Whole-course accompanying robot for disabled old people and automatic control system thereof

Abstract

The invention discloses a whole-course accompanying robot for disabled old people and an automatic control system of the whole-course accompanying robot, relates to the technical field of intelligent accompanying, and is used for solving the problems of inaccurate positioning, untimely task response and poor service continuity in the accompanying process of the disabled old people. The multi-modal perception and state analysis module divides the accompanying environment into a plurality of sub-regions based on region characteristics, and accurately perceives the position and state change of the disabled elderly. And the task generation and scheduling execution module intelligently judges service requirements of disabled old people, and dynamically screens optimal robot resources for task matching and execution. The task feedback closed-loop module collects and analyzes task execution process and result data, and timely rechecks the state of the old person according to feedback. And the self-control module adjusts the accompanying strategy and periodically reviews the priority of the task queue. Through cooperative operation of the above modules, the system can achieve precise perception, intelligent decision and dynamic optimization of accompanying of disabled old people, and remarkably improves continuity, intelligence and reliability of accompanying service.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent companionship, and more specifically, to a robot for full-course companionship of disabled elderly people and its automatic control system. Background Art

[0002] Currently, with the accelerating process of social aging, the number of disabled elderly people has increased rapidly. The companionship and care issues for this group have become important social issues that need to be solved urgently. The traditional companionship mode mainly relies on manual inspections, fixed-point services, and passive responses, and has prominent problems such as high work intensity, many service blind spots, poor real-time performance, and insufficient personalization, and cannot meet the actual needs of disabled elderly people for continuous, accurate, and efficient companionship services. Especially in scenarios with complex environments, dense personnel, or limited resources, the limitations of manual companionship are more obvious, easily leading to problems such as positioning errors, care delays, and service mismatches, seriously affecting the life safety and quality of life of the elderly.

[0003] With the development of technologies such as perception technology, intelligent robots, autonomous decision-making, and task scheduling, some companionship systems have introduced multi-modal perception and robot-assisted execution. However, most existing solutions stay at the basic perception and simple task response levels, lacking precise positioning based on dynamic management of sub-regions, lacking the ability to generate differentiated tasks according to the changes in the state of the elderly, and lacking an efficient task feedback closed-loop and an adaptive optimization mechanism for companionship strategies, resulting in a low intelligent level, poor flexibility of the overall system, and inability to make an efficient response to the real-time changes in the state of the elderly.

[0004] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a robot for full-course companionship of disabled elderly people and its automatic control system to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions: In a preferred embodiment, it includes: a multi-modal perception and state analysis module, a task generation and scheduling execution module, a task feedback closed-loop module, and an automatic control module, and the modules are signal-connected to each other; The multi-modal perception and state analysis module is mainly used to divide the companionship area into multiple sub-regions and configure differentiated positioning schemes for the characteristics of different sub-regions; The task generation and scheduling execution module is mainly used to judge the service state required by the disabled elderly people, dynamically screen and schedule the optimal robot resources to efficiently execute service tasks; The task feedback closed-loop module is mainly used to monitor the execution effect of the task, and quickly review and dynamically update the state of the disabled elderly people based on the feedback data; The automatic control module is mainly used to dynamically optimize the accompanying strategy, periodically re-evaluate and sort the task queue, and control the execution of tasks.

[0007] In a preferred embodiment, the multimodal perception and state analysis module collects the distance and angle in the escort area, generates an environmental map of the escort area, and determines the real-time positioning of the robot.

[0008] In a preferred embodiment, the multimodal perception and state analysis module divides the entire accompanying area unevenly into multiple sub-areas, and determines the disabled elderly positioning influence coefficient of the sub-area by combining the electromagnetic interference data and furniture material characteristic data in the sub-area; sets the affected threshold YL of the AMCL algorithm, and when the disabled elderly positioning influence coefficient L of the sub-area is less than or equal to the affected threshold YL of the AMCL algorithm, it indicates that the sub-area belongs to a low L value area, and when the disabled elderly positioning influence coefficient L of the sub-area is greater than the affected threshold YL of the AMCL algorithm, it indicates that the sub-area belongs to a high L value area, and the positioning scheme for the disabled elderly is selected according to the different L values ​​of the sub-areas.

