Full-speed-domain self-balancing control system and method for unmanned motorcycle

Abstract

The invention belongs to the technical field of unmanned control, and particularly relates to a full-speed-domain self-balancing control system and method for an unmanned motorcycle. The system comprises a balance mechanism and a motion controller, the balance mechanism comprises a rear wheel swinging mechanism and a front wheel steering mechanism; wherein the rear wheel swinging mechanism is composed of an unmanned motorcycle rear wheel swinging steering engine and a swinging connecting rod, and the rear wheel swinging steering engine is connected with the swinging connecting rod to control transverse swinging of a rear wheel; the front wheel swing mechanism is composed of a front wheel swing steering engine and a steering connecting rod, and the front wheel swing steering engine is connected with the rotary connecting rod to control rotation of front wheels of the unmanned motorcycle. The motion controller comprises a rear wheel swing control module and a front wheel steering control module; wherein the rear wheel swing control module is used for outputting a signal to control the rear wheel swing steering engine to control the swing connecting rod when the unmanned motorcycle is static or at a low speed, and the front wheel steering control module is used for outputting a signal to control the front wheel rotation steering engine to control the rotation connecting rod when the unmanned motorcycle is at a medium speed or a high speed.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of unmanned control, and particularly relates to a full-speed-domain self-balancing control system and method for an unmanned motorcycle. BACKGROUND

[0002] With the rapid development of unmanned technology, artificial intelligence technology and sensor technology, the development of intelligent unmanned systems has made great progress. Ground unmanned vehicles have become an important part of military operations due to their low casualty loss in high-risk scenarios, sustainable operations in extreme service environments, and real-time data acquisition in complex battlefield environments. Unmanned motorcycles are widely used in the transportation of wounded personnel and reconnaissance and strike tasks in complex terrain and narrow spaces due to their compact structure and flexible mobility. In addition, their low operating noise and small radar signature make them an ideal platform for covert operations, surveillance tasks, and intelligence gathering tasks. Since the two wheels are arranged on the front and rear axles, they exhibit higher flexibility and stronger climbing and obstacle crossing ability than self-balancing two-wheeled vehicles with left and right wheels. Therefore, the development of unmanned motorcycle technology is of great significance for gaining a favorable situation in war, carrying out combat support tasks, and facilitating people's travel.

[0003] However, unmanned motorcycles are essentially underactuated systems, with a strong coupling between the steering subsystem and the driving subsystem. Unmodeled dynamics and external disturbances make the balance and motion control of unmanned motorcycles extremely challenging. Existing self-balancing control methods for unmanned motorcycles are not effective at low speeds or when stationary. SUMMARY

[0004] Therefore, the present application provides a full-speed-domain self-balancing control system and method for an unmanned motorcycle, which realizes automatic balancing at different speeds through rear wheel swinging, optimizes the universality and versatility of the unmanned motorcycle, and is suitable for more complex terrains and different situations, such as carrying objects.

[0005] The technical solutions of the present application are as follows:

[0006] In a first aspect, the present application provides a full-speed-domain self-balancing control system for an unmanned motorcycle, comprising a balancing mechanism and a motion controller, wherein

[0007] The balancing mechanism comprises a rear wheel swinging mechanism and a front wheel steering mechanism. The rear wheel swinging mechanism is composed of a rear wheel swinging servo and a swinging link, and the rear wheel swinging servo is connected to the swinging link to control the lateral swinging of the rear wheel. The front wheel swinging mechanism is composed of a front wheel swinging servo and a steering link, and the front wheel swinging servo is connected to the steering link to control the rotation of the front wheel of the unmanned motorcycle.

[0008] The motion controller comprises a rear wheel swing control module and a front wheel steering control module; the rear wheel swing control module is used for outputting signals to control the swing linkage of the rear wheel swing servo at the static or low speed of the unmanned motorcycle; and the front wheel steering control module is used for outputting signals to control the rotation linkage of the front wheel steering servo at the medium or high speed of the unmanned motorcycle.

[0009] Optionally, the swing linkage mainly comprises a driving linkage, a driven linkage and a connecting plate, the rear wheel swing servo drives the driving linkage, the driving linkage drives the two symmetric driven linkages through the connecting plate, thereby forming a parallelogram mechanism with three parallel linkages, and swing is realized.

[0010] Optionally, the motion controller further comprises a control strategy switching management module, and the switching logic is as follows: if v<=V tor_MIN , the rear wheel swing control mode is entered; if v>=V tor_MAX , the front wheel steering control mode is entered; if V tor_MIN <v<V tor_MAX , the fuzzy inference is used to determine which control module should be used currently. tor_MIN and V tor_MAX are selected according to the vehicle structure such as the weight, the height of the center of mass, the position of the center of mass and the trailing distance configuration of the unmanned motorcycle, and the allowable values thereof are determined according to the actual situation.

[0011] Optionally, the fuzzy inference outputs a control mode confidence S belonging to [0, 1] according to the vehicle speed and acceleration, and determines whether to switch or to perform hybrid output according to the confidence.

[0012] Contribution of the vehicle speed to S:

[0013]

[0014] Contribution of the acceleration to S:

[0015]

[0016] wherein a min and a max are the minimum acceleration and the maximum acceleration of the bicycle respectively, and x represents the current acceleration value (current instantaneous acceleration) of the vehicle.

[0017] Confidence calculation combining the vehicle speed and the acceleration:

[0018] S = alpha * S v + (1-alpha) * S a

[0019] wherein alpha is a weighting coefficient, and the influence degree of the vehicle speed and the acceleration on S can be adjusted by adjusting alpha.

[0020] Hybrid control strategy:

[0021] u mix = (1-S) · u SMC + S · u NMPC+NTSM .

[0022] wherein μ mix represents the final hybrid control output signal; μ SCM represents the control signal output by the rear wheel swing control mode, i.e. the horizontal swing angle τ(t) of the rear wheel rudder; μ NMPC+NTSM represents the control signal output by the front wheel steering control mode (two-stage control of NMPC+NTSM), i.e. the rudder angle δ(t) of the front wheel rudder.

