A deep learning-based disabled person travel path planning system and method

Abstract

The application provides a kind of system and method for path planning of disabled person based on deep learning, system includes: data acquisition module, for collecting the basic information of disabled person, historical behavior data and location environment data;Behavior data analysis module, for based on deep learning algorithm, the basic information of disabled person and historical behavior data are analyzed, obtain the characteristics of disabled person;Environment data analysis module, for processing location environment data, obtain the environmental characteristics of the location where disabled person is located;Path planning model construction module, for based on the travel demand of disabled person, combine the characteristics of disabled person and environmental characteristics, establish path planning model;Path planning module, for based on path planning model, provide path planning scheme for disabled person user.The technical scheme of the present application can realize more personalized, intelligent and dynamically adaptive path planning service, provide more convenient, safe and comfortable travel experience for disabled people.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of path planning, and particularly relates to a path planning system and method for disabled people based on deep learning. BACKGROUND

[0002] With the rapid development of artificial intelligence technology, deep learning, as an important branch of machine learning, has made remarkable achievements in image recognition, speech recognition, natural language processing and other fields in recent years. Deep learning can automatically learn and extract useful features from a large amount of data through the construction of deep neural network models, providing new ideas and methods for solving complex problems. In the field of path planning for disabled people, deep learning technology can be used to analyze the travel habits, preferences and environmental factors of disabled people, thereby providing more personalized path planning services.

[0003] Traditional path planning systems often ignore the special needs of disabled people, such as the need for barrier-free facilities, the adaptability of traffic flow, etc. Although some systems take into account the needs of disabled people, they lack personalized analysis and dynamic adjustment capabilities, making it difficult to meet the diverse needs of disabled people. SUMMARY

[0004] In view of the deficiencies of the prior art, the application provides a path planning system and method for disabled people based on deep learning, which has dynamic adjustment capability and can automatically optimize path planning schemes according to the real-time needs of disabled people and environmental changes.

[0005] To achieve the above-mentioned purpose, the application provides the following solutions:

[0006] A path planning system for disabled people based on deep learning, the system comprising:

[0007] a data acquisition module for acquiring basic information, historical behavior data and location environment data of disabled people;

[0008] a behavior data analysis module for analyzing the basic information and historical behavior data of disabled people based on a deep learning algorithm to obtain disabled person characteristics; wherein the disabled person characteristics include disability type, disability level and behavior characteristics;

[0009] an environmental data analysis module for processing the location environment data to obtain environmental characteristics of the location where the disabled person is located; wherein the environmental characteristics include traffic characteristics, map characteristics and weather characteristics;

[0010] a path planning model construction module for establishing a path planning model based on the travel needs of disabled people in combination with the disabled person characteristics and the environmental characteristics;

[0011] A path planning module is configured to provide a path planning scheme for the disabled user based on the path planning model.

[0012] Preferably, the behavior data analysis module comprises:

[0013] A behavior data processing unit is configured to clean the basic information and historical behavior data of the disabled person to obtain cleaned data.

[0014] A model construction unit is configured to construct a deep learning model based on a capsule network, and extract features from the cleaned data based on the deep learning model to obtain the disabled person features.

[0015] Preferably, the capsule network comprises a convolution layer for extracting first-order disabled person features, a capsule layer for extracting second-order disabled person features, and an output layer for outputting final disabled person features; and the capsule layer comprises a primary capsule layer and a digital capsule layer.

[0016] Preferably, the model construction unit comprises:

[0017] A squeeze-excitation establishment subunit is configured to construct a squeeze layer based on a global average pooling and construct an excitation layer based on a fully connected layer.

[0018] A squeeze-excitation insertion subunit is configured to insert the squeeze layer and the excitation layer into the primary capsule layer to obtain an improved primary capsule layer.

[0019] A feature extraction subunit is configured to input the first-order disabled person features extracted by the convolution layer into the improved primary capsule layer to obtain weighted second-order disabled person features.

[0020] A dynamic routing subunit is configured to dynamically allocate connection weights for the improved primary capsule layer and the digital capsule layer based on a dynamic routing mechanism, input the weighted second-order disabled person features into the digital capsule layer, and obtain a disabled person feature vector.

[0021] A feature output subunit is configured to output the final disabled person features based on the disabled person feature vector.