[0009] In a preferred embodiment, in the multimodal perception and state analysis module, for the overlap buffer zone, the spectrum fluctuation value and posture feature change rate of the bioelectric signal data of the disabled elderly within the unit time T1 are calculated, and based on the proportional relationship of the change amount, the influence weights of the bioelectric features and posture features are dynamically allocated to determine the dominant features of the disabled elderly in the overlap buffer zone, and the disabled elderly are located based on the dominant features.

[0010] In a preferred embodiment, in the multimodal perception and state analysis module, three-dimensional image data of the disabled elderly are obtained, and the height difference between the head and hips of the disabled elderly is calculated; the physiological parameters of the disabled elderly are collected, and the mean and standard deviation of each collected physiological parameter are calculated; and the emotional characteristics and voice data of the disabled elderly's face are obtained.

[0011] In a preferred embodiment, in the task generation and scheduling execution module, the service status of the disabled elderly is determined, and a task queue is created, a task pool is established, and tasks that need to be performed by the robot are screened; the current coordinate data of the robot and the coordinate data of the task target point are determined, and the shortest path and control instructions are calculated.

[0012] In a preferred embodiment, in the task feedback closed-loop module, the task data is written into the task log, and the current comprehensive status score of the disabled elderly is recalculated.

[0013] In a preferred embodiment, in the self-control module, a behavior state model of accompanying disabled elderly people is established, and the accompanying strategy is optimized through interactive feedback with the environment; according to the status and priority of the current task and the immediate status of the disabled elderly, the task queue is re-evaluated and sorted in each cycle.

[0014] In a preferred embodiment, the method for obtaining electromagnetic interference data is as follows: The electromagnetic interference sensors carried by the robot are used to collect the electromagnetic interference data at fixed points in each sub-region.

[0015] In a preferred embodiment, the method for obtaining furniture material characteristic data is as follows: Scan the main furniture in each sub-region, count the proportion of metal materials in the furniture in the sub-region, and calculate the furniture material characteristic data in the sub-region.

[0016] The technical effects and advantages of the robot for full-course escort of disabled elderly and its automatic control system of the present invention: By introducing the multi-modal perception and state analysis module, the present invention can finely divide the escort area, adopt different positioning strategies for different sub-regions, and achieve high-precision and dynamic positioning and tracking of disabled elderly. Through the task generation and scheduling execution module, it can judge the service demand status of the elderly in real time, intelligently screen and assign the robot to execute personalized tasks, ensuring timely response and efficient use of resources. Relying on the task feedback closed-loop module, it can dynamically update the elderly's status information according to the execution results, forming a continuously optimized service link. The automatic control module realizes the dynamic sorting of task priorities and the autonomous optimization of escort strategies through periodic evaluation and strategy adjustment, improving the flexibility and intelligent decision-making ability of the overall system operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the module structure of the robot for full-course escort of disabled elderly and its automatic control system of the present invention.

[0018] Figure 2 It is a functional flowchart of the robot for full-course escort of disabled elderly and its automatic control system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] Embodiment The present invention discloses a robot for full-course escort of disabled elderly and its automatic control system, as Figure 1 shown, including: a multi-modal perception and state analysis module, a task generation and scheduling execution module, a task feedback closed-loop module, and an automatic control module, and the modules are connected by signals.