[0023] Optionally, the rear wheel swing control module of the present application calculates a sliding surface according to the control state variables, and calculates a saturation function for suppressing chattering according to the sliding surface, and calculates a control output torque to the rear wheel swing rudder based on the saturation function.

[0024] Optionally, the control state variables of the present application: vehicle body lateral inclination angle θ, inclination angle angular velocity vehicle body deflection angle φ, deflection angle change rate

[0025] Sliding surface definition: wherein α i is a positive weight coefficient, i=1,2,3,4;

[0026] Control law form: τ(t)=-K·sat(s)

[0027] wherein τ(t) is the control output torque, used to drive the rear wheel to yaw; K is the control gain; sat is the saturation function:

[0028]

[0029] wherein ε>0 is the boundary layer thickness parameter.

[0030] Optionally, the front wheel rotation control module of the present application is realized by a two-stage composite control composed of a nonlinear model predictive controller and a non-singular terminal sliding mode controller.

[0031] Optionally, the objective function and constraint conditions of the nonlinear model predictive controller of the present application are:

[0032] Let the state vector of the vehicle at the kth sampling time be:

[0033]

[0034] wherein θ kis the roll angle of the vehicle; is the roll angular velocity; k is the forward velocity; k is the front wheel steering angle; k is the lateral position error; k is the heading angle;

[0035] The state of the vehicle evolves at each step according to the nonlinear difference dynamics model:

[0036] x k+1 = f(x k , δ k )

[0037] where f() is the nonlinear discrete dynamics model of the vehicle, which can be obtained from the model of the unmanned motorcycle or measured data;

[0038] The front wheel steering control module constructs the following weighted cost function in the prediction horizon of length N:

[0039]

[0040] where w1-w4 are the terminal attitude accuracy weights; θ N is the roll angle at the end time (k=N).

[0041] Control constraints: rudder angle amplitude limit: δ min ≤ δ k ≤ δ max ; rudder angle change rate limit:

[0042] | δ k - δ k-1 | ≤ Δ max ; attitude roll angle limit: | θ k | ≤ θ safe ;

[0043] Based on the objective function, solve the finite time domain nonlinear optimization problem with constraints:

[0044]

[0045] s.t. x k+1 = f(x k , δ k ), x0 known

[0046] δ k ∈ [ δ min , δ max ]

[0047] | δ k - δ k-1 | ≤ Δ max

[0048] |θ k |≤θ safe

[0049] This optimization problem is solved in each control cycle to obtain the optimal rudder angle input sequence:

[0050]

[0051] Only the first control variable in the sequence will be used as the instruction to be executed in the current cycle.

[0052]

[0053] The remaining part As the initial value for the next optimization cycle;

[0054] Optionally, the control function of the non-singular terminal sliding mode controller of the present invention is:

[0055] Let the system reference input be δref and the actual control quantity be δ, then the tracking error is:

[0056] e = δ - δ ref

[0057] Construct the non-singular terminal sliding surface function s as follows:

[0058]

[0059] Where β>0 is the design gain; p, q are positive odd numbers;

[0060] The control law is designed as follows:

[0061]

[0062] in, The reference acceleration is η; η>0 is the sliding mode gain; and sat(s) is the saturation function.

[0063]

[0064] Where φ>0 is the saturation bandwidth parameter.

[0065] Secondly, the present invention provides a self-balancing control method for an unmanned motorcycle across the entire speed range, the specific process of which is as follows:

[0066] Perception and Mapping: Online mapping of the environment in which the unmanned motorcycle is located and estimation of the vehicle's own pose;

[0067] Path generation: On the established map, the line segment connecting the current pose and the target position is used as the path optimization area, and the planned path is represented by a cubic B-spline curve. The path is then optimized using a multi-objective evolutionary optimization method.

[0068] Minimum time trajectory generation: The geometric path is further mapped to a time trajectory. The goal of time trajectory optimization is to minimize the overall travel time and meet the set constraints. At each time point, a recommended tilt angle is calculated as a feedforward input to assist in adjusting the vehicle's lateral attitude.

[0069] Tracking control and zone coordination: Based on the current vehicle speed, rear wheel sway control and / or front wheel steering control are used to complete path tracking and lateral balance.

[0070] Beneficial effects:

[0071] This invention proposes a full-speed-range self-balancing control system and method for unmanned motorcycles. By introducing a rear wheel lateral oscillation structure and a front wheel full-speed-range tracking mechanism, it overcomes the technical bottlenecks of traditional unmanned motorcycles that cannot maintain balance at low speeds or when stationary, and that lack accuracy and dynamic response in path tracking. It achieves continuous self-balancing and path tracking control throughout the entire process from standstill to high-speed travel, and has the following significant technical effects:

[0072] First, the ability to self-balance and track across the entire speed range.

[0073] This invention introduces a full-speed-domain composite control strategy of "rear wheel attitude compensation + front wheel steering tracking" into the control system of an unmanned motorcycle. In the stationary and low-speed phases, sliding mode control drives the rear wheel yaw to maintain tilt balance. In the medium- and high-speed phases, a combination of nonlinear model predictive control (NMPC) and non-singular terminal sliding mode (NTSM) control is used to drive the front wheel for high-precision path tracking and lateral balance adjustment, achieving stable switching and precise control from zero-speed start to high-speed motion. Path tracking and balance control are decoupled in the control architecture, achieving dynamic coordination through speed-segmented weight adjustment, significantly expanding the speed range applicable to the unmanned motorcycle.

[0074] Second, a compact and highly efficient rear wheel swing mechanism.

[0075] The rear fork employs a parallelogram three-link swing structure, balancing structural rigidity with lateral swing freedom. Without increasing the vehicle's overall dimensions, it achieves low-energy, high-precision rear wheel micro-swing motion for self-balancing support at low speeds. This structure is particularly suitable for miniaturized autonomous vehicle platforms where high mobility, stability, and structural integrability are required, demonstrating excellent engineering feasibility.

[0076] Third, the intelligent fuzzy switching mechanism ensures control stability.

