Electric wheelchair leg support stretching control method and device

Abstract

The invention provides an electric wheelchair leg support stretching control method and device, and relates to the technical field of electric wheelchair control, and the technical scheme is characterized in that user weight distribution, wheelchair posture, motion state and environmental terrain information are acquired; based on the obtained information, the current wheelchair state and the user intention are recognized, and a corresponding control mode is selected; according to the selected control mode, calling a corresponding state space model considering the position and speed of the leg support, the weight distribution of the user and the energy state of the system, and calculating the optimal position and stretching speed of the leg support in real time; and controlling the leg support to execute the stretching action according to the calculated optimal leg support position and the stretching speed. The electric wheelchair leg support stretching control method and device have the advantages that the control mode can be adjusted in a self-adaptive mode according to complex terrains and user requirements, the position and stretching speed of the leg support are accurately controlled in real time, and the use comfort and safety are improved.

Description

Technical Field

[0001] The present application relates to the technical field of electric wheelchair control, and more particularly, to a method and device for controlling the extension of the leg rest of an electric wheelchair. Background Art

[0002] An electric wheelchair is an important auxiliary device that provides mobility convenience for people with limited mobility. In actual use, users often need to move on various complex terrains (such as uneven ground, uphill and downhill, etc.), and at the same time adjust the position and angle of the leg rest according to different body position requirements. These complex usage scenarios pose higher requirements on the leg rest control system of the electric wheelchair.

[0003] However, the existing electric wheelchair leg rest control systems often fail to meet the requirements of these complex scenarios. The traditional control systems lack the ability to adapt to complex terrains and cannot automatically adjust the leg rest position according to different terrain features. This results in the need for users to frequently manually adjust the leg rest when moving between different terrains, greatly reducing the comfort and safety of use.

[0004] The existing systems have deficiencies in terms of real-time performance and are difficult to quickly respond to the user's body position change requirements. When the user needs to quickly adjust the posture or respond to emergencies, the reaction speed of the system may not meet the requirements, affecting the user experience and even potentially causing safety hazards.

[0005] In addition, the existing systems also have problems in terms of energy efficiency and are difficult to achieve long-term battery life while ensuring functions. The battery capacity of the electric wheelchair is limited, and the frequent adjustment of the leg rest consumes a large amount of energy. How to optimize energy use while ensuring functions is another challenge faced by the existing systems.

[0006] Finally, the safety and reliability of the existing systems still need to be improved, especially the ability to handle emergencies in complex environments is insufficient. For example, when going uphill or downhill or on uneven ground, how to ensure the stability of the leg rest and the safety of the user is a problem that the existing systems are difficult to fully solve.

[0007] In view of the above problems, the existing technology urgently needs to be improved. Summary of the Invention

[0008] The purpose of the present application is to provide a method and device for controlling the extension of the leg rest of an electric wheelchair, which has the advantages of being able to adaptively adjust the control mode according to complex terrains and user needs, realizing real-time and precise control of the leg rest position and extension speed, and improving the comfort and safety of use.

[0009] The present application provides a method for controlling the extension of the leg rest of an electric wheelchair, and the technical solution is as follows: The method includes: when it is detected that the terrain complexity where the electric wheelchair is located exceeds a set value or the user needs to adjust the body position, acquiring user weight distribution, wheelchair attitude, motion state, and environmental terrain information; based on the acquired information, identifying the current wheelchair state and user intention, and selecting a corresponding control mode from multiple preset control modes according to the identification result; according to the selected control mode, calling a corresponding state space model that considers leg rest position, speed, user weight distribution, and system energy state, and calculating the optimal leg rest position and extension speed in real time; controlling the leg rest to perform an extension action according to the calculated optimal leg rest position and extension speed.

[0010] Further, the present application also proposes that the step of, based on the acquired information, identifying the current wheelchair state and user intention, and selecting a corresponding control mode from multiple preset control modes according to the identification result includes: performing feature extraction and classification on the acquired user weight distribution, wheelchair attitude, motion state, and environmental terrain information; based on the classification result, calculating the application probability of each preset control mode, and selecting the control mode with the highest application probability as the corresponding control mode, where the classification result is used to represent the current wheelchair state and user intention; it also includes: according to the selected corresponding control mode and the current wheelchair state, calculating the transition control parameters when switching the control mode, so as to achieve a smooth transition between modes during mode switching.

[0011] Further, the present application also proposes that the step of, according to the selected control mode, calling a corresponding state space model that considers leg rest position, speed, user weight distribution, and system energy state, and calculating the optimal leg rest position and extension speed in real time includes: according to the selected control mode, calling a corresponding state space model that considers leg rest position, speed, user weight distribution, and system energy state; acquiring the current system computing resource state and the remaining battery power; based on the acquired system computing resource state and the remaining battery power, adjusting the complexity and calculation frequency of the state space model, where when the system computing resources are sufficient and the remaining battery power is higher than a first preset threshold, increasing the model complexity and calculation frequency, otherwise decreasing the model complexity and calculation frequency; using the adjusted state space model, combined with the leg rest position, speed, user weight distribution, and system energy state acquired in real time, to calculate the optimal leg rest position and extension speed.

[0012] Further, the present application also proposes that the step of calculating the applicable probability of each preset control mode based on the classification result and selecting the control mode with the highest applicable probability as the corresponding control mode includes: when it is detected that the difference in the applicable probabilities of multiple control modes is less than a second preset threshold, obtaining an evaluation matrix including historical control mode selection records, corresponding control effect scores, control mode switching frequencies, and user feedback; based on the historical data in the evaluation matrix, correcting the applicable probabilities of each control mode; and selecting the control mode with the highest applicable probability after correction as the corresponding control mode.

[0013] Further, the present application also proposes that the step of adjusting the complexity and calculation frequency of the state space model based on the obtained system computing resource status and remaining battery power includes: obtaining the model complexity and calculation frequency in the current control mode; determining the target model complexity and target calculation frequency according to the obtained system computing resource status and remaining battery power; generating a progressive adjustment sequence of the model complexity and calculation frequency based on the difference between the current value and the target value in combination with a preset smoothing factor; gradually adjusting the state space model according to the generated adjustment sequence, and at the same time monitoring the control effect, and when it is detected that the control accuracy drops by more than a third preset threshold, pausing the adjustment and reverting to the previous stable state.

[0014] Further, the present application also proposes that the step of calling the corresponding state space model considering the leg rest position, speed, user weight distribution, and system energy state according to the selected control mode includes: obtaining the type of the currently selected control mode, where the control mode includes a flat ground mode, an uphill and downhill mode, and a standing mode; when the control mode is the flat ground mode, calling a first state space model including the leg rest position, speed, user weight distribution, and system energy state; when the control mode is the uphill and downhill mode, calling a second state space model with the wheelchair inclination angle and terrain slope added on the basis of the first state space model; when the control mode is the standing mode, calling a third state space model with the user's standing intention confidence level and leg rest support force added on the basis of the second state space model.

[0015] Further, the present application also proposes that the step of correcting the applicable probabilities of each control mode based on the historical data in the evaluation matrix includes: obtaining the historical data in the evaluation matrix, including historical control mode selection records, corresponding control effect scores, control mode switching frequencies, and user feedback; performing a timeliness evaluation on the historical data and removing the data that exceeds a preset time threshold; performing outlier detection on the remaining data to identify and mark the abnormal data; calculating the historical performance scores of each control mode based on the valid historical data after removing the abnormal data; and performing weighted fusion of the historical performance scores and the currently calculated applicable probabilities to obtain the corrected applicable probabilities of each control mode.

[0016] Further, the present application also proposes that when the control mode is the standing mode, the steps of calling the third state space model with the user's standing intention confidence and the leg rest support force added on the basis of the second state space model include: obtaining the user's standing intention confidence and the leg rest support force; dynamically adjusting the parameter weights in the second state space model according to the user's standing intention confidence and the leg rest support force; applying the adjusted parameter weights to the second state space model to form a third state space model adapted to the standing mode; using the third state space model to calculate the optimal position and extension speed of the leg rest to support the user's transition from the sitting position to the standing position.