[0021] AsFigure 2 As shown in Figure 2 , in the multi-modal perception and state analysis module, the robot's front lidar (2D Lidar) is used to collect distance and angle data within the escort area at a set frequency, and each scan forms a complete frame of lidar scan data. The collected lidar scan data is input into the Gmapping algorithm. At the same time, wheel odometer data is used to provide motion estimation. Then, within Gmapping, based on particle filtering, the pose of the robot and the environmental map are updated in real time to generate an environmental map of the escort area with markings for obstacles, free space, and unknown areas. Furthermore, the environmental map of the escort area, the current wheel odometer data of the robot, and the real-time lidar scan data are input into the AMCL algorithm. Through the particle filter, the possible positions of the robot are represented as a set of particles. The particle weights are continuously adjusted and resampled according to the comparison between the lidar data and the map. Finally, the real-time positioning of the robot in the coordinate system of the escort area environmental map is output, including: position coordinates and orientation angle.

[0022] Among them, the particle weight update formula is specifically as follows: wi = p(z|xi, m) Among them, wi represents the weight of the i-th particle, z represents the current lidar scan observation, xi represents the hypothesized pose represented by the i-th particle; m represents the environmental map of the escort area. Furthermore, considering that when there are multiple large-area metal objects in the indoor environment or there is strong electromagnetic interference in the environment, the positioning accuracy of the adaptive Monte Carlo localization and particle filter algorithm based on the AMCL algorithm will be affected, resulting in an increase in positioning error or even inability to position. Therefore, in this embodiment, before determining the real-time positioning of the disabled elderly, the entire escort area is unevenly divided into multiple sub-regions, each sub-region is defined as a rectangle or polygon, and the sub-region number and boundary coordinates are saved. Then, through the electromagnetic interference sensor carried by the robot, the sampling frequency is set to 1 Hz, and the continuous sampling time is set to T sampling seconds. The electromagnetic interference data within each sub-region is collected at fixed points within each sub-region, and the average electromagnetic interference value E within the sub-region is calculated based on the electromagnetic interference values of multiple sampling points, specifically according to the formula: Among them, N represents the number of sampling points, and Ei represents the electromagnetic interference value of the i-th sampling point; Use an RGB-D camera combined with a machine learning model (CNN network) to scan the main furniture within each sub-region, judge the material category (metal / wood / plastic, etc.), count the proportion of metal materials of the furniture within the sub-region (Mratio), and calculate the furniture material characteristic data within the sub-region, specifically according to the formula: Mratio = metal surface area / total detected surface area Further, by combining the obtained electromagnetic interference data and furniture material characteristic data in the sub-region, the influence coefficient L of the disabled elderly's positioning in this sub-region is determined, specifically according to the formula: L = Q1×GE + Q2×Mratio, where GE represents the normalized value of the electromagnetic interference data, Q1 represents the weight of the normalized value of the electromagnetic interference data, Q2 represents the weight value of the furniture material characteristic data. It should be noted that Q1 and Q2 are determined according to specific environmental tests respectively.

[0023] Further, a threshold value YL affected by the AMCL algorithm is set. When the robot is about to enter a certain sub-region, the influence coefficient L of the disabled elderly's positioning in this sub-region is compared with the threshold value YL affected by the AMCL algorithm. When the influence coefficient L of the disabled elderly's positioning in this sub-region is less than or equal to the threshold value YL affected by the AMCL algorithm, it indicates that this sub-region belongs to a low-L value region. For this type of region, the adaptive Monte Carlo localization and particle filter algorithm based on the AMCL algorithm are used to locate the disabled elderly. When the influence coefficient L of the disabled elderly's positioning in this sub-region is greater than the threshold value YL affected by the AMCL algorithm, it indicates that this sub-region belongs to a high-L value region, and the AMCL algorithm will fail due to the influence of indoor furniture materials and electromagnetic interference intensity. At this time, it is necessary to pause the particle sampling and weight update inside the AMCL, and maintain the current last valid pose data, freeze the output of the AMCL algorithm. Then, start a 360° omnidirectional laser scan to collect the current complete environmental point cloud data, and at the same time start the infrared thermal imaging + RGB-D vision detection model, and identify the human body contour based on the YOLO or pose estimation algorithm OpenPose. The iterative closest point algorithm ICP is used to register the point cloud human body contour with the lidar environmental features to determine the central coordinates of the disabled elderly.