[0077] The controller constructs a fuzzy logic judgment system based on multi-source perception indicators such as speed, attitude, and lateral error. It dynamically adjusts the path tracking weight and attitude control weight according to the confidence level fusion method, replacing the traditional on / off switching mode of the controller. This mechanism effectively avoids frequent control jumps in the speed critical zone, improves the system's robustness to dynamic disturbances, sudden terrain changes, and sensing errors, and enhances the smoothness and safety of the vehicle's ride.

[0078] Fourth, path planning and control work together to improve the overall performance of the system.

[0079] This invention establishes a path planning system oriented towards vehicle physical constraints. Through cubic B-spline modeling and a multi-objective evolutionary algorithm, it comprehensively optimizes path length, curvature smoothness, and obstacle safety distance to generate an analytical path with second-order continuous curvature. In the trajectory generation stage, constraints such as tilt angle, rudder angle, and lateral acceleration are further introduced to construct a minimum time trajectory and feedforward attitude sequence, effectively improving path executability and dynamic control accuracy. Based on this, the control layer constructs a "balance-tracking" dual-objective optimization framework to achieve coordinated optimization of attitude stability, reasonable energy consumption, and path tracking error, ensuring the system's overall operational performance in complex dynamic environments.

[0080] Fifth, the entire system is embedded, making it suitable for integrated engineering applications.

[0081] This system, including its perception mapping, path planning, trajectory generation, and control algorithms, has been tested and run on resource-constrained embedded platforms (such as STM32F4). The overall operating frequency can reach 10–200Hz, featuring low power consumption, low latency, and high responsiveness. System deployment does not rely on GPUs or high-performance industrial computers, making it suitable for batch integration in various application scenarios such as reconnaissance robots, indoor delivery, material handling, and cycling assistance. It demonstrates excellent engineering feasibility and promotional value. Attached Figure Description

[0082] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0083] Figure 1 This is a side view of an unmanned motorcycle.

[0084] Figure 2 A top-down view of an unmanned motorcycle;

[0085] Figure 3(a) is a right view of the unmanned motorcycle;

[0086] Figure 3(b) shows a parallelogram mechanism with three parallel links;

[0087] Figure 4 For control flowchart;

[0088] Figure 5 This is a schematic diagram of the upper-level sensing system;

[0089] Figure 6 This is a flowchart of the path planning and navigation control process for unmanned motorcycles. Detailed Implementation

[0090] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0091] It should be noted that, in the absence of conflict, the following embodiments and features can be combined with each other; and, based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0092] It should be noted that various aspects of the embodiments described below are within the scope of the appended claims. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using other structures and / or functionalities besides one or more of the aspects set forth herein.

[0093] This application provides a self-balancing control system for an unmanned motorcycle across the entire speed range, including a balancing mechanism and a motion controller;

[0094] The balancing mechanism includes a rear wheel swing mechanism and a front wheel steering mechanism; wherein, the rear wheel swing mechanism consists of a rear wheel swing servo motor and a swing linkage of the unmanned motorcycle, the rear wheel swing servo motor being connected to the swing linkage to control the lateral swing of the rear wheel; the front wheel swing mechanism consists of a front wheel swing servo motor and a steering linkage, the front wheel swing servo motor being connected to the rotation linkage to control the rotation of the front wheel of the unmanned motorcycle;

[0095] The motion controller includes a rear wheel sway control module and a front wheel steering control module; wherein, the rear wheel sway control module is used to output a signal to control the rear wheel sway servo to control the sway link when the speed of the unmanned motorcycle is less than a set threshold, and the front wheel steering control module is used to output a signal to control the front wheel rotation servo to control the rotation link when the speed of the unmanned motorcycle is greater than or equal to a set threshold.

[0096] Furthermore, the balancing mechanism in this embodiment specifically includes: a front wheel steering mechanism 2 including a front wheel rotation servo 6 and a steering linkage 7, wherein the front wheel rotation servo 6 drives the steering linkage 7 to realize the front wheel steering function; and a rear wheel swing structure including a rear wheel swing servo 5 and a swing linkage composed of an active linkage, a driven linkage and a connecting plate 1. The rear wheel swing servo 5 drives the active linkage at the rear fork, and the active linkage drives the symmetrical driven linkages on both sides through the connecting plate 1, thereby forming a parallelogram mechanism with three parallel linkages, which is highlighted in red in Figure 3(b). Figure 3(b) is obtained by hiding the front and rear wheels, the drive motor and the rear wheel connection through the A view derived from Figure 3(a). The swing function of the rear wheel is realized through this parallelogram mechanism.

[0097] The balancing mechanism also includes a loading platform 3 and a drive motor 4; wherein...

[0098] The drive motor 4 directly drives the rear wheel. Since the drive motor 4 is placed on one side of the rear wheel, a counterweight is added to the other side of the rear wheel. The mass and geometry of the counterweight are designed through static and dynamic balance calculations.

[0099] Loading platform 3 integrates a sensing system and an embedded computing system to perceive the vehicle's angle and control various servo motors and drive motors. Furthermore, loading platform 3 can be equipped with expansion modules to adapt to different application scenarios, such as a reconnaissance system for environmental perception and monitoring, or a storage module for storing and transporting goods. Its structural diagram is shown below. Figure 1 , Figure 2 And Figure 3.

[0100] Furthermore, the motion controller described in this application embodiment performs composite motion control of the unmanned motorcycle based on speed perception. This means it can adaptively select different control mechanisms under different motion states (such as stationary, low speed, and medium-to-high speed driving) to achieve continuous and stable lateral balance control of the vehicle. In specific implementations, this motion controller integrates advanced control algorithms such as sliding mode control (SMC), nonlinear model predictive control (NMPC), and non-singular terminal sliding mode control (NTSM), and introduces a fuzzy logic-enhanced control strategy switching mechanism, which can effectively improve the vehicle's attitude stability, response speed, and anti-interference capability in multiple scenarios.

[0101] In this embodiment, the motion controller mainly includes the following four functional modules: rear wheel sway control module (suitable for static / low-speed balance control), front wheel steering control module (suitable for medium- and high-speed balance control), and control strategy switching management module (based on fuzzy logic judgment).

