Intelligent power-assisted balance car and control system

Abstract

The invention discloses an intelligent power-assisted balance car and a control system, and relates to the technical field of balance cars, the intelligent power-assisted balance car comprises a child interaction layer, an intelligent decision-making layer, a motion control layer and an execution layer, the child interaction layer is used for interacting with a child, including operation and feedback, and providing an interaction mode of a touch screen and a voice instruction; the intelligent decision-making layer is used for realizing a core part of an intelligent control algorithm, making a decision according to sensing layer data and a target instruction, and outputting a control signal to the execution layer; and the execution layer is used for realizing specific physical actions of the balance car according to the control signal and implementing a safety guarantee mechanism. By integrating various control algorithms and a flexible scene switching mechanism, the stable operation of the vehicle under different road conditions is ensured, meanwhile, the safety, comfort and high efficiency are ensured, the reliability and adaptability of the system are further improved through redundancy design, real-time guarantee and an online learning mechanism, and the system is suitable for large-scale popularization and application. And finally, safe, comfortable and efficient intelligent driving experience is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of balancing vehicles, and in particular to an intelligent power-assisted balancing vehicle and a control system. Background Art

[0002] Children's balance bikes are a means of transportation that helps children learn how to maintain balance, promoting motor skills and coordination. Traditional balance bikes rely on children's feet to maintain balance, which is simple and effective. However, with technological advancements, modern smart power-assisted balance bikes are beginning to incorporate more innovative features, including electric power systems, intelligent control systems, and safety measures, further improving the convenience and safety of riding. In addition to traditional outdoor riding, smart power-assisted balance bikes can also be used for children's daily exercise and entertainment, and even as a new means of transportation.

[0003] At present, the invention patent with application number PCT / CN2020 / 121532 discloses a main control system, control system and balance vehicle for a balance vehicle. The main control system ensures stable communication of the Bluetooth module, simplifies the production process and reduces manufacturing costs. The balance vehicle control system adopts a dual system to realize the control of the balance vehicle. The main control system and the auxiliary control system respectively control one motor of the balance vehicle. However, the existing technology cannot ensure the smooth operation of the vehicle under different road conditions by integrating multiple control algorithms and flexible scene switching mechanisms, while ensuring safety, comfort and efficiency. It cannot further improve the reliability and adaptability of the system through redundant design, real-time guarantee and online learning mechanism, and cannot ultimately achieve a safe, comfortable and efficient intelligent driving experience. Summary of the Invention

[0004] The technical problem solved by the present invention is that the existing technology cannot ensure the smooth operation of the vehicle under different road conditions by integrating multiple control algorithms and flexible scene switching mechanisms, cannot guarantee safety, comfort and efficiency at the same time, cannot further improve the reliability and adaptability of the system through redundant design, real-time guarantee and online learning mechanism, and cannot ultimately achieve a safe, comfortable and efficient intelligent driving experience.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: a control system for an intelligent power-assisted balancing vehicle, comprising a child interaction layer, an intelligent decision-making layer, a motion control layer, and an execution layer:

[0006] The child interaction layer is used to interact with children, including operations and feedback, and provides touch screen and voice command interaction methods;

[0007] The intelligent decision-making layer is used to implement the core part of the intelligent control algorithm, make decisions based on the perception layer data and target instructions, and output control signals to the execution layer;

[0008] The execution layer is used to realize the specific physical actions of the balancing vehicle according to the control signal and implement the safety guarantee mechanism.

[0009] Preferably, the perception layer includes a posture detection unit, an environment perception unit and a data fusion unit:

[0010] The attitude detection unit is used to obtain the vehicle's inclination angle and angular velocity through the MPU6050 gyroscope and accelerometer, remove noise through Kalman filtering, and output the vehicle's pitch and roll angle data in real time;

[0011] The environmental perception unit is used to detect tire ground pressure distribution through a pressure sensor array, calculate pressure variance, build a road bumpiness grade model based on the pressure variance, and trigger active suspension system adjustments. It also uses a lidar to detect obstacles and obtain obstacle avoidance signals, which are transmitted to the decision-making layer via the CAN bus.

[0012] The data fusion unit is used to fuse IMU and pressure sensor data through extended Kalman filtering.