[0024] Considering that in the sub-region division of the escort area, there is a coincidence buffer formed by the intersection of high-L value regions and low-L value regions. In the coincidence buffer, it is impossible to accurately and real-time locate the disabled elderly by the method of dynamically switching the positioning technology according to the sub-region L value. Therefore, when the disabled elderly stands or moves in the buffer area, the electrocardiogram ECG and electroencephalogram EEG sensors are used to collect the bioelectrical signal data of the disabled elderly within the unit time T1, and the frequency of the collected data is set to fbio. Then, the frequency domain analysis is carried out on the collected bioelectrical signal data through the Fourier transform FFT), and the spectral fluctuation value of the bioelectrical signal data within the unit time T1 is calculated, specifically according to the formula: where, σbio represents the fluctuation value of the bioelectrical signal data, Eei represents the i-th sampling point; μ represents the mean value of the bioelectrical signal data, and N represents the number of sampling points; Further, an infrared thermal imaging camera is used to collect the thermal imaging image data of the disabled elderly, form a thermal imaging image, apply the deep learning algorithm pose recognition network OpenPose to the thermal imaging image to identify the pose key points of the disabled elderly within a unit time T1, and extract the pose feature values, including: the relative positions and angles of key parts such as the legs, arms, and torso. Then, calculate the feature change rate of the pose feature values of the disabled elderly within the unit time T1, specifically according to the formula: where θi represents the pose angle at the i-th second, and T represents the time window size; Further, calculate the change rate of the bioelectricity feature fluctuation value of the disabled elderly within the unit time T1, specifically according to the formula: Δσbio = σbio(t) - σbio(t - 1) Take the change rate of the bioelectricity feature fluctuation value and the pose feature change rate of the disabled elderly within the unit time T1 as input features, and dynamically allocate the influence weights based on the proportional relationship of the change amounts, specifically according to the formula: where Wbio(t) represents the dynamic weight of the bioelectricity feature of the disabled elderly within the unit time T1; Wpose(t) represents the dynamic weight of the pose feature of the disabled elderly within the unit time T1.

[0025] Then, compare the dynamic weight of the bioelectricity feature and the dynamic weight of the pose feature of the disabled elderly within the unit time T1. When the dynamic weight of the bioelectricity feature of the disabled elderly within the unit time T1 is greater than or equal to the dynamic weight of the pose feature of the disabled elderly within the unit time T1, it indicates that the bioelectricity feature activity of the disabled elderly is high at this time, so select the bioelectricity feature as the dominant feature and adopt the positioning method for the high L-value area; when the dynamic weight of the bioelectricity feature of the disabled elderly within the unit time T1 is less than the dynamic weight of the pose feature of the disabled elderly within the unit time T1, it indicates that the pose feature activity of the disabled elderly is high at this time, so select the pose feature as the dominant feature and adopt the positioning method for the low L-value area; where, if the difference between the bioelectricity feature and the change rate of the thermal imaging pose feature is large, then the weights need to be readjusted to ensure improving the efficiency and accuracy of the robot's positioning of the disabled elderly.

[0026] It should be noted that the method of selecting the dominant feature for the disabled elderly through the bioelectricity feature fluctuation and the thermal imaging pose feature of the disabled elderly requires relatively high sensors and computing resources. Therefore, the usage cost of this method is relatively high, which is not applicable to ordinary scenarios and is limited to the special application scenarios of the overlapping buffer area.

[0027] After locating the disabled elderly, a depth camera Intel RealSense is configured on the robot to capture the spatial image of the disabled elderly, obtain the three-dimensional image data of the disabled elderly, and then use the YOLOv5 model to detect the key points of the human skeleton in the three-dimensional image data of the disabled elderly, identify the joint positions of the head, hips, torso, etc. of the disabled elderly. At the same time, each frame of the image is processed to obtain the coordinates of the key points of the skeleton of the disabled elderly.

