Self-balancing two-wheeled vehicle queue formation control system

Abstract

The invention relates to the technical field of intelligent traffic control, and discloses a self-balancing two-wheeled vehicle queue formation control system which comprises a cloud scheduling platform, a formation control unit and a vehicle end execution unit. And the cloud scheduling platform generates a global avoidance path based on the accumulated gravitational potential energy change and the pavement flatness coefficient, and executes dynamic position complementing. The formation control unit constructs a communication topology and executes formation switching based on environmental perception; the cooperative control module adopts an attitude position coupling algorithm, an expected acceleration is corrected through a gain function attenuating along with an attitude angle index, and the attitude balance is maintained preferentially when a tracking error is eliminated. The vehicle end execution unit is provided with an environment sensing and composite actuating module, and a hub motor and a control moment gyroscope are used for responding to longitudinal driving and transverse balance instructions respectively, so that dynamic decoupling is achieved. According to the invention, the control conflict between two-wheeled vehicle formation trajectory tracking and attitude stability is effectively solved, and the driving safety of the system is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic control technology, specifically to a self-balancing two-wheeled vehicle platooning control system. Background Technology

[0002] Self-balancing two-wheeled vehicles, with their small footprint, maneuverability, and zero turning radius, have broad application prospects in last-mile logistics delivery, intelligent inspection, and park commuting. With the mature application of Control Moment Gyroscope (CMG) technology and large-scale self-balancing algorithms, advanced self-balancing two-wheeled vehicles have overcome the stability bottleneck of traditional inverted pendulum systems, achieving long-term stable uprightness and balance under all working conditions, including stationary, low-speed, and medium-to-high-speed operation. This lays a solid physical foundation for multi-vehicle collaborative operation and automated platooning.

[0003] However, although the stability issues of individual vehicles have been effectively resolved, forming a platoon of multiple self-balancing two-wheeled vehicles for cooperative driving still faces system-level challenges in path planning and dynamics adaptation, making it difficult to directly transfer and apply existing general platooning control technologies. At the global path planning level, existing navigation schemes typically use the shortest path or the shortest time as the sole optimization objective. This logic applies to four-wheeled vehicles with inherent static stability, but it does not fully consider the special energy consumption model and terrain sensitivity of self-balancing two-wheeled vehicles. Although the CMG system can maintain attitude balance, when traveling on long distances and steep slopes, the vehicle needs to consume a large amount of electrical energy to maintain the anti-gravity torque, which can easily cause some vehicles in the formation to run out of power prematurely and fall behind. In addition, the smoothness of the road surface (such as potholes and speed bumps) greatly interferes with the formation of two-wheeled vehicles, and existing plans lack a weighted assessment mechanism for the risks of road surface unevenness and accumulated gravitational potential energy.

[0004] At the platooning coordination control level, mainstream cooperative adaptive cruise control (CACC) algorithms primarily focus on rapidly eliminating position errors, assuming that the actuators can respond instantly to any acceleration command. However, there is a strong coupling relationship between the longitudinal acceleration and pitch attitude of self-balancing two-wheeled vehicles (i.e., acceleration inevitably leads to backward pitch, and deceleration inevitably leads to forward pitch). If the strong tracking algorithm of traditional four-wheeled vehicles is directly applied, when the platoon undergoes rapid acceleration, deceleration, or formation changes, the upper-level controller will issue acceleration commands exceeding the vehicle's physical boundaries, causing drastic fluctuations in the vehicle's pitch angle. Control strategies that prioritize achieving the correct position while neglecting attitude margins can easily lead to the underlying balancing system entering the saturation zone, and may even trigger instability risks.

[0005] At the underlying execution level, while CMG-based composite actuation systems possess strong disturbance rejection capabilities, their control logic is extremely complex. During platooning, vehicles must respond to path commands from the cloud, follow commands from the vehicle ahead, and simultaneously manage their own balancing. Existing control architectures often lack decoupling mechanisms between longitudinal drive and lateral balancing, making it difficult to effectively allocate commands and dynamically limit them when the CMG frame angles approach singular positions or when motor torque is saturated. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a self-balancing two-wheeled vehicle platooning control system, which solves the problem that existing plans lack a weighted assessment mechanism for road surface unevenness risk and accumulated gravitational potential energy.