[0017] Furthermore, the present application also proposes that the step of calculating the optimal position and extension speed of the leg rest using the third state space model to support the user's transition from a sitting position to a standing position includes: constructing a state vector x(t), including the leg rest position p(t), the leg rest speed v(t), the user's weight distribution m(t), the system energy state E(t), the wheelchair inclination angle α(t), the terrain slope β(t), and the leg rest support force F(t); based on the current control mode θ(t), constructing a state space model: x(t + 1) = A(θ(t), t)x(t) + B(θ(t), t)u(t) + D(θ(t), t)d(t) + w(t), y(t) = C(θ(t), t)x(t) + v_noise(t), where x(t + 1) is the state vector at time step t + 1, y(t) is the output vector at time step t, A(θ(t), t), B(θ(t), t), C(θ(t), t), and D(θ(t), t) are time-varying matrices dependent on the control mode, u(t) is the control input vector, d(t) is the external disturbance vector, w(t) is the process noise, and v_noise(t) is the observation noise; calculating the user's standing intention confidence CI(t); selecting the optimal control mode θ(t) according to the current state x(t) and the user's standing intention confidence CI(t); constructing an objective function: J = Σ(w1(p(t) - p_desired(t))^2 + w2(v(t) - v_desired(t))^2 + w3(F(t) - F_desired(t))^2 + λu(t)^2), where w1, w2, and w3 are weight coefficients, λ is the control cost coefficient, p_desired(t) is the desired leg rest position, v_desired(t) is the desired leg rest speed, and F_desired(t) is the desired leg rest support force; solving the optimization problem of minimizing the objective function J under the premise of satisfying the constraint conditions p_min ≤ p(t) ≤ p_max, v_min ≤ v(t) ≤ v_max, F_min ≤ F(t) ≤ F_max, E(t) ≥ E_min, |α(t)| ≤ α_max, |β(t)| ≤ β_max, to obtain the optimal control sequence u*(t), where p_min and p_max are the minimum and maximum values of the leg rest position respectively, v_min and v_max are the minimum and maximum values of the leg rest speed respectively, F_min and F_max are the minimum and maximum values of the leg rest support force respectively, E_min is the minimum value of the system energy state, α_max is the maximum absolute value of the wheelchair inclination angle, and β_max is the maximum absolute value of the terrain slope; calculating the optimal position and extension speed of the leg rest according to the optimal control sequence u*(t).

[0018] Furthermore, the present application also proposes an electric wheelchair legrest extension control device, which includes: an acquisition module for acquiring user weight distribution, wheelchair attitude, motion state, and environmental terrain information when it is detected that the terrain complexity where the electric wheelchair is located exceeds a set value or the user needs to adjust the body position; an identification module for identifying the current wheelchair state and user intention based on the acquired information, and selecting a corresponding control mode from a variety of preset control modes according to the identification result; a calculation module for calling a corresponding state space model considering legrest position, speed, user weight distribution, and system energy state according to the selected control mode, and calculating the optimal legrest position and extension speed in real time; and a control module for controlling the legrest to perform an extension action according to the calculated optimal legrest position and extension speed.

[0019] As can be seen from the above, an electric wheelchair legrest extension control method and device provided by the present application solve the adaptability and real-time problems of traditional control systems under complex terrains and diverse body position requirements by acquiring and analyzing user and environmental information in real time, adaptively selecting the most suitable control mode, and precisely calculating the legrest position and speed using a state space model. It has the advantages of being able to adaptively adjust the control mode according to complex terrains and user needs, realizing real-time and precise control of the legrest position and extension speed, and improving the comfort and safety of use. Description of the Drawings

[0020] Figure 1 It is a flowchart of an electric wheelchair legrest extension control method provided by the present application.

[0021] Figure 2 It is a schematic structural diagram of an electric wheelchair legrest extension control device provided by the present application.

[0022] In the figure: 210, acquisition module; 220, identification module; 230, calculation module; 240, control module. Detailed Embodiments

[0023] Next, the technical solutions in the present application will be clearly and completely described in conjunction with the accompanying drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the present application usually described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application claimed, but only represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of the present application.

[0024] The real-time precise control of the legrest of an electric wheelchair faces severe challenges under complex terrains and diverse body position requirements. Existing control systems are difficult to adapt to different terrain features and cannot automatically adjust the legrest position according to environmental changes. In addition, the system has deficiencies in real-time performance and is difficult to quickly respond to the changing needs of the user's body position. The energy efficiency issue also restricts the long-term battery life of the system. Most critically, in complex environments, the safety and reliability of existing systems cannot be guaranteed, especially when dealing with emergencies. These problems combined result in the electric wheelchair being unable to provide users with a comfortable, safe, and efficient usage experience.

[0025] Specifically, wheelchairs usually need to switch between various terrains such as flat roads, slopes, steps, and uneven roads. Among them, the terrain complexity can be quantified by parameters such as slope angle and road surface roughness. For example, when the wheelchair enters an uphill section (slope 15°) from flat ground (slope 0°), the legrest position needs to be quickly adjusted to maintain user comfort. However, existing systems cannot accurately estimate the change in the user's weight distribution, resulting in improper adjustment of the legrest and increasing the risk of the user slipping. In addition, frequent adjustment of the legrest will significantly increase energy consumption, possibly reducing the battery life from the original 8 hours to 6 hours.

[0026] In response to this, referring to Figure 1 , this application proposes a method for controlling the extension of the legrest of an electric wheelchair, which includes: S110. When it is detected that the terrain complexity of the electric wheelchair exceeds the set value or the user needs to adjust the body position, obtain the user's weight distribution, wheelchair attitude, motion state, and environmental terrain information; S120. Based on the obtained information, identify the current wheelchair state and user intention, and select the corresponding control mode from multiple preset control modes according to the identification result; S130. According to the selected control mode, call the corresponding state space model considering the legrest position, speed, user weight distribution, and system energy state, and calculate the optimal legrest position and extension speed in real time; S140. Control the legrest to perform the extension action according to the calculated optimal legrest position and extension speed.

[0027] Among them, the terrain complexity refers to the complexity of the terrain features of the environment where the electric wheelchair is located, and can specifically be quantified by parameters such as slope angle and road surface roughness, and is used to judge whether to trigger the legrest control process.

[0028] Among them, the user's weight distribution refers to the weight distribution of the user on the seat of the electric wheelchair, and can specifically be measured by a pressure sensor array, and is used to calculate the optimal legrest position and extension speed.

[0029] Among them, the wheelchair attitude refers to the tilt angle and direction of the electric wheelchair in space, and can be specifically measured by a gyroscope and an accelerometer, which is used to judge the current state of the wheelchair.

[0030] Among them, the motion state refers to dynamic parameters such as the speed and acceleration of the electric wheelchair, and can be specifically obtained by an encoder and an inertial measurement unit, which is used to predict the future state of the wheelchair.

[0031] Among them, the environmental terrain information refers to the terrain features around the wheelchair, and can be specifically obtained by a lidar or a depth camera, which is used to predict the upcoming terrain changes.

[0032] Among them, the control mode refers to the preset leg rest control strategies for different situations, and can be specifically implemented by various algorithms such as fuzzy control and adaptive control, which is used to adapt to different usage scenarios.

[0033] Among them, the state space model is a mathematical model that describes the dynamic characteristics of a system, and can be specifically represented by a matrix equation to show the relationship between the system state, input, and output, which is used to calculate the optimal control parameters in real time.

[0034] The core innovation of this application lies in proposing to achieve the optimal control of the leg rest position and extension speed by obtaining comprehensive environmental and user information in real time, combining multiple preset control modes and state space models. This method significantly improves the adaptability, accuracy, and safety of the electric wheelchair in complex environments.

[0035] The working principle of this application can be divided into the following key steps: First, the system monitors the operating environment and user state of the electric wheelchair in real time through a sensor network. This includes using a pressure sensor array to measure the user's weight distribution, a gyroscope and an accelerometer to measure the wheelchair attitude, an encoder and an inertial measurement unit to obtain the motion state, and a lidar or a depth camera to collect environmental terrain information.