[0028] After that, calculate the height difference Δh between the head and hips of the disabled elderly, specifically according to the formula: Δh = hhead - hhip where hhead represents the height coordinate of the head of the disabled elderly, and hhip represents the height coordinate of the hips of the disabled elderly. Set the standing threshold H1 according to the Δh value when the disabled elderly stands normally. Compare the height difference Δh between the head and hips of the disabled elderly with the standing threshold H1. When the height difference Δh between the head and hips of the disabled elderly exceeds the standing threshold H1, it is considered that the disabled elderly may have fallen at this time.

[0029] After that, configure wearable physiological sensors for the disabled elderly to collect physiological parameters such as heart rate, body temperature, and blood pressure of the disabled elderly. Set the sliding window size W, and use the sliding window algorithm to calculate the mean μHR(t) and standard deviation σHR(t) of each physiological parameter collected by the wearable physiological sensors, specifically according to the formula: where HR(i) represents the heart rate value at the i-th moment.

[0030] Input the mean and standard deviation of each calculated physiological parameter into the health node of ROS for analysis, and use the Z-Score anomaly detection algorithm to determine whether the disabled elderly is in an abnormal physiological health state.

[0031] Furthermore, according to the facial image of the disabled elderly collected by the camera, use the CNN convolutional neural network to perform facial expression recognition on the disabled elderly, and extract the emotional features (such as happy, sad, painful, etc.) of the face of the disabled elderly. At the same time, collect the voice data of the disabled elderly through the microphone array, and extract the pitch and speech rate of the voice as emotional features. Use a support vector machine (SVM) classifier to judge the voice emotion according to the pitch and speech rate data: y = f(X), where X represents the feature vector of the pitch and speech rate, and f(X) represents the emotion label output by the SVM classifier.

[0032] In the task generation and scheduling execution module, the health status data, behavior status data, and psychological status data of the disabled elderly obtained in the above multi-modal perception and status analysis module are quantified. Then, the quantified values of the health status data, behavior status data, and psychological status data of the disabled elderly are weighted and summed to determine the comprehensive status score S of the disabled elderly at this time. Specifically, according to the formula: S = Qj×Shea + Qx×Sbeh + Qp×Spsy, where Shea represents the quantified value of the health status data of the disabled elderly, Sbeh represents the quantified value of the behavior status data of the disabled elderly, Spsy represents the quantified value of the psychological status data of the disabled elderly, Qj represents the weight of the quantified value of the health status data of the disabled elderly, Qx represents the weight of the quantified value of the behavior status data of the disabled elderly, Qp represents the weight of the quantified value of the psychological status data of the disabled elderly. Set the service threshold Ys, and compare the comprehensive status score S of the current disabled elderly with the service threshold Ys. When the comprehensive status score S of the disabled elderly is greater than or equal to the service threshold Ys, it indicates that the disabled elderly needs services at this time; when the comprehensive status score S of the disabled elderly is less than the service threshold Ys, it indicates that the disabled elderly does not need services at this time. After that, based on the obtained service-needed status data of the disabled elderly above, the task planning engine is called, a task queue is created in ROS, and a task pool is established, such as "remind to take medicine", "deliver water", "play news", "interactive chat", "check for falls", etc. Among them, each task includes information such as task type, priority, target time, target location, etc. Then, for each task, it is sorted according to the demand status, task priority, and timestamp of the current disabled elderly, and the tasks are screened using a scoring model to determine the tasks that need to be executed.

[0033] After selecting the task to be executed, use the ROS move_base navigation framework in ROS to determine the current coordinate data of the robot and the coordinate data of the task target point. Then, use the path break point compression and safety cost item algorithm to calculate the shortest path: Among them, Ppath represents the shortest path, C(stepi) represents the cost of the path break point, and n represents the number of break points in the path.