[0013] Preferably, the environment perception unit includes:

[0014] The road surface grade of the road bump grade model is determined by threshold segmentation, and the road surface grade determination includes:

[0015] Low bump: pressure variance <0.5N 2 / cm 4 ;

[0016] Moderate bumps: 0.5 ≤ pressure variance < 2N 2 / cm 4 ;

[0017] High turbulence: variance ≥ 2N 2 / cm 4 ;

[0018] Based on the road surface grade determination, a CNN model is trained, pressure time series data is input, and road surface types are output, where the road surface types include urban roads, off-road roads, and emergency roads;

[0019] When a bump is detected, a command is sent to the suspension controller via the CAN bus;

[0020] The ROS robot operating system is used to receive point cloud data, extract obstacle contours, and perform DBSCAN clustering on the point cloud data into obstacle areas.

[0021] Preferably, the intelligent decision-making layer is used to construct a hierarchical control architecture for a self-balancing vehicle, which includes a PID speed control unit, an LQR attitude control unit, and an MPC prediction control unit:

[0022] The PID speed control unit is used to adjust the vehicle speed and dynamically respond to children's instructions. The basic balance ring adopts position PID (Kp = 120, Ki = 8, Kd = 0.5), and the parameters are optimized by MATLAB genetic algorithm;

[0023] The LQR attitude control unit is used to adjust the vehicle body attitude, with the vehicle body tilt angle error and suspension deformation as state variables, and the weight matrix Q = diag (10, 1, 5, 0.1);

[0024] The MPC prediction control unit is used to dynamically adjust the motor output by predicting the trajectory for a period of time in the future. The rolling horizon is set to 1 second, and the solver adopts QP quadratic programming;

[0025] Dynamically responding to children's instructions includes: receiving instructions from the child interaction layer, the instructions including forward, left turn, right turn, path tracking, emergency braking, cruise and obstacle avoidance, and prioritizing the instructions. The priority ranking content is safety instructions > dynamic instructions > regular instructions. The safety instructions include emergency braking and obstacle avoidance, the dynamic instructions include cruise and path tracking, and the regular instructions include forward, left turn and right turn.

[0026] Preferably, the environment is matched according to the data output by the child interaction layer, and the control strategy is automatically switched, and the control strategy includes:

[0027] On urban roads, the PID speed control unit is used to control the balance, and the MPC prediction control unit is used to track the path.

[0028] The LQR attitude control unit is used for suspension adjustment and fuzzy PID control on off-road surfaces;

[0029] In an emergency, the priority channel is set for quick response, and the PID speed control unit is switched to perform emergency braking control on the balancing vehicle.

[0030] Preferably, the dynamic parameter optimization of the hierarchical control architecture of the self-balancing vehicle is performed, and the optimization process includes:

[0031] A self-balancing vehicle state space is defined, where the states of the self-balancing vehicle state space include vehicle speed, vehicle posture, current road conditions, and control inputs, and the child's riding habits are used as additional state variables. The control inputs include the output of a PID controller, the vehicle posture includes angle θ and angular velocity ω, the current road conditions include wet, flat, and bumpy, and the riding habits include riding speed, perceptual demand data, and motion mode. The perceptual demand data includes:

[0032] Defining a self-balancing vehicle motion space, the self-balancing vehicle motion space including parameter adjustment for a child's instruction in response to the PID speed control unit, the parameters including proportional gain, integral gain, and differential gain;

[0033] The system continuously monitors children's riding habits and vehicle responses per unit time, including acceleration and deceleration frequency and mode preferences. Deep Q-network trial-and-error learning of environmental conditions, including road conditions, vehicle speed, and vehicle posture, is used to optimize the parameters of the self-balancing vehicle's hierarchical control architecture. The MPC prediction model is used to optimize long-term strategies, including slowing down in advance on slippery roads to avoid slipping.

[0034] Preferably, during the trial-and-error learning process, each action is continuously evaluated, and the parameters are adjusted through feedback from the reward function, and the next learning is influenced according to the feedback;

[0035] Define the reward function of the hierarchical control architecture of the self-balancing vehicle. The mathematical expression of the reward function is:

[0036] R=ω1X St +ω2X SC +ω3X UP -ω4X E ;

[0037] Among them, R is the reward function, ω1, ω2, ω3 and ω4 are weight coefficients used to adjust the contribution of each factor to the reward, X St is the stability of attitude control, X SC For speed control, X UP For children's preference matching, X E For energy efficiency.