[0036] Second, the system performs state recognition and control mode selection based on the obtained information. Through data fusion and pattern recognition algorithms, the system can accurately judge the current wheelchair state and user intention. Then, the system selects the most suitable mode from multiple preset control modes. These control modes can include flat ground mode, uphill and downhill mode, etc., and each mode has its specific control strategy.

[0037] Next, the system calls the corresponding state space model according to the selected control mode. These models consider multiple factors such as the leg rest position, speed, user weight distribution, and system energy state. By solving the state space equation, the system can calculate the optimal leg rest position and extension speed in real time.

[0038] Finally, the system converts the calculation result into specific control instructions to drive the leg rest to perform corresponding stretching actions. This process is completed through a precise motor control system to ensure that the leg rest can accurately and smoothly reach the calculated optimal position.

[0039] As a preferred implementation manner, the leg rest stretching control method of the present application can be specifically implemented as follows: First, the electric wheelchair is equipped with a high-precision sensor system, including a pressure sensor array distributed on the seat surface for measuring the user's weight distribution; an inertial measurement unit (IMU) for obtaining the wheelchair attitude and motion state; and a radar for scanning the surrounding environmental terrain.

[0040] When the system detects that the terrain complexity exceeds a set value (for example, the slope is greater than 10° or the road surface roughness index exceeds 0.5) or the user issues an instruction to adjust the body position through the control panel, the leg rest control process is triggered.

[0041] The system immediately starts to obtain various data, including the user weight distribution matrix M, the wheelchair attitude vector A (including pitch angle, roll angle, and yaw angle), the motion state vector V (including velocities in the x, y, and z directions), and the environmental terrain point cloud data P(n×3).

[0042] Next, the system uses a pre-trained deep neural network model to process the obtained data to identify the current wheelchair state and user intention. The input is the above-obtained various data, and the output is the probability distribution of the wheelchair state and user intention. Based on the recognition result, the system selects the mode with the highest applicable probability from the preset control mode library. The control mode library can include but is not limited to flat ground mode, uphill and downhill mode, turning mode, standing mode, etc.

[0043] The system calls the corresponding state space model according to the selected control mode. For example, a certain control mode state space model can be expressed as: x(k + 1)=Ax(k)+Bu(k)+w(k); y(k)=Cx(k)+v(k); Among them, x(k) is the state vector, including the leg rest position, speed, user weight distribution center, and system energy state; u(k) is the control input; y(k) is the observation output; A, B, and C are system matrices; w(k) and v(k) are process noise and observation noise respectively.

[0044] The system uses the Kalman filter to estimate the state and calculates the optimal control input u*(k) by solving the linear quadratic regulator (LQR) problem. The optimization objective function is: J = Σ(x'(k)Qx(k)+u'(k)Ru(k)); Among them, Q and R are weight matrices used to balance control precision and energy consumption. x'(k) represents the transpose of the state vector x(k) at time step k, and u'(k) represents the transpose of the control input vector u(k) at time step k.

[0045] Finally, the system converts the calculated optimal control input u*(k) into specific leg rest position and extension speed instructions, and controls the DC servo motor of the leg rest actuator through PWM signals to achieve the precise extension action of the leg rest.

[0046] Through this implementation manner, the leg rest extension control method of the present application can achieve precise and real-time control under complex terrains and diverse body position requirements, significantly improving the comfort and safety of users.

[0047] In some of the above embodiments of the present application, steps are proposed to identify the current wheelchair state and user intention based on the acquired information, and select the corresponding control mode from multiple preset control modes according to the recognition result to adapt to complex environments and user needs. However, in this process, there are still challenges in accurately identifying the wheelchair state and user intention, and in reasonably selecting and smoothly switching between multiple control modes. Especially when facing complex and changeable environments and user needs, simple control mode selection may not meet the actual requirements, and unstable or discontinuous problems may occur during mode switching.

[0048] In response to this, the present application further proposes to perform feature extraction and classification on the acquired user weight distribution, wheelchair attitude, motion state, and environmental terrain information; based on the classification results, calculate the applicable probabilities of each preset control mode, and select the control mode with the highest applicable probability as the corresponding control mode, where the classification results are used to characterize the current wheelchair state and user intention; it also includes: calculating the transition control parameters during control mode switching according to the selected corresponding control mode and the current wheelchair state, so as to achieve a smooth transition between modes during mode switching.

[0049] The technical solution of the present application improves the recognition accuracy of the wheelchair state and user intention by performing feature extraction and classification on the acquired information. By calculating the applicable probabilities of each preset control mode and selecting the mode with the highest probability, a more intelligent and precise control mode selection is achieved. The introduction of the calculation of transition control parameters solves the possible unstable or discontinuous problems during mode switching, ensuring the smoothness of the control process.

[0050] Specifically, by extracting meaningful features and classifying them, the system can understand the current wheelchair state and the user's intention. Then, based on these recognition results, the system can determine which preset control mode is most suitable for the current situation, calculate the applicable probabilities of each control mode, and finally select the most appropriate control mode to control the extension action of the leg rest.

[0051] This method realizes the selection of control modes in a data-driven manner, improving the system's adaptability to complex environments and diverse user needs. Especially when facing rapidly changing terrains or user states, this method can adjust control strategies in a timely manner, improving the comfort and safety of using electric wheelchairs. At the same time, the introduction of transition control parameters ensures smooth switching between different control modes, avoiding the discomfort that sudden action changes may bring to users. This intelligent control method not only improves the overall performance of the system but also enhances the user's confidence and satisfaction in using electric wheelchairs.

[0052] In the technical solution of this application, the feature extraction and classification steps can be implemented in various ways. For example, machine learning algorithms such as support vector machines or random forests can be used to extract features and classify the acquired information. Another method is to use deep learning models, such as convolutional neural networks or recurrent neural networks, to capture complex spatio-temporal features.

[0053] When calculating the applicable probabilities of each preset control mode, Bayesian inference or fuzzy logic methods can be used. Bayesian inference can calculate the posterior probability of each control mode based on historical data and current observation results. Fuzzy logic can handle uncertainties and map the classification results to applicable probabilities.

[0054] The calculation of transition control parameters can adopt interpolation methods or optimal control theory. Linear interpolation can achieve a simple smooth transition between two control modes. Using optimal control theory, such as model predictive control, can generate an optimal transition trajectory considering system dynamics and constraints.

[0055] The results of feature extraction and classification directly affect the calculation accuracy of applicable probabilities. The calculation results of applicable probabilities in turn determine the selection of corresponding control modes. Finally, the selection of corresponding control modes and the current wheelchair state jointly determine the calculation of transition control parameters.

[0056] When the technical solution of this application solves the problem of controlling the leg rest of an electric wheelchair in a complex environment, it first improves the recognition accuracy of the wheelchair state and user intention through the feature extraction and classification steps. This step can effectively process complex data from multiple sensors, such as user weight distribution, wheelchair attitude, motion state, and environmental terrain information. Through the feature extraction algorithm, the most representative and discriminative features can be extracted from the original data, laying a foundation for the subsequent selection of control modes.

[0057] Based on these extracted features, the system calculates the applicable probabilities of each preset control mode. For example, on a flat road surface, the applicable probability of the flat ground mode may be relatively high; while when an uphill is detected, the applicable probability of the uphill and downhill mode will increase accordingly. This probabilistic selection method greatly improves the system's adaptability to complex environments.

[0058] After selecting the control mode with the highest applicable probability as the corresponding control mode, the system also needs to consider how to smoothly transition from the current mode to the target mode. This requires calculating the transition control parameters. The introduction of transition control parameters solves the possible instability or discontinuity problems during mode switching. For example, when switching from the flat ground mode to the uphill and downhill mode, the system will calculate a series of transition parameters according to the current wheelchair state (such as speed, inclination angle) and the requirements of the target mode, so that the position and speed of the leg rest can be adjusted smoothly, avoiding sudden changes that cause discomfort to the user.

[0059] Suppose the electric wheelchair is traveling on a flat ground and suddenly encounters a gentle slope. The system first obtains the wheelchair attitude information through the acceleration sensor and gyroscope, obtains the user weight distribution information through the pressure sensor, obtains the motion state information through the wheel speed encoder, and obtains the environmental terrain information through the front camera or lidar.