[0034] Furthermore, the linear velocity and angular velocity commands in the robot positioning data obtained in the above multi-modal perception and status analysis module are sent to the STM32 chassis controller through the serial port, and the control commands are calculated through the PID control algorithm to control the differential wheel chassis to move forward or turn: At the same time, during the navigation process, the dynamic window approach (DWA) algorithm is used to perform dynamic obstacle avoidance for the robot: DWA(v,w) = arg max(Safety(v, w), Efficiency(v,ω)) Among them, Safety(v, w) represents the motion safety of the robot, and Efficiency(v, ω) represents the motion efficiency of the robot. When a dynamic obstacle is detected, the DWA algorithm will adjust the speed and direction of the robot to avoid collisions.

[0035] After the robot reaches the task target point, it immediately switches to the service action execution process, such as delivering medicine, reminding to take medicine, delivering water, chatting, reporting weather, etc. After the task is completed, the robot uses the speech synthesis engine to prompt the disabled elderly that the task has been completed, for example: "The medicine has been delivered, please take it on time"; after the robot completes the task and makes a prompt, based on the speech recognition engine (ASR) + keyword triggering, the task result is processed to determine whether the disabled elderly has solved the problem, and this task is recorded.

[0036] In the task feedback closed-loop module, the task execution result information, including data such as task type, task start time, execution duration, task confirmation method, and final status mark, is written into the task log module. According to the task feedback status, the status scoring function of the disabled elderly is updated, and the current comprehensive status score of the current disabled elderly is recalculated; and the status of the current disabled elderly is synchronized to the task generation and scheduling execution module to trigger the next task cycle.

[0037] It should be noted that if multiple consecutive tasks fail or are difficult to confirm, the frequency of this task needs to be increased, or the interaction strategy needs to be adjusted. For example, if the "remind to take medicine" task fails 2 times in a row, increase the execution frequency or add voice / video confirmation methods. If a certain task fails more than twice in a row, create a "call confirmation" or "video call" type task in the task queue, and push the abnormal status in an alarm manner for manual intervention or video confirmation.

[0038] In the self-control module, based on the Markov decision process (MDP) theory, a behavior state model for accompanying the disabled elderly is established, and the state set and behavior set are defined. Among them, the state set includes combinations such as health indicators, emotional levels, and behavior activity levels, and the behavior set includes actions such as reminders, movements, delivering medicine, voice interactions, and remote alarms. Then, the reinforcement learning method is used to adjust the strategy through interaction feedback with the environment, select the optimal behavior in different states, and dynamically update the strategy function using the Q-learning algorithm through each behavior feedback, so as to optimize the accompanying strategy.

[0039] After that, the behavior tree under ROS is adopted as the behavior scheduling framework. Each escort task is split into several action nodes and conditional judgment nodes, and the priority and execution order are used to control the execution of tasks. For example, a health emergency may take precedence over other tasks. Among them, each node of the behavior tree listens to the status input of the disabled elderly in real time, and the status change will trigger a re-evaluation of the behavior tree to decide whether to execute a certain task or adjust the order of tasks, supporting interruption, jump, retry, and rollback mechanisms.

[0040] The self-control module also scores each escort task according to its urgency and importance through a task priority scoring model, establishes a task queue-jumping mechanism, and dynamically adjusts the task queue according to the current task status and priority, combined with the immediate status of the disabled elderly. And re-evaluate and sort the task queue in each control cycle to cope with complex and changing escort needs.

[0041] The above formulas are all dimensionless and take their numerical values for calculation. The formula is a formula obtained by collecting a large amount of data for software simulation to get the closest real situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0042] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0043] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application of the technical solution and the invention constraints. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0044] In addition, the functional modules in each embodiment of this application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0045] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0046] Finally, the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. The whole-process escort robot for the disabled elderly and its automatic control system, characterized in that, Including: A multi-modal perception and state analysis module, a task generation and scheduling execution module, a task feedback closed-loop module, and an automatic control module, with signal connections between the modules; The multi-modal perception and state analysis module is mainly used to divide the escort area into multiple sub-areas and configure different positioning schemes according to the characteristics of different sub-areas; The task generation and scheduling execution module is mainly used to judge the service status of the disabled elderly, dynamically screen and schedule the optimal robot resources to efficiently execute service tasks; The task feedback closed-loop module is mainly used to monitor the task execution effect, quickly review and dynamically update the status of the disabled elderly based on the feedback data; The automatic control module is mainly used to dynamically optimize the escort strategy, periodically re-evaluate and sort the task queue, and control the execution of tasks.