[0022] Preferably, in the environment data analysis module,

[0023] The traffic features comprise road flatness and width, distribution of barrier-free facilities, public transportation, traffic flow and congestion;

[0024] The map features comprise latitude and longitude, altitude, terrain and slope, building distribution, and green coverage;

[0025] The weather features comprise temperature and humidity, precipitation, and wind speed and direction.

[0026] Preferably, the path planning model construction module comprises:

[0027] A population initialization unit is configured to randomly generate a group of initial path planning schemes as an initial population of a genetic algorithm according to the travel demand of the disabled person and the characteristics of the disabled person.

[0028] An adaptability evaluation unit is configured to evaluate the adaptability of each initial path planning scheme according to the environmental characteristics and the characteristics of the disabled person, wherein the evaluation indexes include path length, barrier-free facility coverage rate and travel time.

[0029] A selection unit is configured to select the initial path planning scheme whose adaptability meets a preset threshold as a parent for generating a next generation of offspring according to the adaptability evaluation results.

[0030] A crossover and mutation unit is configured to perform crossover and mutation operations on the selected parent, and introduce a weight disturbance mechanism to dynamically adjust the weights of different environmental characteristics according to the characteristics of the disabled person, and generate new offspring path planning schemes.

[0031] A local search unit is configured to perform local search on each new offspring path planning scheme generated using a simulated annealing algorithm to obtain the adaptability value of the new offspring path planning scheme, and select the offspring path planning scheme whose adaptability value meets a preset target adaptability value as an optimal path planning scheme, thereby completing the construction of the path planning model.

[0032] Preferably, the system further comprises a reinforcement learning module configured to collect feedback data of the disabled person's travel, and construct a reward function based on the feedback data, and optimize the path planning model through the reward function.

[0033] The application also provides a deep learning-based path planning method for the travel of disabled persons, which applies the system and comprises the following steps:

[0034] Collecting basic information, historical behavior data and location environment data of the disabled person;

[0035] Analyzing the basic information and historical behavior data of the disabled person based on a deep learning algorithm to obtain the characteristics of the disabled person, wherein the characteristics of the disabled person include the type of disability, the level of disability and the behavior characteristics.

[0036] Processing the location environment data to obtain the environmental characteristics of the location where the disabled person is located, wherein the environmental characteristics include traffic characteristics, map characteristics and meteorological characteristics.

[0037] Based on the travel demand of the disabled person, the characteristics of the disabled person and the environmental characteristics, a path planning model is established.

[0038] Based on the path planning model, a path planning scheme is provided for the disabled person user.

[0039] Preferably, the method for obtaining the characteristics of the disabled person comprises:

[0040] cleaning the basic information and historical behavior data of the disabled person to obtain cleaned data;

[0041] constructing a deep learning model based on a capsule network, and extracting features from the cleaned data based on the deep learning model to obtain the characteristics of the disabled person;

[0042] The capsule network comprises a convolution layer for extracting first-order characteristics of the disabled person, a capsule layer for extracting second-order characteristics of the disabled person, and an output layer for outputting final characteristics of the disabled person; and the capsule layer comprises a primary capsule layer and a digital capsule layer.

[0043] Compared with the prior art, the beneficial effects of the present application are as follows: the data acquisition module provides a rich data basis for the system, which is helpful for subsequent analysis and path planning. The behavior data analysis module can more accurately understand the travel needs and preferences of the disabled person through a deep learning algorithm, and provide more accurate guidance for subsequent path planning. The environment data analysis module provides comprehensive environmental information support for path planning, which is helpful for planning a safer and more convenient travel route. The path planning model construction module can provide more personalized and intelligent travel services for the disabled person by constructing a special path planning model. In summary, the technical scheme of the present application realizes personalized path planning by accurately analyzing the characteristics of the disabled person through a deep learning algorithm. The influence of environmental characteristics on path planning is comprehensively considered, which improves the feasibility and safety of the path. The system has dynamic adjustment capability and can automatically optimize the path planning scheme according to the real-time needs of the disabled person and environmental changes. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical scheme of the present application, the following briefly introduces the drawings needed in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0045] Figure 1 The figure is a structure diagram of the disabled person travel path planning system based on deep learning of the embodiment of the present application. DETAILED DESCRIPTION

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] Example 1

[0049] like Figure 1 As shown, a deep learning-based travel route planning system for people with disabilities includes: a data acquisition module, a behavioral data analysis module, an environmental data analysis module, a route planning model construction module, and a route planning module.