[0060] These information are input into a pre-trained convolutional neural network for feature extraction and classification. The network output may include classification results of terrain types (flat ground, gentle slope, steep slope, etc.), user postures (normal sitting posture, leaning forward, leaning backward, etc.), and motion states (stationary, uniform speed, accelerating, etc.).

[0061] Based on these classification results, the system uses Bayesian inference to calculate the applicable probabilities of each preset control mode (such as flat ground mode, uphill and downhill mode, etc.). In this example, since a gentle slope is detected, the applicable probability of the uphill and downhill mode may increase significantly, for example, from the original 0.1 to 0.7, while the applicable probability of the flat ground mode may decrease from 0.8 to 0.3.

[0062] The system selects the uphill / downhill mode with the highest applicable probability as the corresponding control mode. Next, the system needs to calculate the transition control parameters from the current flat-ground mode to the uphill / downhill mode. Here, a model predictive control algorithm can be used to calculate an optimal transition trajectory considering the current speed, inclination angle of the wheelchair, the current position of the leg rest, and other states, as well as the target states in the uphill / downhill mode.

[0063] For example, this transition trajectory includes smoothly adjusting the inclination angle of the wheelchair from 0 degrees to 5 degrees within 2 seconds, while adjusting the leg rest angle from 0 degrees to 15 degrees. At the same time, the dynamic model of the wheelchair, comfort constraints (such as the maximum angular acceleration not exceeding 5 degrees / second²) and safety constraints (such as the maximum inclination angle not exceeding 10 degrees) can also be considered to generate a series of control instructions to ensure a smooth, comfortable and safe transition process.

[0064] Through this method, the electric wheelchair can achieve intelligent control mode selection and smooth transition in a complex and changeable environment, greatly improving the comfort and safety of users. At the same time, since the system can adapt to environmental changes in a timely manner, the passability and maneuverability of the electric wheelchair are also improved, enabling users to cope with various terrain conditions more freely.

[0065] In some of the above embodiments of the present application, a state space model considering the leg rest position, speed, user weight distribution and system energy state is called according to the selected control mode to calculate the optimal leg rest position and extension speed in real time for precise control of the leg rest of the electric wheelchair. However, in this process, the dynamic changes in the system computing resources and battery power may affect the performance and efficiency of the control model. How to dynamically adjust the model complexity and calculation frequency according to the system resource state while ensuring the control accuracy to achieve more efficient energy utilization and more durable system operation has become an urgent problem to be solved.

[0066] In response to this, the present application further proposes to call a corresponding state space model considering the leg rest position, speed, user weight distribution and system energy state according to the selected control mode; obtain the current system computing resource state and the remaining battery power; based on the obtained system computing resource state and the remaining battery power, adjust the complexity and calculation frequency of the state space model, where when the system computing resources are sufficient and the remaining battery power is higher than the first preset threshold, increase the model complexity and calculation frequency, otherwise reduce the model complexity and calculation frequency; use the adjusted state space model to calculate the optimal leg rest position and extension speed in combination with the leg rest position, speed, user weight distribution and system energy state obtained in real time.

[0067] This solution addresses the problems of reduced control accuracy or excessive system consumption that may occur in fixed models under limited resources by dynamically balancing control accuracy and system resource consumption. It can extend the system operation time, improve energy utilization efficiency while ensuring control quality, thereby enhancing the adaptability and durability of the electric wheelchair in complex environments.

[0068] In the technical solution of this application, the complexity adjustment of the state - space model can be achieved in various ways. For example, the complexity can be changed by adjusting the number of state variables considered in the model. When system resources are sufficient, the model can include more state variables, such as leg rest position, speed, acceleration, user weight distribution, system energy state, environmental temperature, etc.; when resources are limited, the number of state variables considered can be reduced, only retaining the most critical ones. Another method is to adjust the order of the model. Higher - order models can describe system dynamics more precisely, but also have higher computational complexity.

[0069] The adjustment of the calculation frequency can be achieved by changing the time interval between model updates and control instruction transmissions. For example, when resources are sufficient, the calculation frequency can be increased to once every 50 milliseconds, while when resources are limited, it can be reduced to once every 200 milliseconds.

[0070] The system computing resource status can be evaluated by monitoring metrics such as CPU usage rate and memory occupancy. For example, it can be set that when the CPU usage rate is below 60% and the available memory is greater than 30%, it is considered that resources are sufficient. The remaining battery power can be obtained through the battery management system, and a threshold, such as 30%, can be set as the judgment criterion.

[0071] As a preferred implementation, the technical solution of this application can be realized through the following specific steps: First, when the system is initialized, set the thresholds for resource status evaluation. For example, define that when the CPU usage rate is below 60% and the available memory is greater than 30%, the computing resources are sufficient, and when the remaining battery power is higher than 30%, the battery power is sufficient.

[0072] Second, the system checks the system resource status every certain period (such as every 500 milliseconds). Specifically, obtain the CPU usage rate and available memory amount through the operating system API, and obtain the percentage of the remaining battery power through the battery management system interface.

[0073] Then, based on the obtained resource status, the system decides whether to adjust the model complexity and calculation frequency. If the resource status changes significantly (such as from sufficient to insufficient, or vice versa), the adjustment process is triggered.

[0074] During the adjustment process, the system first determines the target model complexity and calculation frequency. For example, when resources are sufficient, a high-order model with 10 state variables can be selected, and the calculation frequency can be set to once every 50 milliseconds; when resources are insufficient, a low-order model with 5 key state variables can be selected, and the calculation frequency can be reduced to once every 200 milliseconds.

[0075] To avoid the impact of sudden changes on system stability, the adjustment process adopts a progressive method. For example, if it is necessary to reduce from 10 state variables to 5, the system may complete the adjustment within 5 calculation cycles at a rate of reducing 1 state variable each time. Similarly, the adjustment of the calculation frequency can also be carried out step by step.

[0076] During the adjustment process, the system continuously monitors the control effect. If it is found that the control accuracy drops by more than a set value (such as the position error increases by more than 10%), the system will pause the adjustment and roll back to the previous stable state.

[0077] Finally, calculations are performed using the adjusted model. The system inputs the real-time obtained leg rest position, speed, user weight distribution, and system energy state into the adjusted state space model, calculates the optimal leg rest position and extension speed, and sends these control instructions to the actuator.

[0078] Through this specific implementation manner, the technical solution of the present application can achieve a dynamic balance between system resources and control accuracy in practical applications, improving the overall performance and reliability of the electric wheelchair leg rest control system.

[0079] In some of the above embodiments of the present application, it is proposed to calculate the applicable probabilities of each preset control mode based on the classification results, and select the control mode with the highest applicable probability as the corresponding control mode to select the most suitable control mode for the current situation. However, in this process, when the applicable probabilities of multiple control modes are very close, relying only on the currently calculated applicable probabilities may not be able to make an optimal choice. This may lead to instability in the control mode selection, frequent switching of control modes, or selection of a sub-optimal control mode, thus affecting the control effect of the electric wheelchair leg rest extension and the user experience.

[0080] In response to this, the present application further proposes that when it is detected that the difference in the applicable probabilities of multiple control modes is less than the second preset threshold, an evaluation matrix including historical control mode selection records, corresponding control effect scores, control mode switching frequencies, and user feedback is obtained; based on the historical data in the evaluation matrix, the applicable probabilities of each control mode are corrected; and the control mode with the highest corrected applicable probability is selected as the corresponding control mode.

[0081] The technical solution proposed in this application optimizes the selection process of control modes by introducing historical data and user feedback, effectively solving the problems of unstable and sub - optimal selection of control modes that may be caused by relying solely on current state information. This solution considers multi - dimensional historical information, including control mode selection records, control effect scores, control mode switching frequencies, and user feedback, providing a more comprehensive reference basis for the selection of control modes.

[0082] When obtaining information such as the user's weight distribution, wheelchair posture, motion state, and environmental terrain, these information can be combined into a "context vector".

[0083] Establish a historical context database to store historical data and their corresponding context vectors.

[0084] When selecting a control mode, calculate the similarity between the current context vector and each context vector in the historical context database. Methods such as cosine similarity and Euclidean distance can be used.