[0102] This controller, with an STM32 central controller at its core, is responsible for real-time acquisition of various sensor data, determining the current operating mode, executing corresponding control algorithms, and outputting the calculation results to the appropriate actuators. The controller consists of the following components:

[0103] Sensor module: Attitude sensor (e.g., MPU6050): acquires the vehicle's lateral tilt angle θ and angular velocity. Vehicle body deflection angle φ and its angular velocity Wheel speed sensor / encoder: Real-time monitoring of vehicle linear velocity v;

[0104] Control calculation module: rear wheel sway control module (sliding mode control), front wheel steering control module (NMPC+NTSM), control strategy switching management module (based on fuzzy logic judgment);

[0105] The controller controls the following actuators: the rear wheel yaw servo, which controls the rear wheel lateral yaw angle φ; the front wheel steering motor, which adjusts the front wheel steering angle δ; and the rear wheel drive motor, which propels the vehicle forward.

[0106] The controller adaptively selects the control mode based on real-time speed and attitude status, and sends the control output signal to the actuator to achieve closed-loop self-balancing control.

[0107] Furthermore, the rear wheel sway control module (stationary and low-speed control):

[0108] When the vehicle is stationary or at low speed, due to its small inertia, it is difficult to maintain balance by steering. Therefore, a rear wheel yaw control strategy is adopted, which changes the direction of the support reaction force by slightly lateral yaw of the rear wheels, thereby achieving compensation control of the vehicle body tilt. This has almost no significant impact on the driving path, and the path deviation is negligible.

[0109] The rear wheel sway control module employs a sliding mode control (SMC) algorithm for stationary or low-speed states. This algorithm has advantages such as strong robustness, insensitivity to modeling errors, and fast response speed, making it suitable for low-speed nonlinear system control.

[0110] Control state variables: vehicle body lateral tilt angle θ, tilt angle angular velocity Vehicle body deflection angle φ, rate of change of deflection angle

[0111] Sliding surface definition:

[0112]

[0113] Where, α i The weighting coefficients are positive, i = 1, 2, 3, 4.

[0114] Control law form:

[0115] τ(t)=-K·sat(s)

[0116] Where τ(t) is the control output torque used to drive rear wheel yaw; K is the control gain; and sat is the saturation function (to suppress chatter).

[0117]

[0118] Where ε>0 is the boundary layer thickness parameter (also known as the chattering buffer bandwidth).

[0119] The controller outputs a control signal τ(t) to the rear wheel oscillation servo 5. The rear wheel oscillation servo 5 controls the oscillation linkage to adjust the rear wheel yaw angle, thereby achieving vehicle center of gravity reconstruction and rapid balance.

[0120] Furthermore, the front wheel rotation control module (nonlinear model prediction and nonsingular end slip control)

[0121] When the vehicle speed exceeds a set threshold, the system enters a medium-to-high-speed control mode. During this phase, the vehicle retains a certain amount of inertia, and any intervention from the lateral sway of the rear wheels can cause a lateral shift in the vehicle's center of gravity, resulting in a significant deviation from the path and affecting the vehicle's accuracy and stability. Under these conditions, continuing to use rear wheel sway may increase the uncertainty of path control and even induce system oscillations. Therefore, in medium-to-high-speed mode, the motion controller will prioritize front-wheel steering control to achieve lateral balance and path adjustment, ensuring the stability and controllability of the vehicle's trajectory.

[0122] To balance path accuracy, rudder angle smoothness, and tilt stability, this embodiment designs a two-stage composite control structure consisting of a nonlinear model predictive controller (NMPC) and a nonsingular terminal sliding mode controller (NTSM) to meet the requirements of lateral stability and attitude control of unmanned motorcycles in medium- and high-speed driving conditions, thereby achieving high-response and high-robust control during high-speed driving.

[0123] Outer-loop control: Nonlinear Model Predictive Control (NMPC)

[0124] Nonlinear model predictive control is based on the nonlinear discrete dynamic model of the vehicle. In each control cycle, it continuously optimizes the front wheel steering angle input sequence for a future period and outputs the first control command δref to achieve optimized control of the front wheel steering behavior.

[0125] (1) State modeling

[0126] Let the state vector of the vehicle at the k-th sampling time be:

[0127]

[0128] Where: θk The vehicle body camber angle; v is the tilt angle and angular velocity. k Forward velocity; δ k The front wheel steering angle; y k For lateral positional deviation; φ k This is the heading angle.

[0129] The vehicle's state evolves at each step according to the nonlinear differential dynamics model as follows:

[0130] x k+1 =f(x) k δ k )

[0131] Where f() is the nonlinear discrete dynamics model of the vehicle, which can be obtained from unmanned motorcycle models or measured modeling data.

[0132] (2) Optimize the objective function

[0133] To ensure optimal overall performance of the control system in terms of stability, energy consumption, and actuator load, the controller constructs the following weighted cost function in the prediction time domain of length N:

[0134]

[0135] Where: w1 is the attitude stability weight; w2 is the rudder angle energy consumption / amplitude weight; w3 is the control smoothness weight; w4 is the terminal attitude accuracy weight; (δ k -δ k-1 ) 2 To penalize changes in rudder angle and improve handling smoothness.

[0136] (3) Control constraints

[0137] To ensure the physical feasibility and safety of the control behavior, the following control and state constraints are introduced: (1) Rudder angle amplitude limit: δ min ≤δ k ≤δ max (2) Rudder angle change rate limit: |δ k -δ k-1 |≤Δ max (3) Attitude tilt angle limitation: |θ k |≤θ safe Optional restrictions: lateral acceleration, trajectory boundaries, actuator response limits, etc.

[0138] (4) Solving for optimal control

[0139] Based on the above model and objective function, the following finite-time nonlinear constrained optimization problem is constructed:

[0140]

[0141] st x k+1 =f(x) k ,δ k x0 is known.