[0050] The data acquisition module collects basic information, historical behavioral data, and location environment data for people with disabilities. Basic information includes name, age, gender, medical history, and disability type (e.g., visual impairment, hearing impairment, physical disability, etc., which are crucial factors to consider when planning accessible travel routes). Historical behavioral data includes historical travel records (travel time, origin, destination, etc.) and preference settings (travel frequency, route selection, mode of transportation, etc.). Location environment data includes real-time or predicted traffic flow, congestion, traffic accidents, road layout, intersections, sidewalks, detailed locations of accessible facilities (e.g., tactile paving, ramps), weather conditions such as rain, snow, strong winds, and high temperatures, and the location and availability of elevators, accessible restrooms, wheelchair ramps, etc. By collecting a large amount of multi-source heterogeneous data, data support is provided for route planning.

[0051] The behavioral data analysis module is used to analyze the basic information and historical behavioral data of people with disabilities based on deep learning algorithms to obtain their characteristics. These characteristics include disability type, disability level, and behavioral features.

[0052] A further implementation method includes a behavioral data analysis module comprising:

[0053] The behavior data processing unit is configured to clean basic information and historical behavior data of the disabled person, and obtain cleaned data. Specifically, in the embodiment, according to a pre-established data mapping rule and a standardized template, field mapping and format conversion are performed on the cleaned historical behavior heterogeneous data, data of different sources are converted into a unified standardized format, and a structured standardized data set is formed. Data fusion technology is adopted to remove and merge the standardized data, and multiple data records describing the same entity are integrated into a complete data record, so that the completeness and consistency of the data are improved. Through data correlation analysis, the correlation between different data tables and fields is mined, and a correlation graph between the data is constructed, thereby providing support for subsequent data analysis and application.

[0054] The model construction unit is configured to construct a deep learning model based on a capsule network, and perform feature extraction on the cleaned data based on the deep learning model, to obtain disabled person features.

[0055] The capsule network is a set of neurons, including a convolution layer for extracting first-order disabled person features, a capsule layer for extracting second-order disabled person features, and an output layer for outputting final disabled person features. The capsule layer includes a primary capsule layer and a digital capsule layer.

[0056] Further embodiments are configured to include the model construction unit including:

[0057] The squeeze-excitation establishment subunit is configured to construct a squeeze layer based on a global average pooling and construct an excitation layer based on a fully connected layer.

[0058] The squeeze-excitation insertion subunit is configured to insert the squeeze layer and the excitation layer into the primary capsule layer, to obtain an improved primary capsule layer.

[0059] The feature extraction subunit is configured to input the first-order disabled person features extracted by the convolution layer into the improved primary capsule layer, to obtain weighted second-order disabled person features.

[0060] The dynamic routing subunit is configured to dynamically allocate connection weights for the improved primary capsule layer and the digital capsule layer based on a dynamic routing mechanism, input weighted second-order disability features into the digital capsule layer, and obtain a disability feature vector; in this embodiment, the connection weights are calculated and the output of the digital capsule is updated according to the input second-order features and the current state of the digital capsule. Specifically, after the first-order disability features are input into the primary capsule layer, the first-order disability features are rearranged to obtain reconstructed first-order disability features. The prediction vector matrix is obtained by multiplying the reconstructed first-order disability features by their corresponding weights, and the second-order disability features are obtained by weighting the prediction vector. These second-order features not only contain information about local features, but also emphasize the parts related to disability features through weighting processing. The coupling coefficient between the primary capsule layer and the digital capsule layer is integrated to 1, which determines the performance of the capsule network routing algorithm. After multiple iterations, the digital capsule layer outputs a feature vector, which represents the overall features of the disabled person. The length and direction of the feature vector represent the existence probability and specific attributes of the disability features, respectively. In particular, the digital capsule layer obtained after multiple routing is divided into two branches, one of which is used to calculate the length of the feature classifier, and the other branch is used to reconstruct the original data.

[0061] The feature output subunit is configured to output the final disability features based on the disability feature vector.