[0007] To achieve the above objectives, the present invention provides a self-balancing two-wheeled vehicle platoon formation control system, including a cloud-based scheduling platform, a platoon formation control unit, and a vehicle-side execution unit.

[0008] The cloud-based scheduling platform calculates and generates a set of formation scheduling tasks and issues dynamic replanning instructions through a wide area communication network. The formation control unit receives real-time status information of tasks and vehicles, establishes a communication link topology, and resolves cooperative control commands. The vehicle-side execution unit collects local environmental data, responds to coordinated control commands to drive vehicle movement, and prioritizes maintaining vehicle body posture balance during execution.

[0009] At the cloud scheduling level, the cloud scheduling platform has a built-in hierarchical path planning engine and a high-precision map database. When generating global path planning instructions, the hierarchical path planning engine retrieves road attribute information from the high-precision map database and performs multi-objective weighted cost calculation.

[0010] This calculation logic, while considering the physical length of the road segment, introduces the absolute value of the cumulative gravitational potential energy change calculated based on the slope along the road segment, as well as a risk smoothness coefficient characterizing the pavement material and pothole density. The system marks road segments with excessively high weights or average slopes exceeding the safety threshold as impassable areas to avoid overturning risks during the global planning stage.

[0011] For multi-vehicle task allocation, the cloud-based scheduling platform is configured with platooning task generation logic and dynamic replacement logic. The platooning task generation logic uses a cost function minimization strategy to calculate the matching degree between idle vehicles and the current task. The matching degree is calculated based on the Euclidean distance between the vehicle's current position and the task's starting point, as well as the reciprocal of the vehicle's remaining battery power.

[0012] When the dynamic replacement logic detects that a vehicle in the formation is offline, it predicts the future trajectory position based on the current travel speed vector of the formation and the remaining distance. It then calculates the dynamic merging point that can reach the trajectory position the fastest, starting from the current position of the replacement vehicle, and thus generates a dynamic replanning instruction.

[0013] At the formation control level, the queue topology management module in the formation control unit constructs a directed graph-structured communication topology and establishes the leader and follower relationships.

[0014] The formation planning module selects the target geometric configuration based on road environment parameters and calculates the expected position offset of each vehicle relative to the formation reference point. When environmental perception data indicates that the road width is limited, the module uses a fifth-order polynomial interpolation algorithm with constrained curvature derivatives to generate a smooth transition trajectory and updates the expected position offset in real time accordingly to smoothly switch to single-column formation mode.

[0015] The cooperative control computation module employs an attitude and position coupled control algorithm. The system first constructs the formation tracking error based on the virtual navigator state, the actual vehicle state, and the desired position offset.

[0016] Based on model predictive control theory, the module obtains the original desired acceleration required to eliminate formation tracking errors by solving a quadratic programming cost function that includes prediction error terms and control input terms. Building upon this, the module introduces an attitude coupling gain function to dynamically correct the original desired acceleration. This gain function characterizes the vehicle's dynamic margin in the current attitude, and its value is set to decrease exponentially with increasing pitch and roll angles. The system obtains the corrected acceleration command by multiplying the original desired acceleration by this gain function, thereby automatically reducing the tracking weight for position errors when the vehicle attitude approaches the instability boundary.

[0017] The system monitors the Lyapunov function and its derivative of the formation error system in real time to determine the risk of formation divergence and trigger a degraded operation mode that only performs single-vehicle self-balancing.

[0018] At the vehicle-side execution level, the vehicle-side execution unit is equipped with an environmental perception module, an attitude execution module including an inertial measurement unit and a control torque gyroscope component, a power drive module including a hub motor and a steering actuator, and a composite actuation module.

[0019] The environmental perception module uses LiDAR or visual sensors to acquire information about road boundaries, the positions of static obstacles, and relative distances to preceding vehicles. The attitude execution module uses an extended Kalman filter algorithm to fuse data from the inertial measurement unit to calculate real-time 3D attitude and performs zero-speed correction.

[0020] The composite actuation module performs decoupled control of longitudinal drive and lateral balance.

[0021] In terms of lateral balance, the module calculates the required compensation torque based on the current roll angle error, combines the constant angular momentum of the control torque gyroscope flywheel with the current frame angle position, and inversely solves the precession angular velocity command of the universal joint, which also includes singularity avoidance and frame reset logic.

