group behavior of birds and other groups to find the optimal solution. It has been successfully applied in many optimization problems. Its basic idea is to gradually approach the optimal solution by allowing multiple particles to move and communicate with each other in the search space, constantly updating their own positions and speeds.

[0026] Please see Figure 3 , the original particle swarm algorithm optimization steps are as follows: Step 1, initialize the position and velocity of the particle swarm, where each particle represents a set of RBF neural network parameters. The parameters include the target error (variance), the center point of the hidden layer, and the weight from the hidden layer to the output layer. The parameters of these three RBF neural networks are optimized through PSO.

[0027] Step 2, calculate the fitness of each particle, that is, use the RBF neural network with the current parameter configuration to predict the training data and calculate the fitness.

[0028] Step 3, update the individual optimal position and global optimal position of each particle, that is, update the optimal parameter configuration based on the comparison of the current fitness and the historical optimal fitness.

[0029] Step 4, update the speed and position of each particle by considering the current speed, individual optimal position and global optimal position, and iterate using a specific update formula.

[0030] Step 5, repeat Step 2- Step 4 until the requirements are met.

[0031] Step 6, return the parameter configuration corresponding to the global optimal position as the optimized RBF neural network parameters.

[0032] The above steps can be implemented using existing technologies and will not be described in detail here.

[0033] In the particle optimization process, the particle speed It is composed of the particle's inertial velocity, particle self-cognition and global information sharing. Using the inertial weight factor It can change the algorithm's global search ability and local optimization ability. When it is large, the global search capability is strong. When it is smaller, the local optimization ability is strong. As the number of iterations increases, the inertia weight factor decreases linearly.

[0034] The speed update formula is: In the formula For particles of The speed of the dimension; and is the learning factor, ranging from [0,4]; It is a particle of Dimension individual optimal position; It is a particle of Dimensional global optimal position; and are two independent random numbers, ranging from 0 to 1; is the inertia weight factor.

[0035] From the particle speed update formula, we can know that the main influencing factors in the particle swarm algorithm optimization process are: inertia weight w, individual learning factor c1 and social learning factor c2.

[0036] The inertia weight factor w determines the continuity of the particle speed. A larger w value helps the particle maintain its original direction and speed, thereby enhancing the global search ability. A smaller w value makes the particle more susceptible to the influence of individual and social learning factors, enhancing the local search ability. The individual learning factor c1 controls the particle to search based on its own experience. A higher c1 value will prompt the particle to search more within its own local range. If c1 is too large, it is easy for the particle to search too much locally and ignore the exploration of the global optimal solution. The social learning factor c2 controls the particle to search based on the group experience. A higher c2 will attract more particles to the group's optimal position and fall into the local optimal solution too early. Therefore, in the early stage of the algorithm's search, the diversity of particles in the group can be increased by setting a larger c1 value and a smaller c2 value. As the number of iterations increases, gradually adjust c1 to decrease linearly and c2 to increase linearly, thereby enhancing the particle's ability to converge to the global optimal point.

[0037] The focus of this embodiment is not on c1 and c2, but mainly on optimizing the inertia weight factor w to study the impact on the IP-R algorithm. Specifically, the following steps are included: Select the average fitness change rate , the formula is: in, is the average fitness of the particle at this moment, It is the difference between the average fitness of the particle at two consecutive moments.

[0038] Based on the above average fitness change rate , this embodiment improves the inertia weight function and obtains a particle swarm optimization function for adaptively adjusting the inertia weight factor: Where t is the current iteration number, is the maximum number of iterations, is the maximum inertia weight factor, is the minimum inertia weight factor.

[0039] In step S3, based on the particle swarm optimization function, the parameters of the radial basis neural network are optimized to obtain the IP-R model. The particle swarm optimization function of the IPSO algorithm for adaptively adjusting the inertia weight factor is: The parameter combination in the radial basis neural network is optimized to help the radial basis function find the optimal parameters (such as variance, hidden layer center point, and weight from hidden layer to output layer), thereby improving its prediction accuracy.