[0038] Preferably, the execution layer includes a safety assurance unit and a drive control unit:

[0039] The safety assurance unit includes a dual-mode decision-making mechanism and a fuse mechanism;

[0040] The dual-mode decision-making mechanism includes: the main controller and the backup controller work alternately, compare the outputs in real time, and if the difference between the main controller and the backup controller exceeds a preset difference threshold, trigger an alarm or switch to the backup controller;

[0041] The fusing mechanism includes automatically switching to a backup sensor or entering a degraded mode when sensor data is abnormal, including gyroscope drift and reading failure.

[0042] Preferably, the drive control unit includes an instruction arbitration mechanism, a real-time control mechanism and a state feedback mechanism:

[0043] The command arbitration mechanism includes: when a child intervenes, the autonomous driving algorithm is suspended and manual commands are executed first;

[0044] The real-time control mechanism includes: using the FreeRTOS real-time operating system to schedule tasks and divide task priorities, wherein the task priorities include first priority, second priority, third priority and basic task level, first priority tasks are executed every 10ms, second priority tasks are executed every 100ms, and third priority tasks are executed every 1s, and tasks at the basic task level are executed by accelerating matrix operations through FPGA chips;

[0045] The state feedback mechanism includes: informing the child of the current control mode through voice prompts.

[0046] An intelligent power-assisted balancing vehicle comprises a handlebar, a vehicle frame, wheels, a child interaction layer, an intelligent decision-making layer, a motion control layer and an execution layer.

[0047] The beneficial effects of the present invention are as follows: by integrating multiple control algorithms and a flexible scene switching mechanism, the smooth operation of the vehicle under different road conditions is ensured, while safety, comfort and efficiency are guaranteed. The reliability and adaptability of the system are further improved through redundant design, real-time guarantee and online learning mechanism, ultimately achieving a safe, comfortable and efficient intelligent driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of the basic flow of a control system for an intelligent power-assisted balancing vehicle provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0050] Reference Figure 1 , which is an embodiment of the present invention, provides an intelligent power-assisted balancing vehicle and control system, including a child interaction layer, an intelligent decision-making layer, a motion control layer, and an execution layer:

[0051] The child interaction layer is used to interact with children, including operations and feedback, and provides touch screen and voice command interaction methods;

[0052] The intelligent decision-making layer is used to implement the core part of the intelligent control algorithm, make decisions based on the perception layer data and target instructions, and output control signals to the execution layer;

[0053] The execution layer is used to realize the specific physical actions of the balancing vehicle according to the control signals and implement the safety guarantee mechanism.

[0054] The perception layer includes a posture detection unit, an environment perception unit, and a data fusion unit:

[0055] The attitude detection unit is used to obtain the vehicle's inclination and angular velocity through the MPU6050 gyroscope and accelerometer, remove noise through Kalman filtering, and output the vehicle's pitch and roll angle data in real time;

[0056] The environmental perception unit detects tire ground pressure distribution through a pressure sensor array, calculates pressure variance, builds a road surface roughness model based on the pressure variance, and triggers adjustments to the active suspension system. It also uses lidar to detect obstacles and obtain obstacle avoidance signals, which are then transmitted to the decision-making layer via the CAN bus.

[0057] The data fusion unit is used to fuse the IMU and pressure sensor data through extended Kalman filtering.

[0058] The environmental perception unit includes:

[0059] The road surface grade of the road bump grade model is determined by threshold segmentation. The road surface grade determination includes:

[0060] Low bump: pressure variance <0.5N 2 / cm 4 ;

[0061] Moderate bumps: 0.5 ≤ pressure variance < 2N 2 / cm 4 ;

[0062] High turbulence: variance ≥ 2N 2 / cm 4 ;

[0063] The CNN model is trained based on road surface grade determination, with pressure time series data as input and road surface type output, including urban roads, off-road roads, and emergency roads.