[0022] In terms of longitudinal drive, the module is based on an inverted pendulum dynamics model and uses a linear quadratic regulator to calculate the hub motor torque command. This command consists of a feedforward term based on the modified acceleration command and a feedback term based on vehicle state information. The gain coefficient of the feedback term is obtained by solving the algebraic Riccati equation and is used to weight the vehicle pitch angle, pitch rate and vehicle speed difference, thereby simultaneously meeting the requirements of longitudinal speed tracking and longitudinal pitch balance.

[0023] This invention provides a self-balancing two-wheeled vehicle platooning control system. It has the following advantages: 1. This invention effectively improves the energy efficiency and traffic safety of self-balancing two-wheeled vehicle platoons by introducing accumulated gravitational potential energy changes and road surface risk smoothness coefficients into cloud-based path planning. Because self-balancing vehicles are sensitive to slope and have limited battery capacity, the system automatically avoids long, steep slopes and uneven road sections with high energy consumption, reducing the risk of tipping over and extending operational range. Furthermore, the system's dynamic replacement logic can quickly integrate substitute vehicles based on trajectory prediction, ensuring mission continuity in the event of a single vehicle failure.

[0024] 2. This invention utilizes an attitude-position coupling control algorithm to resolve the contradiction between trajectory tracking accuracy and individual vehicle attitude stability in formation control. By introducing an attitude coupling gain function that decays exponentially with increasing pitch and roll angles, the system can automatically reduce the correction force for position errors when the vehicle attitude approaches the instability boundary. This prioritizes passive balancing over active tracking, preventing system divergence due to actuator saturation and ensuring high robustness of the formation.

[0025] 3. This invention achieves decoupled control of longitudinal drive and lateral balance through a composite actuation module, improving the stability of self-balancing two-wheeled vehicles under low-speed and stationary conditions. The system utilizes a control torque gyroscope to compensate for lateral roll and a hub motor to handle longitudinal balance, overcoming the limitations of traditional two-wheeled vehicles that rely on forward speed for balance. The decoupled control logic allows the platoon to maintain a stable formation configuration without mechanical support in zero-speed scenarios such as traffic congestion or stop-and-go traffic lights. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the architecture of the self-balancing two-wheeled vehicle platooning control system of the present invention; Figure 2This is a schematic diagram of the cloud-based global path planning and task allocation process of the present invention; Figure 3 This is a schematic diagram of the automatic switching and smooth trajectory generation of the formation planning module of the present invention; Figure 4 This is a schematic diagram of the attitude and position coupling control logic of the collaborative control calculation module of the present invention; Figure 5 This is a schematic diagram of the hardware configuration and data flow of the vehicle-side execution unit of the present invention; Figure 6 This is a schematic diagram of the longitudinal and lateral decoupling control logic of the composite actuation module of the present invention. Figure 7 This is a schematic diagram of the dynamic replacement and merging logic during platooning in this invention.

[0027] Among them, 110 is the cloud-based scheduling platform; 120 is the formation control unit; 121 is the queue topology management module; 122 is the formation planning module; 123 is the collaborative control and calculation module; 130 is the vehicle-side execution unit; 131 is the environmental perception module; 132 is the attitude execution module; 133 is the power drive module; 134 is the vehicle-mounted communication module; and 135 is the composite actuation module. Detailed Implementation

[0028] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Please refer to the appendix. Figure 1 The present invention provides a self-balancing two-wheeled vehicle platoon formation control system, comprising: a cloud scheduling platform 110, a platoon formation control unit 120, and a vehicle-side execution unit 130.

[0030] The cloud-based dispatching platform 110 is configured on a remote server or cloud cluster. The cloud-based dispatching platform 110 establishes a data connection with the formation control unit 120 via a wide area communication network. The cloud-based dispatching platform 110 responds to task requests, calculates and generates global path planning instructions and formation scheduling task sets. The task set includes a vehicle ID allocation list, task start and end point coordinates, and preset time window constraints. The cloud-based dispatching platform 110 also monitors the status data uploaded by vehicles in real time and issues dynamic replanning instructions when anomalies are detected or sudden traffic control measures are implemented.