[0040] In step S4, the motion data of the moving objects in the coking environment are collected, and the IP-R model is trained to improve the prediction accuracy of the IP-R model for the motion trajectory of the moving objects in the coking environment; the trained IP-R model is used to predict the motion trajectory of the moving objects in the coking environment to obtain the prediction result of the target object.

[0041] This embodiment collects 5 sets of data of three types of moving targets in a simulated coking environment: pedestrians, bicycles, and electric bicycles, each with 15 two-dimensional coordinate points. Four sets of data are randomly selected from each data set as training sets, and the remaining set of data is used as a test set.

[0042] In order to more intuitively evaluate the performance differences of different models, this paper uses a test set with three types of moving targets as the performance indicator to evaluate different models.

[0043] The RMSE formula is as follows: Among them, pred is the predicted value and true is the measured value.

[0044] In order to verify the effectiveness and superiority of the improved IP-R algorithm in predicting the motion trajectory of mobile targets in the harsh environment of coking, the IP-R model provided in this embodiment is compared with the original PSO-RBF model and HGWO-SVM model.

[0045] Figure 4 The motion trajectory prediction diagrams of the IP-R model, PSO-RBF model and HGWO-SVM model in the same data test set under the coking environment are shown, where the arrows represent the direction of motion. It can be clearly seen that no matter what kind of moving target, the motion trajectory direction predicted by the IP-R model is more consistent with the line connecting the collected coordinate points, which shows that the prediction results of the improved algorithm are closer to the actual motion trend and trajectory of the moving target.

[0046] Table 1 shows the running time of the three models IP-R, PSO-RBF and HGWO-SVM in four groups of pedestrian training data sets. It can be seen from Table 1 that compared with the original PSO-RBF model, the HGWO-SVM model and the IP-R model can improve the convergence speed of the algorithm and reduce the running time. Among them, the IP-R algorithm can significantly shorten the running time of the algorithm, which can be reduced by 10~12s.

[0047] Table 1. Comparison of running time At the same time, in order to show the performance of the IP-R algorithm in three sets of four data sets, the RMSE value of IP-R in these 12 sets of data sets was calculated, and the calculation results were added according to the type of data set to obtain the average value. Finally, the average RMSE of the three data sets of pedestrians, bicycles, and electric bicycles was 0.151321, 0.144374, and 0.135672, respectively, which shows that the movement trajectory of pedestrians is more difficult to predict than that of bicycles and electric bicycles. Because pedestrians are more likely to change direction significantly within a small range during walking, if you want to predict the movement trajectory of pedestrians more accurately, it is not enough to rely on position coordinates alone, but also needs to be combined with other factors such as the environment and the pedestrian's own walking gait.

[0048] In addition, the various performance prediction evaluation indicators of the neural network sample data of the three models in the same test set of three different types of data sets, namely pedestrians, bicycles, and electric bicycles, are shown in Table 2. From Table 2, it can be seen that the root mean square error of the improved IP-R algorithm for the prediction of the trajectory position of each group is smaller than that of the original PSO-RBF and HGWO-SVM, which is reduced by 0.15 and 0.1 respectively. This shows that the improved algorithm has high accuracy. The experiment only predicts the direction of movement of the object, combined with Figure 4 From the motion trajectory prediction diagram, it can be found that the improved IP-R prediction algorithm has a smaller root mean square error, a more accurate overall prediction direction, less running time and higher efficiency, and can be applied to the motion trajectory prediction of pedestrians and non-motor vehicles in the area.

[0049] Table 2 RMSE of moving target position prediction This embodiment aims at the trajectory prediction of mobile targets in a typical coking environment in a carbon deposition environment, and uses the PSO-RBF algorithm as a reference to adaptively improve the inertia weight factor. The motion trajectories of three types of mobile targets in a simulated coking environment were collected for simulation experiments. Compared with the prediction results of the original PSO-RBF algorithm, the improved IP-R algorithm can accurately predict the motion trajectories of pedestrians and two non-motor vehicles in the coking environment, with a shorter running time and a smaller RMSE value, and can be applied to the prediction of the motion trajectories of pedestrians and non-motor vehicles in the region.