[0142] δ k ∈[δ min ,δ max ]

[0143] |δ k -δ k-1 |≤△ max

[0144] |θ k |≤θ safe

[0145] This optimization problem is solved within each control cycle using numerical algorithms such as the interior-point method (IPOPT), sequential quadratic programming (SQP), or a symbolic optimization solver (e.g., CaSADi) to obtain the optimal rudder angle input sequence.

[0146]

[0147] (5) Rolling execution and closed-loop control

[0148] Only the first control variable in the sequence will be used as the instruction to be executed in the current cycle.

[0149]

[0150] The remaining part It can be used as the initial value (Warm Start) for the next optimization cycle to improve computational efficiency and form a closed-loop rolling optimization control strategy. This method has good dynamic adaptability, constraint handling ability and stability, and is suitable for fine-grained control of the path and attitude of unmanned motorcycles at high speeds.

[0151] Inner loop control: Non-singular terminal sliding mode control (NTSM)

[0152] To ensure that the optimal reference rudder angle δref output by the outer loop NMPC controller can be tracked by the actual front wheel rudder angle δ of the vehicle with high precision and high speed, and to maintain system stability and response performance under conditions of actuator non-ideal, modeling error or external disturbance, a non-singular terminal sliding mode controller (NTSM) is designed as the inner loop servo controller.

[0153] The controller constructs a rudder angle error feedback loop, which has finite-time convergence, strong robustness, and anti-chattering capability. The specific design process is as follows:

[0154] (1) Definition of error variable

[0155] Let the system reference input be δref and the actual control quantity be δ. Then the tracking error is:

[0156] e = δ - δ ref

[0157] Its derivative is:

[0158]

[0159] (2) Sliding mode surface design

[0160] Construct a non - singular terminal sliding mode surface function s as:

[0161]

[0162] where: β > 0 is the design gain; p, q are positive odd numbers, satisfying 0 < p / q < 1, ensuring that the convergence rate of the sliding mode surface is enhanced when the error approaches zero; the exponential term e p / q constitutes the terminal term, ensuring that the error converges within a finite time; this design avoids the problem of singularity of the control law when e → 0 in the classical terminal sliding mode.

[0163] (3) Reaching law design

[0164] To ensure that the system state can quickly and effectively approach the sliding mode surface s(t) when there is an initial deviation, the following exponential reaching law is designed for this controller:

[0165]

[0166] where: k1 > 0 is the linear reaching gain, adjusting the exponential decay rate of the sliding mode surface; k2 > 0 is the constant disturbance suppression term, enhancing the robustness of the system to uncertainties; sat(s) is the saturation function, defined as:

[0167]

[0168] where, φ > is the saturation bandwidth parameter.

[0169] This reaching law can ensure that when the system state has not entered the sliding mode surface (i.e., s ≠ 0), the state trajectory approaches the sliding mode surface at an exponential rate, ensuring the control accuracy and response speed of the system.

[0170] (4) Control law design

[0171] Considering the dynamics of the front - wheel steering motor or servo - actuator is approximated as an inertial system:

[0172]

[0173] Where J is the equivalent steering inertia and τ is the control torque. The objective is to design a control law that keeps the system invariant at the sliding surface s=0, thereby achieving error convergence.

[0174] The control law is selected as follows:

[0175]

[0176] in: The reference acceleration is η (which can be estimated through numerical differentiation); η>0 is the sliding mode gain, which controls the convergence speed; sat(s) is the saturation function, used to suppress high-frequency chattering.

[0177]

[0178] Where φ>0 is the saturation bandwidth parameter.

[0179] Control characteristics description:

[0180] Finite-time convergence: The control law uses a terminal exponent term to ensure that the error e→0 converges within a finite time. Non-singular design: Avoids exponential singularity (e.g., e→0) when e→0. -1 This avoids control divergence issues. It boasts strong robustness: sliding mode control is inherently robust to modeling errors and disturbances. It also exhibits strong anti-chattering capability: the saturation function replaces the sign function, effectively reducing high-frequency jitter and adapting to motor actuators.

[0181] Overall control logic based on outer loop control and inner loop control:

[0182] The NMPC+NTSM two-level control structure operates as follows: the outer-loop NMPC controller outputs δ_ref at a lower frequency (e.g., 50Hz) for long-term global optimization, while the inner-loop NTSM controller performs short-term fast tracking compensation of δ_ref at a higher frequency (e.g., 200Hz). Working together, they balance global constraint coordination capabilities with actuator fast response robustness, achieving excellent performance in medium- and high-speed path tracking and attitude control.

[0183] Furthermore, the control strategy switching management module (fuzzy logic enhanced judgment) in this application embodiment:

[0184] To avoid the critical speed point (such as V) tor_MIN –V tor_MAX Frequent controller switching causes system jitter. This system is equipped with a state switching mechanism enhanced by fuzzy logic.

[0185] Switching logic:

[0186] If v≤V tor_MIN Enter rear wheel sway control mode;

[0187] If v≥V tor_MAX Enter front wheel steering control mode;

[0188] If V tor_MIN <v<V tor_MAX A fuzzy inference system is used to determine the control method that should be used at the current time.

[0189] The fuzzy inference system takes vehicle speed, acceleration, and tilt angle fluctuation amplitude as inputs, and outputs a control mode confidence score S∈[0,1]. The system determines whether to switch modes or perform a mixed output based on this value.

[0190] The contribution of vehicle speed to S:

[0191]

[0192] The contribution of acceleration to S:

[0193]

[0194] Among them, a min and a max These are the minimum and maximum accelerations of the bicycle, respectively.

[0195] Calculation of confidence levels combining vehicle speed and acceleration:

[0196] S=α·S v +(1-α)·S a

[0197] Here, α is a weighting coefficient, and by adjusting α, the degree of influence of vehicle speed and acceleration on S can be adjusted.

[0198] Hybrid control strategy:

[0199] u mix =(1-S)·u SMC +S·u NMPC+NTSM

[0200] This approach enables a smooth transition between the two controllers, avoiding instability caused by sudden changes in the controller.

[0201] Furthermore, embodiments of this application also introduce a multi-objective control optimization mechanism.