[0064] When a bump is detected, a command is sent to the suspension controller via the CAN bus;

[0065] The ROS robot operating system is used to receive point cloud data, extract obstacle contours, and perform DBSCAN clustering on the point cloud data into obstacle areas. If the obstacle distance is less than 50 cm, emergency braking or detour strategies are triggered.

[0066] The intelligent decision-making layer is used to build a hierarchical control architecture for the self-balancing vehicle. The hierarchical control architecture includes a PID speed control unit, an LQR attitude control unit, and an MPC prediction control unit:

[0067] The PID speed control unit is used to adjust the vehicle speed and dynamically respond to the child's instructions. The basic balance loop uses a position PID (Kp = 120, Ki = 8, Kd = 0.5), and the parameters are optimized using the MATLAB genetic algorithm;

[0068] The LQR attitude control unit is used to adjust the vehicle body attitude, with the vehicle body tilt angle error and suspension deformation as state variables, and the weight matrix Q = diag(10,1,5,0.1);

[0069] The MPC predictive control unit is used to dynamically adjust the motor output by predicting the trajectory over a period of time in the future. The rolling horizon is set to 1 second, and the solver uses QP quadratic programming to cope with complex road conditions.

[0070] Dynamic response to children's commands includes: receiving commands from the child interaction layer, including forward, left turn, right turn, path tracking, emergency braking, cruise and obstacle avoidance, and prioritizing the commands. The priority ranking content is safety commands > dynamic commands > regular commands. Safety commands include emergency braking and obstacle avoidance, dynamic commands include cruise and path tracking, and regular commands include forward, left turn and right turn.

[0071] The system matches the environment based on the data output by the child interaction layer and automatically switches the control strategy. The control strategy includes:

[0072] On urban roads, the PID speed control unit is used to control balance, and the MPC prediction control unit is used for path tracking, which is suitable for low-speed comfortable driving.

[0073] On off-road surfaces, the LQR attitude control unit is used for suspension adjustment and fuzzy PID control to cope with complex terrain and limit vehicle speed to ensure safety;

[0074] In an emergency, the priority channel is set for quick response, and the PID speed control unit is switched to perform emergency braking control on the balancing vehicle, ignoring energy consumption optimization to ensure safety.

[0075] Fuzzy PID is used to handle nonlinear scenarios (such as driving on sand) and enhance the dynamic response capability of the vehicle.

[0076] Scene recognition and strategy switching, for example:

[0077] Urban road scenario: Enable PID balancing + MPC path tracking, limit speed to ≤ 20 km / h, and prioritize comfort.

[0078] Off-road scenario: Activate LQR suspension adjustment + fuzzy PID anti-disturbance control, enable lidar obstacle avoidance, and maintain a speed of ≤10 km / h.

[0079] Emergency avoidance scenario: Switch to pure PID fast response and ignore energy consumption optimization.

[0080] The scene adaptation logic table includes:

[0081]

[0082]

[0083] Dynamic parameter optimization of the hierarchical control architecture of the self-balancing vehicle is performed. The optimization process includes:

[0084] Define the state space of the self-balancing vehicle. The states in the state space include vehicle speed, vehicle posture, current road conditions, and control inputs. The child's riding habits are used as additional state variables. The control inputs include the output of the PID controller, the vehicle posture includes angle θ and angular velocity ω, the current road conditions include wet, flat, and bumpy, and the riding habits include riding speed, perceptual demand data, and motion mode. Perceptual demand data includes:

[0085] Define the motion space of the self-balancing vehicle. The motion space includes parameter adjustments for the child's instructions in response to the PID speed control unit. The parameters include proportional gain, integral gain, and differential gain.

[0086] For example: If a child prefers "sports mode", the PID parameters will automatically increase the proportional gain.

[0087] The system continuously monitors children's riding habits and vehicle responses over time, including acceleration and deceleration frequency and mode preferences. A deep Q-network is used to learn from environmental conditions, including road conditions, vehicle speed, and body posture, to optimize the parameters of the balancing vehicle's hierarchical control architecture. The system then uses an MPC predictive model to optimize long-term strategies, including preemptive deceleration on slippery surfaces to prevent skidding. Whenever environmental conditions (such as slippery surfaces) change, the system adjusts the PID controller parameters based on real-time feedback to ensure optimal control performance.