[0062] The environmental data analysis module is configured to process location environmental data to obtain environmental features of the location where the disabled person is located; wherein the environmental features include traffic features, map features, and weather features; in a further embodiment, in the environmental data analysis module, the traffic features include road flatness and width, distribution of barrier-free facilities, public transportation, traffic flow and congestion; the map features include latitude and longitude, altitude, terrain and slope, building distribution, and green coverage;

[0063] The weather features include temperature and humidity, precipitation, and wind speed and direction.

[0064] The path planning model construction module is configured to establish a path planning model based on the travel needs of the disabled person, in combination with the disability features and the environmental features; in a further embodiment, the path planning model construction module includes:

[0065] The population initialization unit is configured to randomly generate a set of initial path planning schemes as the initial population of the genetic algorithm according to the travel needs of the disabled person and the disability features;

[0066] The fitness evaluation unit is configured to evaluate the fitness of each initial path planning scheme according to the environmental features and the disabled person features; wherein the evaluation indexes include path length, barrier-free facility coverage rate and travel time; in this embodiment, the traffic environment factor (road congestion, accident situation, etc.) is introduced into the evaluation indexes to disturb the weights of the evaluation indexes, and the temperature annealing factor in the simulated annealing algorithm is introduced to make the fitness function adaptively change by controlling the temperature annealing factor. In particular, the weights of the evaluation indexes are disturbed according to the user portrait of the current disabled user based on the disabled person features of the disabled user, and different weight distribution schemes can be preset for different types of disabled users based on the user portrait.

[0067] The selection unit is configured to select the initial path planning scheme with fitness satisfying a preset threshold as a parent according to the fitness evaluation result, and to generate a next generation of offspring.

[0068] The crossover and mutation unit is configured to perform crossover and mutation operations on the selected parent, and introduce a weight disturbance mechanism to dynamically adjust the weights of different environmental features according to the disabled person features to generate a new offspring path planning scheme; in particular, the adaptive mutation probability is calculated based on a constant factor, a heuristic mutation factor and preset upper and lower limit values of the adaptive mutation probability; in this embodiment, the temperature annealing factor is introduced into the calculation formula of the adaptive mutation probability, and the adaptive mutation probability is calculated when the fitness value of the mutated individual is greater than the average fitness value and the temperature annealing factor is greater than a preset threshold. The adaptive mutation probability is used to control whether the parent is mutated.

[0069] The local search unit is configured to perform local search on each new offspring path planning scheme generated by using the simulated annealing algorithm to obtain the fitness value of the new offspring path planning scheme, and select the offspring path planning scheme with fitness satisfying a preset target fitness value as the optimal path planning scheme to complete the construction of the path planning model.

[0070] The path planning module is configured to provide a path planning scheme for the disabled person user based on the path planning model.

[0071] Further embodiments are that the system further comprises a reinforcement learning module configured to collect disabled person travel feedback data and construct a reward function based on the feedback data to optimize the path planning model by using the reward function.

[0072] In this embodiment, the path planning model is continuously optimized according to user feedback:

[0073] First, the state space and the action space are defined, wherein the state space includes the current position, the destination, the traffic condition, the environmental features, etc.; and the action space includes possible path selection.

[0074] Then, a reward function is designed based on user feedback (satisfaction score, frequency of choosing a certain path). For example, if users prefer to avoid crowded areas, a positive reward is given when the model recommends a path that avoids crowded areas. The reinforcement learning module continuously optimizes the path planning model through the method of policy iteration. In each iteration, the module selects a path according to the current policy and collects feedback data to update the reward function. According to the feedback of the reward function, the module adjusts the parameters of the path planning model to maximize the cumulative reward. This process is repeated until the model converges to the optimal solution or reaches the predetermined number of iterations.