[0031] The formation control unit 120 is the logical center for coordinating the control of multiple self-balancing two-wheeled vehicles. The physical carrier of the formation control unit 120 is either an independent roadside edge computing device or a specific self-balancing two-wheeled vehicle designated as the navigator within the convoy. The formation control unit 120 includes a queue topology management module 121, a formation planning module 122, and a cooperative control calculation module 123.

[0032] The queue topology management module 121 is used to establish and maintain the communication link topology between multiple self-balancing two-wheeled vehicles. Based on the ID allocation list issued by the cloud scheduling platform 110, the queue topology management module 121 assigns a logical number and role attribute to each vehicle joining the network within the queue. Role attributes are divided into navigator and follower. The queue topology management module 121 updates the topology connection matrix in real time when vehicles join or leave the queue.

[0033] The formation planning module 122 is used to select a target geometric configuration from a preset formation database based on current road environment parameters and task type. Target geometric configurations include, but are not limited to, single-column formations, double-column parallel formations, and V-shaped formations. The formation planning module 122 calculates the expected positional offset of each controlled vehicle relative to the formation reference point.

[0034] The cooperative control calculation module 123 is used to receive real-time status information from each vehicle. The real-time status information includes position coordinates, velocity vectors, and attitude parameters. Based on a preset attitude-position coupling control algorithm, the cooperative control calculation module 123 calculates the desired acceleration command and desired heading command for each vehicle and sends the commands to the vehicle-side execution unit 130.

[0035] The vehicle-side actuator 130 is installed on each self-balancing two-wheeled vehicle. The vehicle-side actuator 130 is a physical entity that performs specific motion control and maintains its own balance. The vehicle-side actuator 130 includes an environmental perception module 131, an attitude execution module 132, a power drive module 133, an on-board communication module 134, and a composite actuation module 135.

[0036] The environmental perception module 131 includes at least one of a lidar, a visual sensor, or an ultrasonic sensor. The environmental perception module 131 is used to collect local environmental data surrounding the vehicle. This local environmental data includes road boundaries, the positions of static obstacles, and relative distances to preceding vehicles. The environmental perception module 131 transmits the collected raw data to the formation control unit 120 after preprocessing or for local obstacle avoidance calculations.

[0037] The attitude execution module 132 is used to maintain the self-balance of the vehicle body during static and dynamic processes. The attitude execution module 132 includes an inertial measurement unit (IMU) and a control torque gyroscope (CMG) assembly. The IMU collects the vehicle body's three-axis acceleration and three-axis angular velocity in real time, calculating the current pitch and roll angles. The control torque gyroscope assembly is configured to generate gyroscopic torque by adjusting the precession angular velocity under low-speed or stationary conditions, directly controlling the roll attitude of the vehicle body.

[0038] The power drive module 133 includes a hub motor and a steering actuator. The power drive module 133 responds to hub motor torque commands and desired heading commands issued by the composite actuation module 135. The power drive module 133 outputs a target driving torque to the hub motor. This driving torque simultaneously meets the vehicle's longitudinal speed tracking and longitudinal pitch balance requirements.

[0039] The vehicle communication module 134 includes a V2X communication module or a dedicated short-range communication (DSRC) module. The vehicle communication module 134 is used to enable high-frequency data interaction between the vehicle-side execution unit 130 and the formation control unit 120. The interactive data includes uplink vehicle state vectors and downlink control command vectors.

[0040] The composite actuation module 135 is the underlying control hub within the vehicle-side execution unit 130, configured to perform decoupled control of longitudinal drive and lateral balance. The composite actuation module 135 receives coordinated control commands from the upper layer and decomposes them based on the vehicle dynamics model into drive commands for the power drive module 133 and balance commands for the attitude execution module 132, ensuring that the vehicle maintains its dynamic stability while performing formation tasks.

[0041] In this embodiment, the cloud-based scheduling platform 110, the formation control unit 120, and the vehicle-side execution unit 130 form a closed-loop control circuit. During system operation, the environmental perception module 131 continuously updates local environmental data. The formation control unit 120 dynamically adjusts the formation configuration and calculates cooperative control commands based on environmental changes. While executing formation commands, the vehicle-side execution unit 130 prioritizes responding to the balance requests from the attitude execution module 132 to ensure the vehicle's attitude stability during acceleration and deceleration.