[0088] During the trial-and-error learning process, each action is continuously evaluated, and the parameters are adjusted through the feedback of the reward function. The feedback influences the next learning, allowing the system to gradually adapt to new road conditions and improve control accuracy.

[0089] Define the reward function of the hierarchical control architecture of the balance car. The mathematical expression of the reward function is:

[0090] R=ω1X St +ω2X SC +ω3X UP -ω4X E ;

[0091] Among them, R is the reward function, ω1, ω2, ω3 and ω4 are weight coefficients used to adjust the contribution of each factor to the reward, XSt is the stability of attitude control, X SC For speed control, X UP For children's preference matching, X E For energy efficiency.

[0092] Stability specifically includes: if the vehicle body posture remains within a predetermined range, a positive reward is given;

[0093] Speed control specifically includes: giving positive rewards if the vehicle speed is close to the child’s desired speed and does not exceed the speed limit;

[0094] The child preference matching specifically includes: if the adjustment of the PID controller matches the child’s riding habits, a reward is given;

[0095] Energy efficiency specifically includes: while ensuring the driving experience, the higher the energy efficiency, the higher the reward. For example, rewards are given when the PID controller can adapt to low power consumption mode without sacrificing performance.

[0096] Combining reinforcement learning with online learning mechanisms to update PID controller parameters, using real-time data and a trial-and-error learning process, not only optimizes the control strategy but also improves the system's adaptability in different environments. Reinforcement learning continuously optimizes long-term strategies through exploration and exploitation, while online learning dynamically adjusts parameters during children's actual use, enhancing the personalized control experience.

[0097] The execution layer includes the safety assurance unit and the drive control unit:

[0098] The safety assurance unit includes a dual-mode decision-making mechanism and a fuse mechanism;

[0099] The dual-mode decision-making mechanism includes: the main controller and the backup controller work alternately, and compare the output in real time. If the difference between the main controller and the backup controller exceeds the preset difference threshold, an alarm is triggered or the backup controller is switched to.

[0100] The fuse mechanism includes: automatically switching to a backup sensor or entering a degraded mode to ensure safety when sensor data is abnormal. Abnormalities include gyroscope drift and reading failure.

[0101] The drive control unit includes a command arbitration mechanism, a real-time control mechanism, and a status feedback mechanism:

[0102] The command arbitration mechanism includes: when a child intervenes, the autonomous driving algorithm is suspended and manual commands are executed first;

[0103] The real-time control mechanism includes: using the FreeRTOS real-time operating system for task scheduling and task priority classification. Task priorities include first priority, second priority, third priority, and basic task levels. First priority tasks are executed every 10ms, second priority tasks are executed every 100ms, and third priority tasks are executed every 1s. Basic task level tasks use FPGA chips to accelerate matrix operations to perform task execution operations, such as solving the Riccati equation in LQR, to improve real-time processing capabilities.

[0104] The status feedback mechanism includes: informing children of the current control mode through voice prompts, improving the transparency of the driving experience.

[0105] By integrating multiple control algorithms and a flexible scenario switching mechanism, the intelligent decision-making layer ensures smooth vehicle operation under varying road conditions, while guaranteeing safety, comfort, and efficiency. Redundancy, real-time performance, and online learning mechanisms further enhance the system's reliability and adaptability.

[0106] The decision-making layer achieves intelligent control through multi-algorithm collaboration and scenario adaptation, ultimately achieving a safe, comfortable and efficient intelligent driving experience.

[0107] Parallel or serial computing: Depending on the control architecture design, the calculations are performed:

[0108] PID speed control: Use position PID (example parameters Kp = 120, Ki = 8, Kd = 0.5) to adjust the motor output to maintain vehicle speed and basic upright posture.

[0109] LQR Stance Optimization: If enabled, uses an LQR controller (based on state variables such as tilt error and suspension deformation, and the weight matrix Q) to calculate the optimal control force / torque to adjust the stance and suppress road disturbances. This involves solving the Riccati equation (FPGA acceleration).

[0110] MPC predictive control: If activated, based on the vehicle model and predictions for a period of time in the future (e.g., 1 second, 20 steps), optimization methods such as QP quadratic programming are used to solve the optimal control sequence (such as motor output and suspension adjustment) to achieve trajectory tracking, obstacle avoidance, or smooth driving in complex road conditions.