[0075] Because the travel trajectory data of disabled users contains location coordinates, timestamps, travel purposes, etc. Take wheelchair users as an example, through the mobile application, record the start and end point, route selection, time consumption, etc. of each trip. Users can score and evaluate each trip experience, such as whether the route is convenient, whether the accessibility facilities are perfect, etc. These data constitute the basic data set for model optimization. In the model optimization process, the reward function optimization method. For example, a user feedback that the sidewalk of a certain path is too narrow and not suitable for wheelchair travel generates a negative reward function, and the system will reduce the weight of this road segment in path planning according to the negative reward function. With the continuous accumulation of feedback data, the model is gradually optimized, and the recommendation effect continues to improve. By analyzing the user's historical travel data, we can find out their travel preferences. For example, a visually impaired user is used to choosing a road segment with voice prompt equipment, and the system will take this preference as an important reference for path planning. If a user often goes to a certain place at 8 am, the system will record this rule and actively recommend appropriate routes during the corresponding period. User satisfaction evaluation directly affects the path recommendation strategy. Set the satisfaction threshold to 3.5 out of 5, if the average score of a certain path is lower than the threshold, the system will accordingly reduce its recommendation frequency. For example, a certain path is temporarily diverted due to construction, and the user's score is low, so the system will adjust the recommendation strategy in time. Collaborative filtering algorithm finds similar user groups by analyzing user characteristics. For example, users who also use wheelchairs and often move in the same area have similar path selection. The system will filter out suitable routes for the current user from the high-score paths of similar users and recommend them. When considering multiple factors, the system gives different weights to different factors. The weight of accessibility facility perfection is 0.4, the weight of pedestrian density is 0.3, and the weight of road slope is 0.3. For example, although a certain commercial area is crowded, it has complete accessibility facilities and a gentle slope, it may still be recommended by the system. The system continuously tracks the path usage, records whether the user drives according to the recommended route, whether the route is changed along the way, whether the actual time consumption meets the expectations, etc. These data are used to evaluate the recommendation effect and serve as an important basis for model optimization. When it is found that the actual travel time of a recommended path is generally longer than the estimated value, the system will adjust the time estimation model accordingly.

[0076] Embodiment two

[0077] The application also provides a deep learning-based path planning method for the disabled, and an application system, comprising:

[0078] Basic information, historical behavior data and location environment data of the disabled are collected;

[0079] The basic information and historical behavior data of the disabled are analyzed based on a deep learning algorithm to obtain the characteristics of the disabled, wherein the characteristics of the disabled include the type of disability, the level of disability and the behavior characteristics;

[0080] The location environment data are processed to obtain the environmental characteristics of the location where the disabled are located, wherein the environmental characteristics include traffic characteristics, map characteristics and meteorological characteristics;

[0081] Based on the travel needs of the disabled, the characteristics of the disabled and the environmental characteristics, a path planning model is established;

[0082] Based on the path planning model, a path planning scheme is provided for the disabled users.

[0083] Further embodiments are that the method for obtaining the characteristics of the disabled comprises:

[0084] The basic information and historical behavior data of the disabled are cleaned to obtain cleaned data;

[0085] A deep learning model based on a capsule network is constructed, and the cleaned data are subjected to feature extraction based on the deep learning model to obtain the characteristics of the disabled;

[0086] The capsule network comprises a convolution layer for extracting first-order characteristics of the disabled, a capsule layer for extracting second-order characteristics of the disabled and an output layer for outputting final characteristics of the disabled; the capsule layer comprises a primary capsule layer and a digital capsule layer.

[0087] The above-described embodiments are only descriptions of the preferred modes of the application and do not limit the scope of the application. Without departing from the design spirit of the application, various modifications and improvements to the technical solutions of the application made by those skilled in the art shall fall within the protection scope of the claims of the application.