[0042] Reference Appendix Figure 2 The cloud-based dispatch platform 110 establishes a two-way connection with the vehicle platoon through a standardized communication interface. Its configuration transforms macro-level operational needs into digital instructions executable by the platoon. The cloud-based dispatch platform 110 integrates a high-precision map parsing module, a global path search module, and a multi-agent task allocation algorithm module.

[0043] After receiving a job request, the cloud-based dispatch platform 110 parses the parameters in the request, including the coordinates of the job start point, the coordinates of the job end point, the task priority, and the required payload. To adapt to the urban road network, the cloud-based dispatch platform 110 incorporates a hierarchical path planning engine. This engine calculates the globally optimal route at the topology level based on the road resistance function, generating a series of waypoint sequences with geographic information. The high-precision map database pre-stores road attribute information, including the geometric topology of road segments, average slope data, and road surface smoothness level data. When generating the waypoint sequence, the path planning engine retrieves road attribute information and marks road segments with an average slope exceeding a preset safety threshold (e.g., 15 degrees) or a road surface smoothness level below a preset value as impassable areas, thereby mitigating the risk of self-balancing two-wheeled vehicles tipping over at the global path level.

[0044] For platoon formation task generation, the cloud-based scheduling platform 110 employs task allocation logic based on minimizing a cost function. This is determined when the job request requires a certain number of vehicles. At that time, the platform determines and generates a platooning task. The cloud-based dispatch platform 110 reads the location coordinates of idle vehicles in the vehicle status database in real time. and remaining battery power The platform uses the following cost function. Calculate the first Matching degree between idle vehicles and current tasks: ; in, Indicates the first The current location coordinates of the idle vehicle; Indicates the starting point coordinates (or rendezvous point coordinates) of the formation operation task; Indicates the first The remaining battery charge of an idle vehicle; This represents the Euclidean distance between the vehicle's current position and the mission's starting point. and These are the preset weighting coefficients.

[0045] Platform selects cost function smallest The vehicles form a target vehicle group and generate a unique formation mission ID.

[0046] After locking onto the vehicle group, the cloud-based dispatch platform 110 generates a task data packet containing multiple sub-instructions. The task data packet explicitly defines the initial role of each vehicle. For a vehicle designated as the Leader, the data packet contains a complete global waypoint sequence and a preset cruise speed curve; for a vehicle designated as a Follower, the data packet contains its platoon ID and the Leader's communication address. This pre-assignment mechanism avoids communication congestion caused by multiple vehicles vying for roles during self-organization.

[0047] The cloud-based dispatch platform 110 features dynamic rescheduling capabilities. During platooning, the platform periodically receives operational status reports uploaded by the navigator. If a report indicates traffic congestion or construction on a certain road segment, the platform triggers a replanning process, calculates detour routes, and issues updated waypoint sequences. If a vehicle in the platoon goes offline due to a malfunction or runs out of power, the platform generates a dynamic replacement instruction based on the remaining workload. This instruction dispatches a new idle vehicle to plan an interception route, which is then added to the platoon during the platoon's journey to replace the malfunctioning vehicle. The conventional data transmission protocols, database read / write operations, and server load balancing technologies are well-known in the field and will not be elaborated upon here.

[0048] The global path point sequence distributed from the cloud is represented as ,in, This represents a global path point sequence distributed from the cloud. Indicates the formation of a global path An ordered set of specific waypoint elements. ,in, In mathematical expression, it refers to any general element in a sequence; It is a subscript index variable in geodetic coordinates; its value range is... , These represent the x-coordinate, y-coordinate, and path curvature of the point, respectively. This sequence serves as the direct reference input for the trajectory tracking algorithm in the formation control unit.

[0049] Reference Appendix Figure 3 The formation control unit 120 provided by the present invention includes: a queue topology management module 121, a formation planning module 122, and a collaborative control calculation module 123.

[0050] The queue topology management module 121 is configured to establish a communication connection matrix between vehicles based on the V2X communication protocol. During the platoon initialization phase, the queue topology management module 121 receives a list of vehicle IDs from the cloud and constructs a communication topology with a directed graph structure. In this topology, nodes represent vehicles, and directed edges represent information flow directions. The queue topology management module 121 periodically sends heartbeat frames to monitor node status. When it detects that a vehicle node's communication interruption exceeds a preset time limit (e.g., 200 milliseconds), it triggers topology reconstruction logic, marks the node as offline, and updates the adjacency matrix. This is to ensure that the remaining vehicles can re-establish connectivity.