[0085] Set a similarity threshold. Only when the similarity between the historical context and the current context is higher than this threshold, the historical data is considered to have reference value.

[0086] Use the similarity as a weight to weight the historical data. The higher the similarity, the greater the weight.

[0087] Specifically, when it is detected that the difference in the applicable probabilities of multiple control modes is less than the second preset threshold, the system will trigger the acquisition and correction mechanism of the evaluation matrix. This indicates that the decision - making confidence provided by the current state information is low, and historical data needs to be used for auxiliary judgment. The system first constructs the current context vector and calculates the similarity (such as cosine similarity) with the context vectors in the historical context database. Only historical data with high similarity will be included in the correction process and weighted according to the similarity. The correction process adopts a dynamic weight adjustment strategy, and the formula is as follows: Corrected probability = α1 * current applicable probability+(1 - α1)* similarity - weighted historical performance score; Among them, α1 is a weight dynamically adjusted based on the confidence of the current applicable probability. The higher the confidence, the greater α1; the similarity - weighted historical performance score is the result of weighting the historical performance score according to the context similarity. The calculation of the historical performance score considers the control effect score, user feedback satisfaction, and control mode stability (related to the switching frequency), and classifies and processes abnormal data (eliminating systematic errors and smoothing accidental errors).

[0088] In this way, based on the current state information, the selection of the control mode can be adjusted by combining highly relevant historical performance and user experience, improving the stability and accuracy of the control mode selection, reducing unnecessary mode switches, and better adapting to the personalized needs of users and environmental changes.

[0089] In some of the above embodiments of the present application, it is proposed to adjust the complexity and calculation frequency of the state space model based on the obtained system computing resource status and remaining battery power to achieve real-time precise control of the leg rest of an electric wheelchair. However, in this process, the existing system has deficiencies in resource management and computing efficiency, resulting in an inability to maintain a stable control effect in complex environments.

[0090] In response to this, the present application further proposes to obtain the model complexity and calculation frequency in the current control mode; determine the target model complexity and target calculation frequency according to the obtained system computing resource status and remaining battery power; generate a progressive adjustment sequence of the model complexity and calculation frequency based on the difference between the current value and the target value, in combination with a preset smoothing factor; gradually adjust the state space model according to the generated adjustment sequence, while monitoring the control effect, and when it is detected that the control accuracy drops by more than a third preset threshold, pause the adjustment and roll back to the previous stable state.

[0091] By dynamically adjusting the complexity and calculation frequency of the state space model, the present application can effectively address the real-time control problem of an electric wheelchair under complex terrains and diverse body position requirements. This solution ensures that when resources are sufficient, the computing power of the model is enhanced to improve control accuracy by obtaining the system resource status and battery power; while when resources are insufficient, the model complexity is reduced to maintain the stability and safety of the system. In this way, the technical solution of the present application can improve the adaptability and energy efficiency of the electric wheelchair in complex environments while ensuring its functions.

[0092] The technical solution of the present application further enhances the stability and reliability of the system by introducing a progressive adjustment and control effect monitoring mechanism. Specifically, the present application first obtains the model complexity and calculation frequency in the current control mode, which provides a benchmark for subsequent adjustments. Then, according to the system computing resource status and remaining battery power, the target model complexity and target calculation frequency are determined. This step ensures that the system can make appropriate adjustment decisions based on the currently available resources.

[0093] Furthermore, the present application generates a progressive adjustment sequence of the model complexity and calculation frequency based on the difference between the current value and the target value, in combination with a preset smoothing factor. This progressive adjustment method can avoid sudden large changes, thereby reducing the impact on system stability. The introduction of the smoothing factor further enhances the controllability of the adjustment process and can be flexibly set according to the requirements of different scenarios.

[0094] During the actual adjustment process, the present application gradually adjusts the state space model according to the generated adjustment sequence while monitoring the control effect. This real-time monitoring mechanism ensures that the system can promptly detect potential problems. When it is detected that the control accuracy drops by more than the third preset threshold, the system will suspend the adjustment and roll back to the previous stable state. This feature greatly improves the robustness of the system and prevents performance degradation caused by over-adjustment.

[0095] By combining the state space model and the multi-mode control strategy in the previous embodiments, the dynamic adjustment mechanism of the present application can more precisely adapt to different control scenarios. For example, in the uphill and downhill modes, the system may require a higher model complexity to process complex terrain information, and the solution of the present application can increase the model complexity when resources permit, thereby improving the control accuracy.

[0096] As a preferred implementation manner, the present application can set multiple predetermined complexity levels and calculation frequency levels. For example, three complexity levels of low, medium, and high can be set, corresponding to different numbers of state variables and model parameters respectively. Similarly, the calculation frequency can also be set to three levels of low (such as 1Hz), medium (such as 5Hz), and high (such as 10Hz). The system can switch between these preset levels according to the current resource state, thereby achieving more efficient resource management.

[0097] Specifically, the implementation process of the present application can be as follows: First, the system obtains the current model complexity (for example, medium complexity, including 5 state variables) and calculation frequency (for example, 5Hz). Then, the system detects that the remaining battery power drops below 30%, and at the same time, the system calculation resource utilization rate exceeds 80%. Based on this information, the system determines that it is necessary to reduce the model complexity and calculation frequency, and sets the target to low complexity (3 state variables) and low-frequency calculation (1Hz).

[0098] Next, the system generates a progressive adjustment sequence. Assuming that the smoothing factor is set to 0.2, then the system will complete the adjustment in 5 steps. In each step, the model complexity will be reduced by 0.4 state variables (rounded to the nearest integer), and the calculation frequency will be reduced by 0.8Hz. The system starts to execute this adjustment sequence and monitors the control effect every time a step is executed.

[0099] Assume that when executing to the third step, the system detects that the control accuracy drops by 15%, exceeding the preset 10% threshold. At this time, the system will immediately suspend the adjustment and roll back to the state of the second step. Finally, the system successfully reduces the model complexity to 4 state variables and the calculation frequency to 3Hz, effectively reducing resource consumption while ensuring the control effect.

[0100] In this way, the technical solution of the present application can maintain the stability and accuracy of the leg rest control of the electric wheelchair in a complex and changeable environment, while optimizing the usage efficiency of system resources. This not only improves the adaptability of the electric wheelchair to various terrains and user requirements, but also extends the battery usage time and enhances the reliability and user experience of the overall system.

[0101] In some of the above embodiments of the present application, according to the selected control mode, a corresponding state space model considering the leg rest position, speed, user weight distribution, and system energy state is called to calculate the optimal leg rest position and extension speed in real time. However, in this process, different control modes may need to consider different parameters and state variables. Simply using a unified state space model may not fully meet the special requirements of various control modes, thus affecting the accuracy and effect of control.

[0102] In response to this, the present application further proposes that the steps of calling the corresponding state space model considering the leg rest position, speed, user weight distribution, and system energy state according to the selected control mode include: obtaining the type of the currently selected control mode, where the control mode includes a flat ground mode, an uphill and downhill mode, and a standing mode; when the control mode is the flat ground mode, calling a first state space model including the leg rest position, speed, user weight distribution, and system energy state; when the control mode is the uphill and downhill mode, calling a second state space model with the wheelchair inclination angle and terrain slope added on the basis of the first state space model; when the control mode is the standing mode, calling a third state space model with the user's standing intention confidence and leg rest support force added on the basis of the second state space model.

[0103] This technical solution designs dedicated state space models for different control modes to meet the special requirements of various scenarios. Specifically, first, the type of the currently selected control mode is obtained, including the flat ground mode, the uphill and downhill mode, and the standing mode, which lays the foundation for subsequent selection of the appropriate state space model. For the flat ground mode, a first state space model including the leg rest position, speed, user weight distribution, and system energy state is used. This model is suitable for the relatively simple flat ground environment and can effectively control the basic movement of the leg rest. For the uphill and downhill mode, the wheelchair inclination angle and terrain slope are added on the basis of the first state space model to form a second state space model. This enables the system to adapt to slope changes and improves the control accuracy in non-flat ground environments. For the standing mode, the user's standing intention confidence and leg rest support force are further added on the basis of the second state space model to form a third state space model. This enables the system to better support the user's transition from a sitting position to a standing position and improves safety and comfort.