[0202] To enhance the overall performance of the control system, this embodiment introduces a multi-objective optimization mechanism into the controller design. Attitude stability, energy consumption, actuator load, and control smoothness are incorporated into the control objective function. The controller automatically weighs these objectives and outputs the optimal control action. The main control objectives include: minimizing attitude error (improving stability); smoothing control input (reducing mechanical shock); low energy consumption (extending range); small path deviation (improving driving accuracy); and reasonable control output constraints (ensuring actuator lifespan). This multi-objective optimization structure is particularly suitable for the NMPC+NTSM control framework, improving overall operating efficiency and system stability.

[0203] Based on the above embodiments, the control process and system characteristics of the unmanned motorcycle are as follows:

[0204] The control process is as follows: Real-time acquisition of speed and attitude sensor data determines the current vehicle speed range. If it is in a transition zone, fuzzy logic is used to determine the switching intention, and the corresponding controller (SMC or NMPC) is executed to calculate the control quantity. The control output is transmitted to the rear wheel yaw servo or front wheel steering servo, and the control parameters are corrected in real time, operating in a closed loop. The flowchart is as follows: Figure 4 As shown.

[0205] System characteristics: Adaptable to full-speed domain self-balancing requirements; nonlinear control algorithm improves system response and robustness; fuzzy enhancement switching mechanism improves system continuity; multi-objective control optimization structure enhances system operating efficiency and engineering practicality; can run in real time on microcontroller platform and has good deployability.

[0206] Furthermore, the unmanned motorcycle full-speed-domain self-balancing control system of this application possesses capabilities such as autonomous mapping, physical feasible path generation, minimum time trajectory optimization, and multi-mode control closed loop. It can achieve dynamic obstacle avoidance and path tracking in unknown and complex environments, while also considering vehicle attitude stability and actuator constraints. The overall structure of this unmanned motorcycle is a four-layer serial mode: perception-path-trajectory-control. Each layer achieves real-time coordination and closed-loop response through interface information flow.

[0207] The upper-level sensing system has the following structure: Figure 5 As shown, it is mainly used for external environment perception, path planning, and obstacle avoidance tasks. The lidar 18 can generate environmental grid point clouds in real time and work in conjunction with the binocular depth camera 19 to obtain the distance and position information of obstacles. To achieve all-round environmental observation, the observation and reconnaissance system is equipped with two stepper motors 11 and 21. The output shaft of the stepper motor 11 is connected to the rotating platform 15 through the motor flange 14, and the rotating platform is installed in the bearing 13 and bearing seat 12 to drive the camera 19 to achieve horizontal rotation. A bracket 16 is installed on the rotating platform, and a pitch cradle 20 is installed on the inner side of the bracket. The stepper motor 21 works with the cradle 20 to adjust the camera pitch angle.

[0208] The data flow for path planning is as follows: LiDAR 18 continuously scans the surrounding environment, generating obstacle point clouds and local grid maps; the lower-level computer 17 receives and processes the radar data via the communication bus, combining it with images and depth information acquired by the binocular depth camera to provide basic data support for target recognition, path planning, and obstacle avoidance judgment. Simultaneously, the lower-level computer 17 precisely controls the stepper motor's movement based on preset logic and real-time sensor feedback, achieving dynamic adjustment of the observation angle, thereby assisting in completing reconnaissance and path analysis tasks.

[0209] The embodiment of this application is based on the above-mentioned control system and the unmanned motorcycle full-speed-domain self-balancing control method, the specific process of which is as follows:

[0210] Perception and Mapping: Online mapping of the environment in which the unmanned motorcycle is located and estimation of the vehicle's own pose;

[0211] Path generation: On the established map, the line segment connecting the current pose and the target position is used as the path optimization area, and the planned path is represented by a cubic B-spline curve. The path is then optimized using a multi-objective evolutionary optimization method.

[0212] Minimum time trajectory generation: The geometric path is further mapped to a time trajectory. The goal of time trajectory optimization is to minimize the overall travel time and meet the set constraints. At each time point, a recommended tilt angle is calculated as a feedforward input to assist in adjusting the vehicle's lateral attitude.

[0213] Tracking control and zone coordination: Based on the current vehicle speed, rear wheel sway control and / or front wheel steering control are used to complete path tracking and lateral balance.

[0214] The above process will be explained in detail below:

[0215] (1) Perception and Layering

[0216] This layer enables online mapping of the environment and vehicle pose estimation, which is a prerequisite for path planning. The vehicle's perception modules include: a 360° LiDAR (LDS-01, providing real-time point cloud data); an encoder (to obtain linear velocity and mileage); a six-axis IMU (for attitude angular velocity); and an optional UWB positioning base station (providing absolute distance information). After multi-source data input, the system uses an improved Rao-Blackwellized particle filter SLAM algorithm, whose core structure decouples map state from trajectory state as follows:

[0217] p(q 1：t m t |l 1：t o 1：t )=p(m t |q 1:t , l1：t )·p(q 1：t |l 1：t o 1：t )

[0218] Where q1:t represents the vehicle trajectory, mt represents the map grid, and l1:t and o1:t represent the observation and action sequences, respectively. By updating the pose-map pair in parallel within the particle ensemble, the computational coupling and error accumulation risks can be effectively reduced, thereby improving map construction accuracy.

[0219] The final output local raster map has centimeter-level resolution and covers an area of ​​3 to 5 meters in front, providing high-precision obstacle constraint information for subsequent path planning.

[0220] (2) Path generation layer

[0221] On the local map provided by the mapping module, the system uses the line segment connecting the current pose and the target position as the path optimization area, and represents the planned path as a cubic B-spline curve:

[0222]

[0223] Where Pi is the control vertex and Bi(s) is the spline basis function. This path model possesses second-order curvature continuity, which is beneficial for smooth execution of subsequent rudder angles.

[0224] The path optimization employs a multi-objective evolutionary optimization (MOOA) method, with the objective function comprehensively considering: 1. Total path length: 2. Maximum curvature smoothness: max s |κ(s)|;3. Obstacle safety boundary: minimum distance min s d obs (s).