Claims

1. A deep learning-based path planning system for a person with a disability, characterized by, The system comprises: a data acquisition module for acquiring basic information, historical behavior data and location environment data of the disabled person; a behavior data analysis module for analyzing the basic information and historical behavior data of the disabled person based on a deep learning algorithm to obtain the disabled person characteristics; wherein the disabled person characteristics include disability type, disability level and behavior characteristics; an environment data analysis module for processing the location environment data to obtain the environmental characteristics of the location where the disabled person is located; wherein the environmental characteristics include traffic characteristics, map characteristics and weather characteristics; a path planning model construction module for establishing a path planning model based on the travel needs of the disabled person in combination with the disabled person characteristics and the environmental characteristics; a path planning module for providing a path planning scheme for the disabled person user based on the path planning model; the behavior data analysis module comprises: a behavior data processing unit for cleaning the basic information and historical behavior data of the disabled person to obtain cleaned data; a model construction unit for constructing a deep learning model based on a capsule network and performing feature extraction on the cleaned data based on the deep learning model to obtain the disabled person characteristics; wherein the capsule network comprises a convolution layer for extracting first-order disabled person characteristics, a capsule layer for extracting second-order disabled person characteristics, and an output layer for outputting final disabled person characteristics; the capsule layer comprises a primary capsule layer and a digital capsule layer; the model construction unit comprises: a squeeze excitation establishment subunit for constructing a squeeze layer based on a global average pooling and constructing an excitation layer based on a fully connected layer; a squeeze excitation insertion subunit for inserting the squeeze layer and the excitation layer into the primary capsule layer to obtain an improved primary capsule layer; a feature extraction subunit for inputting the first-order disabled person characteristics extracted by the convolution layer into the improved primary capsule layer to obtain weighted second-order disabled person characteristics; a dynamic routing subunit for dynamically assigning connection weights to the improved primary capsule layer and the digital capsule layer based on a dynamic routing mechanism, inputting the weighted second-order disabled person characteristics into the digital capsule layer to obtain a disabled person characteristic vector; a feature output subunit for outputting the final disabled person characteristics based on the disabled person characteristic vector.

2. The deep learning-based travel path planning system for disabled persons according to claim 1, characterized by, In the environment data analysis module, the traffic characteristics include road flatness and width, barrier-free facility distribution, public transportation, traffic flow and congestion; the map characteristics include latitude and longitude and altitude, terrain and slope, building distribution and green coverage; the weather characteristics include temperature and humidity, precipitation and wind speed and direction. 3.The deep learning-based disabled person travel path planning system of claim 1, wherein, The path planning model construction module comprises: a population initialization unit for randomly generating a set of initial path planning schemes as the initial population of a genetic algorithm according to the travel needs of the disabled person and the disabled person characteristics; a fitness evaluation unit for evaluating the fitness of each initial path planning scheme according to the environmental characteristics and the disabled person characteristics; wherein the evaluation indexes include path length, barrier-free facility coverage and travel time; The selection unit is configured to select, according to the fitness evaluation results, an initial path planning scheme whose fitness satisfies a preset threshold as a parent for generating a next generation of offspring; The crossover and mutation unit is configured to perform crossover and mutation operations on the selected parent, and introduce a weight disturbance mechanism to dynamically adjust the weights of different environmental characteristics according to the characteristics of the disabled person, to generate a new offspring path planning scheme; The local search unit is configured to perform local search on each new offspring path planning scheme generated using a simulated annealing algorithm to obtain the fitness value of the new offspring path planning scheme, and select an offspring path planning scheme whose fitness value satisfies a preset target fitness value as an optimal path planning scheme, thereby completing the construction of the path planning model. 4.The deep learning-based disabled person trip path planning system of claim 1, wherein, The system further comprises a reinforcement learning module configured to collect feedback data of the disabled person's travel, and construct a reward function based on the feedback data, and optimize the path planning model through the reward function.

5. A deep learning-based path planning method for the travel of disabled people, applying the system of any one of claims 1-4, characterized in that, The method comprises: collecting basic information, historical behavior data and location environment data of the disabled person; analyzing the basic information and historical behavior data of the disabled person based on a deep learning algorithm to obtain characteristics of the disabled person, wherein the characteristics of the disabled person include disability type, disability level and behavior characteristics; processing the location environment data to obtain environmental characteristics of the location where the disabled person is located, wherein the environmental characteristics include traffic characteristics, map characteristics and weather characteristics; establishing a path planning model based on the travel needs of the disabled person in combination with the characteristics of the disabled person and the environmental characteristics; providing a path planning scheme for the disabled person based on the path planning model. 6.The deep learning-based path planning method for the disabled person traveling according to claim 5, wherein, The method for obtaining the characteristics of the disabled person comprises: cleaning the basic information and historical behavior data of the disabled person to obtain cleaned data; constructing a deep learning model based on a capsule network, and performing feature extraction on the cleaned data based on the deep learning model to obtain the characteristics of the disabled person; wherein the capsule network comprises a convolution layer for extracting first-order characteristics of the disabled person, a capsule layer for extracting second-order characteristics of the disabled person, and an output layer for outputting final characteristics of the disabled person; and the capsule layer comprises a primary capsule layer and a digital capsule layer.

Images