[0111] Fuzzy PID control: If activated, it is used to deal with nonlinear and highly uncertain scenarios (such as sandy areas, slippery roads, and sharp turns), dynamically adjust PID parameters, or directly output the control quantity.

[0112] Collaborative decision-making: The outputs of each activated algorithm are integrated or arbitrated to generate the final comprehensive control intention.

[0113] An intelligent power-assisted balancing vehicle comprises a handlebar, a vehicle frame, wheels, a child interaction layer, an intelligent decision-making layer, a motion control layer and an execution layer.

[0114] This invention integrates multiple control algorithms and a flexible scene switching mechanism to ensure the smooth operation of the vehicle under different road conditions, while guaranteeing safety, comfort and efficiency. It further improves the reliability and adaptability of the system through redundant design, real-time guarantee and online learning mechanism, ultimately achieving a safe, comfortable and efficient intelligent driving experience.

[0115] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium may be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0116] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A control system for an intelligent power-assisted balancing vehicle, characterized in that: It includes the child interaction layer, intelligent decision-making layer, motion control layer and execution layer: The child interaction layer is used to interact with children, including operations and feedback, and provides touch screen and voice command interaction methods; The intelligent decision-making layer is used to implement the core part of the intelligent control algorithm, make decisions based on the perception layer data and target instructions, and output control signals to the execution layer; The execution layer is used to realize the specific physical actions of the balancing vehicle according to the control signal and implement the safety guarantee mechanism.

2. The control system of the intelligent power-assisted balancing vehicle according to claim 1, characterized in that: The perception layer includes a posture detection unit, an environment perception unit and a data fusion unit: The attitude detection unit is used to obtain the vehicle's inclination angle and angular velocity through the MPU6050 gyroscope and accelerometer, remove noise through Kalman filtering, and output the vehicle's pitch and roll angle data in real time; The environmental perception unit is used to detect tire ground pressure distribution through a pressure sensor array, calculate pressure variance, build a road bumpiness grade model based on the pressure variance, and trigger active suspension system adjustments. It also uses a lidar to detect obstacles and obtain obstacle avoidance signals, which are transmitted to the decision-making layer via the CAN bus. The data fusion unit is used to fuse IMU and pressure sensor data through extended Kalman filtering.

3. The control system of the intelligent power-assisted balancing vehicle according to claim 2, characterized in that: The environment perception unit includes: The road surface grade of the road bump grade model is determined by threshold segmentation, and the road surface grade determination includes: Low bump: pressure variance <0.5N 2 / cm 4 ; Moderate bumps: 0.5 ≤ pressure variance < 2N 2 / cm 4 ; High turbulence: variance ≥ 2N 2 / cm 4 ; Based on the road surface grade determination, a CNN model is trained, pressure time series data is input, and road surface types are output, where the road surface types include urban roads, off-road roads, and emergency roads; When a bump is detected, a command is sent to the suspension controller via the CAN bus; The ROS robot operating system is used to receive point cloud data, extract obstacle contours, and perform DBSCAN clustering on the point cloud data into obstacle areas.

4. The control system of the intelligent power-assisted balancing vehicle according to claim 3, characterized in that: The intelligent decision-making layer is used to build a hierarchical control architecture for the self-balancing vehicle, which includes a PID speed control unit, an LQR attitude control unit, and an MPC prediction control unit: The PID speed control unit is used to adjust the vehicle speed and dynamically respond to children's instructions. The basic balance ring adopts position PID (Kp = 120, Ki = 8, Kd = 0.5), and the parameters are optimized by MATLAB genetic algorithm; The LQR attitude control unit is used to adjust the vehicle body attitude, with the vehicle body tilt angle error and suspension deformation as state variables, and the weight matrix Q = diag (10, 1, 5, 0.1); The MPC prediction control unit is used to dynamically adjust the motor output by predicting the trajectory for a period of time in the future. The rolling horizon is set to 1 second, and the solver adopts QP quadratic programming; Dynamically responding to children's instructions includes: receiving instructions from the child interaction layer, the instructions including forward, left turn, right turn, path tracking, emergency braking, cruise and obstacle avoidance, and prioritizing the instructions. The priority ranking content is safety instructions > dynamic instructions > regular instructions. The safety instructions include emergency braking and obstacle avoidance, the dynamic instructions include cruise and path tracking, and the regular instructions include forward, left turn and right turn.