[0050] Embodiment 2 Based on the same concept, this embodiment provides a motion trajectory prediction device based on the IP-R algorithm. Figure 5 , the device comprises: RBF module, used to initialize the radial basis neural network and determine the structure of the radial basis neural network; IPSO module is used to initialize the particle population, select the average fitness change rate, and construct a particle swarm optimization function that adaptively adjusts the inertia weight factor; The IP-R module is used to optimize the parameters of the radial basis neural network based on the particle swarm optimization function to obtain the IP-R model; A training module is used to collect motion data of moving objects in a coking environment and train the IP-R model to improve the prediction accuracy of the IP-R model for the motion trajectory of moving objects in the coking environment; The prediction module is used to predict the motion trajectory of moving objects in the coking environment using the trained IP-R model to obtain the prediction result of the target object.

[0051] Among them, the IPSO module is configured as: Based on the fitness function of the PSO algorithm, calculate the average fitness change rate , the formula is: in, is the average fitness of the particle at time t, is the difference in the average fitness of the particle at two consecutive moments; Based on the average fitness change rate, a particle swarm optimization function for adaptively adjusting the inertia weight factor is constructed to optimize the parameter combination in the radial basis neural network. The formula is: Where t is the current iteration number, is the maximum number of iterations, is the maximum inertia weight factor, is the minimum inertia weight factor.

[0052] Through the particle swarm optimization function, the parameter combination of the target error (variance), the center point of the hidden layer, the weight from the hidden layer to the output layer in the radial basis neural network is optimized.

[0053] The device is used to implement the method in the above-mentioned embodiment 1, and its functions and specific methods are not described in detail here.

[0054] Embodiment 3 The above-mentioned second embodiment describes in detail the motion trajectory prediction device based on the IP-R algorithm of the present invention from the perspective of modular functional entities. The following describes in detail the motion trajectory prediction device based on the IP-R algorithm of the present invention from the perspective of hardware processing.

[0055] Please see Figure 6 The motion trajectory prediction device 500 based on the IP-R algorithm may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 510 (for example, one or more processors) and a memory 520, and one or more storage media 530 (for example, one or more mass storage devices) storing application programs 533 or data 532. Among them, the memory 520 and the storage medium 530 can be short-term storage or permanent storage. The program stored in the storage medium 530 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the motion trajectory prediction device 500 based on the IP-R algorithm.

[0056] Furthermore, the processor 510 may be configured to communicate with the storage medium 530 to execute a series of instruction operations in the storage medium 530 on the motion trajectory prediction device 500 based on the IP-R algorithm.

[0057] The motion trajectory prediction device 500 based on the IP-R algorithm may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input and output interfaces 560, and / or one or more operating systems 531, such as Windows Serve, Vista, etc.

[0058] Those skilled in the art will understand that Figure 6 The structure of the motion trajectory prediction device based on the IP-R algorithm shown does not constitute a limitation on the motion trajectory prediction device based on the IP-R algorithm, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0059] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, which, when executed on a computer, cause the computer to execute the steps of the motion trajectory prediction method based on the IP-R algorithm in the first embodiment.

[0060] If the module in the second embodiment is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of software. The computer software is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0061] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and equipment can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0062] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Even if various changes are made to the present invention, if these changes fall within the scope of the claims of the present invention and their equivalents, they still fall within the protection scope of the present invention.

Claims

1. A motion trajectory prediction method based on IP-R algorithm, characterized in that: include: Initialize the radial basis neural network and determine the structure of the radial basis neural network; Initialize the particle population, select the average fitness change rate, and construct a particle swarm optimization function that adaptively adjusts the inertia weight factor; Based on the particle swarm optimization function, the parameters of the radial basis neural network are optimized to obtain the IP-R model; Collect motion data of moving objects in the coking environment and train the IP-R model to improve the prediction accuracy of the IP-R model for the motion trajectory of moving objects in the coking environment; The trained IP-R model is used to predict the motion trajectory of moving objects in the coking environment to obtain the prediction results of the target objects.