[0104] Furthermore, the technical solution of the present application provides multiple possible implementation manners. For example, when obtaining the currently selected control mode type, it can be achieved through sensor data fusion. Specifically, the data of the gyroscope, accelerometer, and pressure sensor can be combined, and the current control mode can be identified through machine learning algorithms (such as decision trees or support vector machines). This method can improve the accuracy and robustness of mode recognition.

[0105] For different state space models, different mathematical expression forms can be adopted. For example, the first state space model can be expressed as: x(t + 1) = A1 * x(t) + B1 * u(t); y(t) = C1 * x(t); Among them, x(t + 1) is the state vector at time step t + 1, y(t) is the output vector at time step t, including the leg rest position, speed, user weight distribution, and system energy state; u(t) represents the control input; y(t) is the output vector at time step t; A1, B1, and C1 are the corresponding system matrices.

[0106] The second state space model can be expressed as: x'(t + 1) = A2 * x'(t) + B2 * u(t) + D2 * d(t); y'(t) = C2 * x'(t); Among them, x'(t + 1) adds the wheelchair inclination angle and terrain slope on the basis of x(t + 1); d(t) represents the external disturbance; D2 is the disturbance matrix.

[0107] The third state space model can be expressed as: x''(t + 1) = A3 * x''(t) + B3 * u(t) + D3 * d(t) + E3 * s(t); y''(t) = C3 * x''(t); Among them, x''(t + 1) adds the user's standing intention confidence and leg rest support force on the basis of x'(t + 1); s(t) represents the special input related to the user's standing; E3 is the corresponding input matrix.

[0108] This hierarchical and progressive model design method enables the system to select the most suitable state space model for different control modes, thereby improving the accuracy and adaptability of leg rest extension control. By introducing specific key parameters, such as wheelchair inclination angle, terrain slope, user's standing intention confidence, etc., in different models, the pertinence and effectiveness of the models are improved. This method solves the problem that a single model is difficult to adapt to multiple complex scenarios, and significantly improves the usage experience and safety of electric wheelchairs in various environments.

[0109] As a preferred implementation, the technical solution of the present application can be realized as follows: First, the system monitors the environment and user status in real time through multi-sensor fusion technology. For example, a gyroscope and an accelerometer are used to detect the inclination angle and motion state of the wheelchair, a pressure sensor is used to detect the user's weight distribution, and a distance sensor is used to detect terrain changes.

[0110] When the system detects that the control mode needs to be switched (such as when entering an uphill from flat ground), it will trigger the mode switching program. Assuming that the current mode is switched from the flat ground mode to the uphill / downhill mode, the system will perform the following steps: Obtain the current state: the leg rest position p0 = 0.3m (relative to the seat reference plane), the speed v0 = 0m / s, the user weight distribution m0 = [60%, 40%] (front-back distribution), and the system energy state E0 = 80%.

[0111] Detect new parameters: the wheelchair inclination angle α = 10° (uphill), and the terrain slope β = 12°.

[0112] Construct a new state vector: x' = [p0, v0, m0, E0, α, β].

[0113] Apply the second state space model: x'(t + 1) = A2 * x'(t) + B2 * u(t) + D2 * d(t); Where A2, B2, and D2 are matrices pre-trained through machine learning algorithms.

[0114] Solve the optimization problem to obtain the optimal control sequence u*(t).

[0115] Calculate the new leg rest position and speed according to u*(t): p1 = 0.35m (slightly increased to relieve the user's leg pressure); v1 = 0.02m / s (slowly adjusted to the new position); Execute the leg rest adjustment while continuously monitoring the user feedback and system status.

[0116] In this way, the system can smoothly transition from the flat ground mode to the uphill / downhill mode, providing a more comfortable and safe riding experience for the user. At the same time, since the model takes into account the system energy state, it can also optimize energy use while ensuring functionality.

[0117] The technical solution of this application significantly improves the accuracy and adaptability of the leg rest control of an electric wheelchair by designing a dedicated state space model for different control modes. Compared with using a single model, this method can better cope with complex and variable usage environments and provide more precise control. For example, in the uphill / downhill mode, considering the wheelchair inclination and terrain slope, the system can more accurately adjust the leg rest position, improving the comfort and safety of the user on the ramp. In the standing mode, by introducing the confidence level of the user's standing intention and the leg rest support force, the system can provide more precise standing assistance, reducing the discomfort and fall risk of the user during the standing process. In addition, this hierarchical and progressive model design method also improves the overall robustness and scalability of the system, facilitating the addition of new control modes in the future.

[0118] In some of the above embodiments of this application, a step of correcting the applicable probability of each control mode based on the historical data in the evaluation matrix is proposed to select the most suitable control mode for the current situation. However, in this process, the historical data may contain outdated or abnormal information, and directly using this data may lead to deviations in the selection of the control mode, affecting the accuracy and efficiency of the electric wheelchair leg rest extension control.

[0119] In response to this, the steps for correcting the applicable probability of each control mode based on the historical data in the evaluation matrix further proposed in this application include: obtaining the historical data in the evaluation matrix, including historical control mode selection records, corresponding control effect scores, control mode switching frequencies, and user feedback; performing a timeliness evaluation on the historical data to eliminate data that exceeds a preset time threshold; performing outlier detection on the remaining data to identify and mark abnormal data; calculating the historical performance scores of each control mode based on the valid historical data after removing the abnormal data; and performing weighted fusion of the historical performance scores and the currently calculated applicable probabilities to obtain the corrected applicable probabilities of each control mode.

[0120] The technical solution proposed in this application optimizes the control mode selection process through multiple steps. First, by obtaining comprehensive historical data, including historical control mode selection records, control effect scores, switching frequencies, and user feedback, a comprehensive data basis is provided for subsequent analysis. Second, by performing a timeliness evaluation to eliminate data that exceeds a preset time threshold, it is ensured that the historical data used has timeliness and avoids the influence of outdated information on the current decision. Third, by performing outlier detection to identify and mark abnormal data, it is prevented that outliers have an improper impact on the selection of the control mode, improving data reliability. Then, based on the valid historical data after removing the abnormal data, the performance scores of each control mode are calculated to reflect the actual effects of each mode in past use. Finally, the historical performance scores and the currently calculated applicable probabilities are weighted and fused, taking into account both historical performance and the current situation, to obtain a more accurate applicable probability of the control mode.

[0121] In some of the above embodiments of the present application, a third state - space model is proposed to calculate the optimal position and extension speed of the leg rest to support the user's transition from a sitting position to a standing position. However, in this process, how to accurately construct the state - space model, how to effectively consider multiple key parameters, and how to optimize the control strategy under multiple constraint conditions still remains a complex challenge. Especially in practical applications, multiple aspects such as user safety, system performance, and energy efficiency need to be considered simultaneously, which requires a more comprehensive and refined control method.

[0122] To this end, the present application further proposes to construct a state vector x(t), including the leg - rest position p(t), the leg - rest speed v(t), the user's weight distribution m(t), the system energy state E(t), the wheelchair inclination angle α(t), the terrain slope β(t), and the leg - rest support force F(t); based on the current control mode θ(t), construct a state - space model: x(t + 1)=A(θ(t),t)x(t)+B(θ(t),t)u(t)+D(θ(t),t)d(t)+w(t), y(t)=C(θ(t),t)x(t)+v_noise(t); calculate the user's standing - intention confidence CI(t); according to the current state x(t) and the user's standing - intention confidence CI(t), select the optimal control mode θ(t); construct an objective function: J = Σ(w1(p(t)-p_desired(t))^2+w2(v(t)-v_desired(t))^2+w3(F(t)-F_desired(t))^2+λu(t)^2); under the premise of satisfying the constraint conditions p_min≤p(t)≤p_max, v_min≤v(t)≤v_max, F_min≤F(t)≤F_max, E(t)≥E_min, |α(t)|≤α_max, |β(t)|≤β_max, solve the optimization problem of minimizing the objective function J to obtain the optimal control sequence u*(t); according to the optimal control sequence u*(t), calculate the optimal position and extension speed of the leg rest.