[0225] The multi-objective objective function is combined with weights w1 and w2 to form the overall fitness function:

[0226]

[0227] The search strategy employs a genetic evolutionary process involving elite preservation, single-point crossover, and displacement mutation to ensure that the optimal path simultaneously satisfies both smoothness and obstacle avoidance capabilities.

[0228] (3) Minimum Time Trajectory Generation Layer

[0229] After path planning, to meet the requirements of real-time invocation by the control layer, the system further maps the geometric path P(s) to a time trajectory T(t), where each discrete time point contains information such as position, attitude, velocity, and curvature:

[0230] T(t)={x(t),y(t),φ(t),v(t),κ(t)}

[0231] The objective of trajectory optimization is to minimize the overall travel time while satisfying the following inequality constraints: 1. Inclination angle constraint: |θ(t)|≤θ max 2. Rudder angle limit: |δ(t)|≤δ max , 3. Lateral acceleration constraint: |a y (t)|=v 2 ·κ(t)≤a y，max .

[0232] The optimization process employs a quasi-convex approximation approach. Initially, path parameters are allocated at a constant speed, and then trajectory time-scale scaling and speed adjustment are performed based on the minimum time objective. The final trajectory is reconstructed into an analytical function using B-spline interpolation, facilitating high-frequency reading by the controller.

[0233] Meanwhile, the system calculates the recommended tilt angle as the feedforward input at each time point:

[0234]

[0235] This value reflects the vehicle's physical equilibrium state at the current speed and path curvature, helping the control layer improve dynamic stability. The recommended tilt angle θ... ref (t) is input as an attitude control reference signal to the rear wheel sliding mode controller (SMC) or the front wheel two-stage control structure (NMPC+NTSM) in the motion controller to assist in adjusting the lateral attitude of the vehicle and improve the response speed and anti-disturbance stability when turning.

[0236] (4) Tracking control and zonal coordination

[0237] The path tracking and lateral balance are achieved by rear wheel oscillation and front wheel steering across the entire speed range, using the motion control system described above.

[0238] (5) Path replanning mechanism

[0239] When the system detects that any of the following conditions are met: 1- A new obstacle intrudes into the planned path; 2- The current rudder angle is approaching its limit; 3- The lateral deviation continues to exceed the set threshold; the replanning flag replan_needed = true is triggered. The system will pause the current path queue, re-invoke the path optimization and trajectory module to generate a new local path, and smoothly connect it to the original trajectory flow without stopping.

[0240] (6) System robustness and security strategy

[0241] To cope with uncertainties such as non-ideal terrain and disturbing winds, the system incorporates multi-level control strategies:

[0242] First, if the rudder angle remains within the boundary value range for an extended period, the controller reduces the lateral error weight to prevent the actuator from reaching its limit. Second, if a sudden obstacle cuts into the middle of the path, a rapid local path recalculation is triggered. Third, if rapid tilting is detected (e.g., due to crosswinds or slippery conditions), a forced switch to low-speed control mode is initiated, with the rear wheels maintaining stability through swaying. This strategy ensures that the vehicle's balance is maintained and navigation tasks are successfully completed even in unexpected situations.

[0243] In summary, within the "perception-path-trajectory-control" closed loop, the vehicle first utilizes LiDAR, encoders, and IMU to run an improved RBPF-SLAM within the perception and mapping module, outputting a real-time centimeter-level local grid map. Subsequently, the path generation module uses cubic B-splines as the curve model and, through an elite-preserving multi-objective evolutionary algorithm, obtains the shortest path with second-order continuous curvature while ensuring a safe distance from obstacles. This path is then fed into the minimum time trajectory generation module: under constraints of tilt angle, rudder angle, and lateral acceleration, the system performs quasi-convex optimization on the velocity-curvature relationship and adds a feedforward tilt angle to obtain an analytical time-trajectory sequence. Next, tracking control and zone coordination uniformly employ an "outer loop NMPC + inner loop NTSM" front-wheel steering strategy and a rear-wheel yaw strategy. If a new obstacle, rudder angle limit, or lateral deviation exceeding the threshold is detected during operation, the replanning mechanism immediately returns to the path generation layer for local recalculation; otherwise, it directly outputs the rudder angle torque for closed-loop execution. The parallel system robustness and safety strategy on the right dynamically adjusts weights, rapidly reduces speed, or switches to mitigate gains to ensure safe operation under external disturbances or extreme conditions.

[0244] This invention provides a self-balancing unmanned motorcycle configuration, control, and path planning method that utilizes front-wheel steering for path finding across the entire speed range, while the rear wheel handles balance compensation at low speeds. A parallelogram structure consisting of three links is introduced at the rear fork to control rear wheel oscillation, further enhancing the bicycle's balance control. The drive system uses a rear-mounted motor to directly drive the rear wheel, and the handlebars employ a servo motor via a four-bar linkage for steering. A platform is equipped with a sensing system and an embedded computing system, with additional expandable modules to enhance the bicycle's applicability and versatility. Speed ​​control methods are designed for different speed conditions, and a fuzzy logic-enhanced judgment mechanism is used to achieve precise switching between speed control modes. This provides theoretical and technical support for bicycle operation and adaptation under complex and variable operating conditions.

[0245] The unmanned motorcycle structure and motion control system described in this invention has the advantages of reasonable structural design, advanced control strategy and wide adaptability to various scenarios, providing solid technical support for the application of intelligent mobile platforms in complex terrain conditions, and has high engineering transformation value and application promotion potential.

[0246] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A self-balancing control system for an unmanned motorcycle across the entire speed range, characterized in that, It includes: A balance mechanism and a motion controller; among which, The balance mechanism includes a rear-wheel swing mechanism and a front-wheel steering mechanism; among which, the rear-wheel swing mechanism consists of an unmanned motorcycle rear-wheel swing servo and a swing link, and the rear-wheel swing servo connects the swing link to control the lateral swing of the rear wheel; the front-wheel swing mechanism consists of a front-wheel swing servo and a steering link, and the front-wheel swing servo connects the rotating link to control the rotation of the front wheel of the unmanned motorcycle; The motion controller includes a rear-wheel swing control module and a front-wheel steering control module; among which, the rear-wheel swing control module is used to output a signal to control the rear-wheel swing servo to control the swing link when the unmanned motorcycle is stationary or at low speed, and the front-wheel steering control module is used to output a signal to control the front-wheel rotation servo to control the rotating link when the unmanned motorcycle is at medium or high speed.