2. The disabled elderly full-course escort robot and its automatic control system according to claim 1, characterized in that: The multi-modal perception and state analysis module collects distance and angle data in the escort area, generates an environmental map of the escort area, and determines the real-time positioning of the robot.

3. The disabled elderly full-course escort robot and its automatic control system according to claim 2, characterized in that: The multi-modal perception and state analysis module unevenly divides the entire escort area into multiple sub-areas, combines the electromagnetic interference data and furniture material property data in the sub-areas to determine the influence coefficient of the disabled elderly's positioning in the sub-area; sets the influence threshold YL of the AMCL algorithm. When the influence coefficient L of the disabled elderly's positioning in the sub-area is less than or equal to the influence threshold YL of the AMCL algorithm, it indicates that the sub-area belongs to the low L-value area. When the influence coefficient L of the disabled elderly's positioning in the sub-area is greater than the influence threshold YL of the AMCL algorithm, it indicates that the sub-area belongs to the high L-value area, and selects a targeted positioning scheme for the disabled elderly according to the different L-values of the sub-areas.

4. The disabled elderly full-course escort robot and its automatic control system according to claim 3, characterized in that; In the multi-modal perception and state analysis module for the overlapping buffer area, calculate the spectral fluctuation value and the posture feature change rate of the bioelectrical signal data of the disabled elderly within the unit time T1, dynamically allocate the influence weights of the bioelectrical features and posture features based on the proportional relationship of the change amounts, determine the dominant feature of the disabled elderly in the overlapping buffer area, and locate the disabled elderly based on the dominant feature.

5. The incapacitated elderly full-course escort robot and its automatic control system according to claim 4, characterized in that: In the multi-modal perception and state analysis module, obtain the three-dimensional image data of the disabled elderly, calculate the height difference between the head and the buttocks of the disabled elderly; collect the physiological parameters of the disabled elderly, calculate the mean and standard deviation of each collected physiological parameter; obtain the emotional features and voice data on the face of the disabled elderly.

6. The disabled elderly full-course escort robot and its automatic control system according to claim 5, characterized in that: In the task generation and scheduling execution module, judge the service status of the disabled elderly, create a task queue, establish a task pool, and screen the tasks that need to be executed by the robot; determine the current coordinate data of the robot and the coordinate data of the task target point, and calculate the shortest path and control instructions.

7. The disabled elderly full-course escort robot and its automatic control system according to claim 6, characterized in that: In the task feedback closed-loop module, write the task data into the task log and recalculate the current comprehensive status score of the current disabled elderly.

8. The disabled elderly full-course escort robot and its automatic control system according to claim 7, characterized in that; In the automatic control module, establish a model of the escort behavior state of the disabled elderly, optimize the escort strategy through interaction feedback with the environment; according to the status and priority of the current task, combined with the immediate status of the disabled elderly, re-evaluate and sort the task queue in each cycle.

9. The incapacitated elderly full-course escort robot and its automatic control system according to claim 8, characterized in that: The method for obtaining electromagnetic interference data is as follows: Fixed-point collect the electromagnetic interference data in each sub-area through the electromagnetic interference sensor carried by the robot.

10. The disabled elderly full-course escort robot and its automatic control system according to claim 9, characterized in that: The method for obtaining furniture material property data is as follows: Scan the main furniture in each sub-region, count the proportion of metal materials of the furniture in the sub-region, and calculate the furniture material characteristic data in the sub-region.

Images