[0123] This technical solution involves multiple key technical features, including the state vector x(t), the state - space model, the user's standing - intention confidence CI(t), the objective function J, and the constraint conditions. These features cooperate with each other to jointly solve the problems of precise control, safety guarantee, and energy - efficiency optimization.

[0124] Specifically, the state vector x(t) includes multiple parameters such as the leg rest position, speed, user weight distribution, system energy state, wheelchair inclination angle, terrain slope, and leg rest support force. This comprehensive consideration of parameters enables the system to more accurately describe the dynamic characteristics of the electric wheelchair. For example, by considering the user weight distribution, the system can better adapt to the needs of users with different body types; by including the wheelchair inclination angle and terrain slope, the system can maintain stability in complex terrains.

[0125] The state space model is constructed based on the control mode θ(t) and is a time-varying model. This design allows the system to dynamically adjust its behavior according to different control modes. For example, in the flat ground mode and uphill / downhill mode, the system may adopt different parameter matrices A(θ(t),t), B(θ(t),t), C(θ(t),t), and D(θ(t),t) to achieve more precise control. In addition, the model includes external disturbances d(t), process noise w(t), and observation noise v_noise(t), which makes the model closer to the actual situation and capable of handling various uncertainties.

[0126] The introduction of the user's standing intention confidence CI(t) is an innovation point. This parameter can help the system better understand and respond to the user's intention. By calculating and updating CI(t) in real time, the system can react in advance when the user is about to stand up, providing more timely and smooth support. For example, when CI(t) gradually increases, the system may pre-adjust the leg rest position in preparation for the upcoming standing action.

[0127] The objective function J comprehensively considers the deviations of the leg rest position, speed, and support force from the expected values, as well as the control cost. This multi-objective optimization method can ensure control accuracy while also considering the energy efficiency of the system. By adjusting the weight coefficients w1, w2, w3, the system can flexibly balance the importance of each objective according to the requirements of different scenarios. For example, during the user's standing process, the weight of the support force term may be increased to ensure the user's safety.

[0128] The setting of the constraint conditions ensures that the system operates within a safe range. These constraints cover multiple aspects such as the leg rest position, speed, support force, system energy, wheelchair inclination angle, and terrain slope. By strictly adhering to these constraints, the system can provide optimal performance while always maintaining safety. For example, by restricting the wheelchair inclination angle |α(t)|≤α_max, the wheelchair can be prevented from tipping over on a steep slope.

[0129] The synergistic effect of these technical features enables the system to achieve precise, safe, and efficient legrest control in complex environments. For example, during the transition of the user from a sitting position to a standing position, the system comprehensively considers factors such as the user's weight distribution, the inclination angle of the wheelchair, and the terrain slope, and dynamically adjusts the control strategy according to the confidence level CI(t) of the user's standing intention. On the premise of meeting safety constraints, the system calculates the optimal legrest position and extension speed, which can not only provide sufficient support force but also ensure the smoothness of the movement.

[0130] By solving the optimization problem of minimizing the objective function J, the system can obtain the optimal control sequence u*(t). This control sequence not only considers the current state but also the prediction of several future time steps, thus achieving more forward-looking and coherent control. For example, when the user is about to stand up, the system may start adjusting the legrest position several time steps in advance to achieve a smoother transition.

[0131] Finally, the optimal legrest position and extension speed calculated according to the optimal control sequence u*(t) can provide precise, safe, and comfortable support for the user. This control method can not only adapt to complex terrains and user needs but also optimize energy use while ensuring performance, extending the battery life of the electric wheelchair.

[0132] As a preferred implementation manner, the following specific embodiments can be considered: Suppose a user is using an electric wheelchair equipped with this control system and is about to transition from a sitting position to a standing position. The system first constructs a state vector x(t), which includes the following parameters: The legrest position p(t) = 0.3m (distance from the ground); The legrest speed v(t) = 0 m / s (initial stationary state); The user's weight distribution m(t) = [60%, 40%] (60% of the weight on the seat and 40% on the legrest); The system's energy state E(t) = 80% (remaining battery charge); The wheelchair inclination angle α(t) = 2° (slightly forward tilted); The terrain slope β(t) = 5° (slightly uphill); The legrest support force F(t) = 200N; Based on the current control mode θ(t) (assumed to be the standing mode), the system constructs a state space model. In this example, A(θ(t),t), B(θ(t),t), C(θ(t),t), and D(θ(t),t) are time-varying matrices obtained in advance through system identification and machine learning methods.

[0133] The system calculates the confidence level CI(t) of the user's standing intention as 0.8 (indicating that the user is very likely to be about to stand up). Based on the current state x(t) and CI(t), the system confirms that the optimal control mode θ(t) is the standing assistance mode.

[0134] Next, the system constructs the objective function J, where the weight coefficients are set as w1 = 0.4, w2 = 0.3, w3 = 0.3, and the control cost coefficient λ = 0.1. The expected values are set as follows: p_desired(t) = 0.7m (the target height of the leg rest); v_desired(t) = 0.1 m / s (the slow rising speed); F_desired(t) = 400N (the gradually increasing supporting force); The system solves the optimization problem on the premise of satisfying the following constraints: 0.2m ≤ p(t) ≤ 0.8m; 0 m / s ≤ v(t) ≤ 0.2 m / s; 100N ≤ F(t) ≤ 600N; E(t) ≥ 20%; |α(t)| ≤ 10°; |β(t)| ≤ 15°; By solving this optimization problem, the system obtains the optimal control sequence u*(t). Based on u*(t), the system calculates the optimal position and extension speed of the leg rest. For example, within the next 5 seconds, the position of the leg rest may change according to the following sequence: 0.3m → 0.4m → 0.5m → 0.6m → 0.7m; The corresponding speed sequence may be: 0.1 m / s → 0.12 m / s → 0.08 m / s → 0.1 m / s → 0.05 m / s; This control strategy can achieve a smooth rise of the leg rest, and at the same time dynamically adjust the speed according to the user's reaction, providing safe and comfortable standing assistance.

[0135] By implementing the technical solution of the present application, the leg rest control system of the electric wheelchair can more precisely, safely and efficiently support the user to transition from a sitting position to a standing position. This solution takes into account multiple key parameters, including the user's weight distribution, the energy state of the system, etc., making the control more comprehensive and precise. The introduction of the confidence level of the user's standing intention enhances the intelligence of the system and the ability to respond to the user's needs. The multi-objective optimization method is adopted, which not only ensures the control accuracy but also considers the energy efficiency. Comprehensive constraint conditions are set to effectively ensure the safety of the system. This method not only solves the problem of precise control but also significantly improves in terms of safety, energy efficiency and user experience.

[0136] In a second aspect, referring to Figure 2 , the present application further provides an electric wheelchair leg rest extension control device, which includes: An acquisition module 210, configured to acquire the user's weight distribution, wheelchair attitude, motion state and environmental terrain information when it is detected that the terrain complexity of the electric wheelchair exceeds a set value or the user needs to adjust the body position; An identification module 220, configured to identify the current wheelchair state and the user's intention based on the acquired information, and select a corresponding control mode from a variety of preset control modes according to the identification result; A calculation module 230, configured to call a corresponding state space model considering the leg rest position, speed, user's weight distribution and system energy state according to the selected control mode, and calculate the optimal leg rest position and extension speed in real time; A control module 240, configured to control the leg rest to perform an extension action according to the calculated optimal leg rest position and extension speed.

[0137] By acquiring and analyzing the user and environmental information in real time, adaptively selecting the most suitable control mode, and accurately calculating the leg rest position and speed using the state space model, the adaptability and real-time performance problems of the traditional control system under complex terrains and diverse body position requirements are solved. It has the advantages of being able to adaptively adjust the control mode according to complex terrains and user needs, realizing real-time and precise control of the leg rest position and extension speed, and improving the use comfort and safety.

[0138] In addition, in some preferred embodiments, an electric wheelchair leg rest extension control device proposed by the present application can execute any one of the steps in the above method.