2. The unmanned motorcycle full-speed-range self-balancing control system according to claim 1, characterized in that, The swing link is mainly composed of a driving link, a driven link and a connecting plate. The rear-wheel swing servo drives the driving link, and the driving link drives the symmetrically arranged driven links on both sides through the connecting plate, thus forming a parallelogram mechanism with three parallel links to achieve swing.

3. The unmanned motorcycle full-speed-range self-balancing control system according to claim 1 or 2, characterized in that, The motion controller also includes a control strategy switching management module, and the switching logic is: if v ≤ Vtor_MIN, enter the rear-wheel swing control mode; if v ≥ Vtor_MAX, enter the front-wheel steering control mode; if Vtor_MIN < v < Vtor_MAX, use fuzzy inference for hybrid control of the two control modes.

4. The unmanned motorcycle full-speed-range self-balancing control system according to claim 3, characterized in that, The fuzzy inference outputs a control mode confidence level S ∈ [0, 1] according to the vehicle speed and acceleration, and determines whether to switch or perform hybrid output according to the confidence level; Contribution of vehicle speed to S: Contribution of acceleration to S: Among them, a min and a max These are the minimum and maximum accelerations of the bicycle, respectively, and x represents the current acceleration value of the vehicle. Calculation of confidence level combining vehicle speed and acceleration: S=α·S v +(1-α)·S a Among which, α is a weighting coefficient, and by adjusting α, the influence degree of vehicle speed and acceleration on S can be adjusted. Hybrid control strategy: in mix =(1-S)·u SMC +S·u NMPC+NTSM 。 Where, μ mix This represents the final hybrid control output signal; μ SCM This indicates the control signal output by the rear wheel sway control mode; μ NMPC+NTSM This indicates the control signal output by the front wheel steering control mode.

5. The unmanned motorcycle full-speed-range self-balancing control system according to claim 1, characterized in that, The rear-wheel swing control module calculates a sliding surface according to the control state variable, calculates a saturation function for suppressing chattering according to the sliding surface, and calculates a control output torque to the rear-wheel swing servo based on the saturation function.

6. The unmanned motorcycle full-speed-range self-balancing control system according to claim 5, characterized in that, The control state variables are: vehicle body lateral tilt angle θ, tilt angle angular velocity. Vehicle body deflection angle φ, rate of change of deflection angle Sliding surface definition: Where, α i The weighting coefficients are positive, i = 1, 2, 3, 4; Control law form: τ(t) = -K·sat(s) Among which, τ(t) is the control output torque for driving the rear-wheel yaw; K is the control gain; sat is the saturation function: Among which, ε > 0 is the boundary layer thickness parameter.

7. The unmanned motorcycle full-speed-range self-balancing control system according to claim 1, characterized in that, The front-wheel rotation control module is realized by a two-stage composite control composed of a nonlinear model predictive controller and a nonsingular terminal sliding mode controller.

8. The unmanned motorcycle full-speed-range self-balancing control system according to claim 7, characterized in that, The objective function and constraint conditions of the nonlinear model predictive controller are: Let the state vector of the vehicle at the kth sampling moment be: Where: θ k The vehicle body camber angle; v is the tilt angle and angular velocity. k Forward velocity; δ k The front wheel steering angle; y k For lateral positional deviation; φ k For heading angle; The state of the vehicle evolves according to the nonlinear differential dynamics model at each step as: x k+1 =f(x k ,δ k ) Among which, f() is the nonlinear discrete dynamics model of the vehicle, which can be obtained according to the unmanned motorcycle model or measured modeling data; The front-wheel rotation control module constructs the following weighted cost function within a prediction time domain of length N: Where: w1~w4 are the terminal attitude accuracy weights; θ N The tilt angle is the angle at the endpoint (k = N). Control constraints: Rudder angle amplitude limit: δ min ≤δ k ≤δ max Rudder angle change rate limit: |δ k -δ k-1 |≤Δ max Attitude tilt angle constraint: |θ k |≤θ safe ; Based on the objective function, solve the finite-time domain nonlinear constrained optimization problem: stx k+1 =f(x) k δ k x0 is known d k ∈[δ min ,d max ] |d k -d k-1 |≤Δ max |θ k |≤θ safe This optimization problem is solved in each control cycle to obtain the optimal rudder angle input sequence: Only take the first control quantity in the sequence as the current cycle execution instruction output: The remaining part Used as the initial value for the next optimization cycle.

9. The unmanned motorcycle full-speed-range self-balancing control system according to claim 8, characterized in that, The control function of the non-singular terminal sliding mode controller is: Let the system reference input be δref and the actual control quantity be δ, then the tracking error is: e=δ-δ ref Construct the non-singular terminal sliding surface function s as follows: Where β>0 is the design gain; p, q are positive odd numbers; The control law is designed as follows: in, η is the reference acceleration; η is the sliding mode gain; sat(s) is the saturation function; Where φ>0 is the saturation bandwidth parameter.

10. A self-balancing control method for an unmanned motorcycle across the entire speed range, characterized in that, The specific process is as follows: Perception and Mapping: Online mapping of the environment in which the unmanned motorcycle is located and estimation of the vehicle's own pose; Path generation: On the established map, the line segment connecting the current pose and the target position is used as the path optimization area, and the planned path is represented by a cubic B-spline curve. The path optimization is performed using a multi-objective evolutionary optimization method. Minimum time trajectory generation: The geometric path is further mapped to a time trajectory. The goal of time trajectory optimization is to minimize the overall travel time and meet the set constraints. At each time point, a recommended tilt angle is calculated as a feedforward input to assist in adjusting the vehicle's lateral attitude. Tracking control and zone coordination: Based on the current vehicle speed, rear wheel sway control and / or front wheel steering control are used to complete path tracking and lateral balance.

Images