5. The control system of the intelligent power-assisted balancing vehicle according to claim 4, characterized in that: The environment is matched according to the data output by the child interaction layer, and the control strategy is automatically switched. The control strategy includes: On urban roads, the PID speed control unit is used to control the balance, and the MPC prediction control unit is used to track the path. The LQR attitude control unit is used for suspension adjustment and fuzzy PID control on off-road surfaces; In an emergency, the priority channel is set for quick response, and the PID speed control unit is switched to perform emergency braking control on the balancing vehicle.

6. The control system of the intelligent power-assisted balancing vehicle according to claim 5, characterized in that: Dynamic parameter optimization is performed on the hierarchical control architecture of the self-balancing vehicle. The optimization process includes: A self-balancing vehicle state space is defined, where the states of the self-balancing vehicle state space include vehicle speed, vehicle posture, current road conditions, and control inputs, and the child's riding habits are used as additional state variables. The control inputs include the output of a PID controller, the vehicle posture includes angle θ and angular velocity ω, the current road conditions include wet, flat, and bumpy, and the riding habits include riding speed, perceptual demand data, and motion mode. The perceptual demand data includes: Defining a self-balancing vehicle motion space, the self-balancing vehicle motion space including parameter adjustment for a child's instruction in response to the PID speed control unit, the parameters including proportional gain, integral gain, and differential gain; The system continuously monitors children's riding habits and vehicle responses per unit time, including acceleration and deceleration frequency and mode preferences. Deep Q-network trial-and-error learning of environmental conditions, including road conditions, vehicle speed, and vehicle posture, is used to optimize the parameters of the self-balancing vehicle's hierarchical control architecture. The MPC prediction model is used to optimize long-term strategies, including slowing down in advance on slippery roads to avoid slipping.

7. The control system of the intelligent power-assisted balancing vehicle according to claim 6, characterized in that: In the process of trial and error learning, each action is continuously evaluated, and the parameters are adjusted through the feedback of the reward function, and the next learning is affected by the feedback; Define the reward function of the hierarchical control architecture of the self-balancing vehicle. The mathematical expression of the reward function is: R=ω1X St +ω2X SC +ω3X UP -ω4X E ; Among them, R is the reward function, ω1, ω2, ω3 and ω4 are weight coefficients used to adjust the contribution of each factor to the reward, X St is the stability of attitude control, X SC For speed control, X UP For children's preference matching, X E For energy efficiency.

8. The control system of the intelligent power-assisted balancing vehicle according to claim 7, characterized in that: The execution layer includes a safety assurance unit and a drive control unit: The safety assurance unit includes a dual-mode decision-making mechanism and a fuse mechanism; The dual-mode decision-making mechanism includes: the main controller and the backup controller work alternately, compare the outputs in real time, and if the difference between the main controller and the backup controller exceeds a preset difference threshold, trigger an alarm or switch to the backup controller; The fusing mechanism includes automatically switching to a backup sensor or entering a degraded mode when sensor data is abnormal, including gyroscope drift and reading failure.

9. The control system of the intelligent power-assisted balancing vehicle according to claim 8, characterized in that: The drive control unit includes an instruction arbitration mechanism, a real-time control mechanism, and a state feedback mechanism: The command arbitration mechanism includes: when a child intervenes, the autonomous driving algorithm is suspended and manual commands are executed first; The real-time control mechanism includes: using the FreeRTOS real-time operating system to schedule tasks and divide task priorities, wherein the task priorities include first priority, second priority, third priority and basic task level, first priority tasks are executed every 10ms, second priority tasks are executed every 100ms, and third priority tasks are executed every 1s, and tasks at the basic task level are executed by accelerating matrix operations through FPGA chips; The state feedback mechanism includes: informing the child of the current control mode through voice prompts.

10. An intelligent power-assisted balancing vehicle, comprising a handlebar, a vehicle frame, wheels, a child interaction layer, an intelligent decision-making layer, a motion control layer, and an execution layer.

Images