2. The motion trajectory prediction method based on the IP-R algorithm according to claim 1, characterized in that: The selecting of the average fitness conversion rate and constructing the particle swarm optimization function for adaptively adjusting the inertia weight factor further includes: Based on the fitness function of the PSO algorithm, calculate the average fitness change rate , the formula is: in, is the average fitness of the particle at time t, is the difference in the average fitness of the particle at two consecutive moments; Based on the average fitness change rate, a particle swarm optimization function for adaptively adjusting the inertia weight factor is constructed to optimize the parameter combination in the radial basis neural network. The formula is: Where t is the current iteration number, is the maximum number of iterations, is the maximum inertia weight factor, is the minimum inertia weight factor.

3. The motion trajectory prediction method based on the IP-R algorithm according to claim 1, characterized in that: The initialization of the particle population further comprises: Initialize the position and velocity of the particle population, each particle represents a set of radial basis neural network parameters; Calculate the fitness of each particle according to the individual optimal position of each particle and the global optimal position; and calculate the average fitness based on the fitness of each particle.

4. The motion trajectory prediction method based on the IP-R algorithm according to claim 1, characterized in that: Based on the particle swarm optimization function, the parameters of the radial basis neural network are further optimized including: Based on the particle swarm optimization function, an update formula is constructed to update the speed and position of each particle until the preset requirements are met; the update formula is: in, For particles of The speed of the dimension; and is the learning factor; It is a particle of Dimension individual optimal position; It is a particle of Dimensional global optimal position; and are two independent random numbers, ranging from 0 to 1; It is a particle of The location of the dimension; is the inertia weight factor.

5. The motion trajectory prediction method based on the IP-R algorithm as claimed in claim 1, characterized in that: The IP-R model includes an input layer, a hidden layer and an output layer; The input layer includes n neurons representing n input variables, and there are Q training samples, x q ,n represents the nth variable of the qth sample, then for the qth sample, the neural network input is: x q =[x q ,1 ,x q ,2 ,...,x q ,N ] n=1 ,2 ,...,N The hidden layer uses a Gaussian function as the activation function of the hidden layer, and the formula is: in, is an n-dimensional input sample vector; It is The center vector of the hidden layer node determines the position of the basis function; It is The expansion coefficient of the basis function of the hidden layer nodes; The output layer performs linear weighted summation on the output of the hidden layer, and the formula is: In the formula, is the weight from the hidden layer to the output layer.

6. The motion trajectory prediction method based on the IP-R algorithm according to claim 1, characterized in that: The motion data of moving objects in the coking environment are collected, and the coordinates of the moving objects are used as feature vectors. The collected motion data are divided into a training set and a test set. The IP-R model is trained based on the training set, and the prediction accuracy of the IP-R model is evaluated based on the test set.

7. A motion trajectory prediction device based on IP-R algorithm, characterized in that: include: RBF module, used to initialize the radial basis neural network and determine the structure of the radial basis neural network; IPSO module is used to initialize the particle population, select the average fitness change rate, and construct a particle swarm optimization function that adaptively adjusts the inertia weight factor; The IP-R module is used to optimize the parameters of the radial basis neural network based on the particle swarm optimization function to obtain the IP-R model; A training module is used to collect motion data of moving objects in a coking environment and train the IP-R model to improve the prediction accuracy of the IP-R model for the motion trajectory of moving objects in the coking environment; The prediction module is used to predict the motion trajectory of the moving object in the coking environment using the trained IP-R model to obtain the prediction result of the target object.

8. A motion trajectory prediction device based on IP-R algorithm, characterized in that: The IPSO module is configured to: Based on the fitness function of the PSO algorithm, calculate the average fitness change rate , the formula is: in, is the average fitness of the particle at time t, is the difference in the average fitness of the particle at two consecutive moments; Based on the average fitness change rate, a particle swarm optimization function for adaptively adjusting the inertia weight factor is constructed to optimize the parameter combination in the radial basis neural network. The formula is: Where t is the current iteration number, is the maximum number of iterations, is the maximum inertia weight factor, is the minimum inertia weight factor.

9. A motion trajectory prediction device based on IP-R algorithm, characterized in that: include: A memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via a line; The at least one processor calls the instruction in the memory so that the motion trajectory prediction device based on the IP-R algorithm executes the motion trajectory prediction method based on the IP-R algorithm as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the motion trajectory prediction method based on the IP-R algorithm as described in any one of claims 1 to 6 is implemented.

Images