[0139] The above are only the embodiments of the present application and are not intended to limit the protection scope of the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for controlling the extension of a leg support of an electric wheelchair, characterized in that: The method includes: When it is detected that the complexity of the terrain where the electric wheelchair is located exceeds the set value or the user needs to adjust the body position, the user's weight distribution, wheelchair posture, movement status and environmental terrain information are obtained; Based on the acquired information, the current wheelchair state and the user's intention are identified, and a corresponding control mode is selected from a plurality of preset control modes according to the identification result; According to the selected control mode, the corresponding state space model considering the leg rest position, speed, user weight distribution and system energy state is called to calculate the optimal leg rest position and extension speed in real time; The leg support is controlled to perform the stretching action according to the calculated optimal leg support position and stretching speed.

2. The electric wheelchair leg support extension control method according to claim 1, characterized in that: The step of identifying the current wheelchair state and the user's intention based on the acquired information, and selecting a corresponding control mode from a plurality of preset control modes according to the identification result comprises: Extract and classify the acquired user weight distribution, wheelchair posture, motion status and environmental terrain information; Based on the classification results, the applicability probabilities of each preset control mode are calculated, and the control mode with the highest applicability probability is selected as the corresponding control mode, wherein the classification results are used to characterize the current wheelchair state and the user's intention; Also includes: According to the selected corresponding control mode and the current wheelchair state, the transition control parameters when the control mode is switched are calculated to achieve a smooth transition between modes when the mode is switched.

3. The electric wheelchair leg support extension control method according to claim 1, characterized in that: The step of calling the corresponding state space model considering the leg rest position, speed, user weight distribution and system energy state according to the selected control mode to calculate the optimal leg rest position and extension speed in real time comprises: According to the selected control mode, the corresponding state space model considering the leg rest position, speed, user weight distribution and system energy state is called; Get the current system computing resource status and remaining battery power; Based on the acquired system computing resource status and the remaining battery power, adjust the complexity and calculation frequency of the state space model, wherein when the system computing resources are sufficient and the remaining battery power is higher than a first preset threshold, increase the model complexity and calculation frequency, otherwise reduce the model complexity and calculation frequency; The optimal leg rest position and extension speed are calculated using the adjusted state-space model in combination with the real-time acquired leg rest position, speed, user weight distribution, and system energy state.

4. The electric wheelchair leg support extension control method according to claim 2, characterized in that: The step of calculating the applicability probability of each preset control mode based on the classification result and selecting the control mode with the highest applicability probability as the corresponding control mode includes: When it is detected that the difference in applicability probabilities of the multiple control modes is less than a second preset threshold, obtaining an evaluation matrix including historical control mode selection records, corresponding control effect scores, control mode switching frequencies, and user feedback; Based on the historical data in the evaluation matrix, the applicability probability of each control mode is modified; The control mode with the highest probability of applicability after correction is selected as the corresponding control mode.

5. The electric wheelchair leg support extension control method according to claim 3, characterized in that: The step of adjusting the complexity and calculation frequency of the state space model based on the acquired system computing resource state and battery remaining power includes: Get the model complexity and calculation frequency under the current control mode; Determine the target model complexity and target calculation frequency based on the acquired system computing resource status and battery remaining power; Based on the difference between the current value and the target value, combined with the preset smoothing factor, a progressive adjustment sequence of model complexity and calculation frequency is generated; The state space model is gradually adjusted according to the generated adjustment sequence, and the control effect is monitored at the same time. When it is detected that the control accuracy drops by more than a third preset threshold, the adjustment is suspended and returns to the previous stable state.

6. The electric wheelchair leg support extension control method according to claim 3, characterized in that: The step of calling the corresponding state space model considering the leg rest position, speed, user weight distribution and system energy state according to the selected control mode includes: Obtain the type of the currently selected control mode, wherein the control modes include flat ground mode, uphill and downhill mode, and standing mode; When the control mode is the flat ground mode, a first state space model including the leg rest position, speed, user weight distribution and system energy state is called; When the control mode is the uphill and downhill mode, a second state space model is called based on the first state space model to add the wheelchair inclination angle and the terrain slope; When the control mode is the standing mode, a third state space model is called based on the second state space model to add the user's standing intention confidence and the leg support force.

7. The electric wheelchair leg support extension control method according to claim 4, characterized in that: The step of correcting the applicable probability of each control mode based on the historical data in the evaluation matrix includes: Obtain historical data in the evaluation matrix, including historical control mode selection records, corresponding control effect scores, control mode switching frequency, and user feedback; Evaluate the timeliness of historical data and remove data that exceeds the preset time threshold; Perform outlier detection on the remaining data to identify and mark abnormal data; Based on the valid historical data after removing abnormal data, calculate the historical performance score of each control mode; The historical performance score is weightedly integrated with the currently calculated applicability probability to obtain the revised applicability probability of each control mode.

8. The electric wheelchair leg support extension control method according to claim 6, characterized in that: When the control mode is the standing mode, the step of calling the third state space model based on the second state space model with the user's standing intention confidence and the leg support force added includes: Obtain the user's standing intention confidence and leg support strength; Dynamically adjust the parameter weights in the second state space model according to the user's standing intention confidence and the leg support force; applying the adjusted parameter weights to the second state-space model to form a third state-space model adapted to the standing mode; The optimal position and extension velocity of the leg rest are calculated using a third state-space model to support the user's transition from a sitting to a standing posture.

9. The electric wheelchair leg support extension control method according to claim 8, characterized in that: The step of using the third state space model to calculate the optimal position and extension speed of the leg rest to support the user's transition from a sitting position to a standing position includes: Construct the state vector x(t), which includes the leg support position p(t), leg support velocity v(t), user weight distribution m(t), system energy state E(t), wheelchair inclination angle α(t), terrain slope β(t) and leg support force F(t); Based on the current control mode θ(t), the state space model is constructed: x(t+1)=A(θ(t),t)x(t)+B(θ(t),t)u(t)+D(θ(t),t)d(t)+w(t) y(t)=C(θ(t),t)x(t)+v_noise(t) Where x(t+1) is the state vector at time step t+1, y(t) is the output vector at time step t, A(θ(t),t), B(θ(t),t), C(θ(t),t) and D(θ(t),t) are time-varying matrices that depend on the control mode, u(t) is the control input vector, d(t) is the external disturbance vector, w(t) is the process noise, and v_noise(t) is the observation noise; Calculate the user's standing intention confidence CI(t); According to the current state x(t) and the user's standing intention confidence CI(t), the optimal control mode θ(t) is selected; Construct the objective function: J=Σ(w1(p(t)-p_desired(t))^2+w2(v(t)-v_desired(t))^2+w3(F(t)-F_desired(t))^2+λu(t)^2) Among them, w1, w2, w3 are weight coefficients, λ is the control cost coefficient, p_desired(t) is the desired leg support position, v_desired(t) is the desired leg support velocity, and F_desired(t) is the desired leg support force; Under the premise of satisfying the constraints p_min≤p(t)≤p_max, v_min≤v(t)≤v_max, F_min≤F(t)≤F_max, E(t)≥E_min, |α(t)|≤α_max, |β(t)|≤β_max, solve the optimization problem of minimizing the objective function J and obtain the optimal control sequence u*(t), where p_min and p_max are the minimum and maximum values ​​of the leg rest position, v_min and v_max are the minimum and maximum values ​​of the leg rest speed, F_min and F_max are the minimum and maximum values ​​of the leg rest support force, E_min is the minimum value of the system energy state, α_max is the maximum absolute value of the wheelchair inclination, and β_max is the maximum absolute value of the terrain slope; According to the optimal control sequence u*(t), the optimal position and extension speed of the leg support are calculated.

10. An electric wheelchair leg support extension control device, characterized in that: The device includes: An acquisition module is used to acquire the user's weight distribution, wheelchair posture, motion state and environmental terrain information when it is detected that the terrain complexity of the electric wheelchair exceeds a set value or the user needs to adjust the body position; An identification module, used to identify the current wheelchair state and user intention based on the acquired information, and select a corresponding control mode from a plurality of preset control modes according to the identification result; A calculation module, for invoking a corresponding state space model that considers the leg rest position, speed, user weight distribution and system energy state according to the selected control mode, and calculating the optimal leg rest position and extension speed in real time; The control module is used to control the leg rest to perform the stretching action according to the calculated optimal leg rest position and stretching speed.

Images