[0051] The formation planning module 122 stores a preset formation geometry parameter library and calculates the desired relative positions of each vehicle based on environmental constraints. In this embodiment, the formation geometry parameter library defines at least three basic configurations: single-column formation, double-column parallel formation, and V-formation. The formation planning module 122 uses relative position vectors... Definition of the first The expected geometric relationship between the vehicle and the navigator, whereby... Indicates the first The expected relative position vector of the vehicle with respect to the navigator; This represents the vertical offset in the relative position vector; This represents the lateral offset in the relative position vector.

[0052] For a single-column formation, set ,in For longitudinal safety distance; For a double-row side-by-side arrangement, set ,in This refers to the lateral spacing between lanes. For V-shaped formation, set ,in It is a V-shaped opening angle.

[0053] The formation planning module 122 also executes automatic switching logic based on environmental perception. This is done when the vehicle-mounted sensors return valid road width data. When the lateral width is less than the minimum required for the current formation, the formation planning module 122 switches the target formation parameters to a single-column column mode. During the formation switching process, the formation planning module 122 generates a smooth transition trajectory. This trajectory uses a fifth-order polynomial interpolation algorithm to connect the current position before the switch with the desired position after the switch. By limiting the curvature derivative of the interpolation curve, excessive lateral acceleration of vehicles during formation changes is suppressed.

[0054] Reference Appendix Figure 4The collaborative control computation module 123 is configured to implement attitude and position coupled control. The collaborative control computation module 123 introduces attitude stability constraints and constructs a two-level closed-loop control architecture. The collaborative control computation module 123 first calculates the state vector of the virtual leader... With the The actual state vector of the vehicle Calculate formation tracking error Actual state vector Formation tracking error Represented as: ; The collaborative control computation module 123 calculates the original desired acceleration based on model predictive control (MPC) theory. To eliminate formation tracking error Specifically, the collaborative control computation module 123 solves the following quadratic programming cost function. To obtain the minimum value : ; in, For prediction in the time domain; It is a semi-positive definite state weighting matrix; It is a positive definite control weighting matrix; and Indicates the first The formation tracking error vector and control input vector of the vehicles; and Indicates the first The vehicle in the prediction time domain The prediction error vector and control input vector at each time step; and This represents the transpose of the prediction error vector and the control input vector.

[0055] Based on this, the cooperative control calculation module 123 introduces an attitude coupling gain function. Dynamically correct the original desired acceleration. Attitude coupling gain function. The mathematical expression for the dynamic margin used to characterize the current attitude is: ; in, The current pitch angle of the vehicle. The current roll angle of the vehicle. and These are the safety attenuation constants for pitch and roll, respectively.

[0056] The corrected acceleration command is finally output by the collaborative control calculation module 123. for: ; This control law causes the system to automatically decrease when the vehicle's attitude angle (pitch or roll angle) increases. This reduces the tracking effort for formation position errors. This mechanism suppresses attitude divergence caused by forcibly tracking position errors, ensuring that the underlying balance control loop has sufficient actuator saturation margin.

[0057] The collaborative control operation module 123 also includes formation stability prediction logic. This logic monitors the Lyapunov function and its derivative of the formation error system in real time. When predicted If the time window is exceeded for an extended period, the collaborative control calculation module 123 determines that the formation is in a state of divergence risk, sends an alarm signal to the cloud dispatch platform 110, and instructs all vehicles to enter a degraded operation mode, in which only single-vehicle self-balancing and emergency braking operations are performed.

[0058] Reference Appendix Figure 5 -Appendix Figure 6 In this embodiment, the vehicle-side execution unit 130 includes an environment perception module 131, an attitude execution module 132, a power drive module 133, an in-vehicle communication module 134, and a composite actuation module 135.

[0059] The environmental perception module 131 is configured to collect environmental data surrounding the vehicle and the vehicle's own motion state data. The environmental perception module 131 integrates a LiDAR, a vision camera, a six-axis inertial measurement unit (IMU), and a wheel speed encoder. The environmental perception module 131 performs spatiotemporal registration of the LiDAR point cloud data and the camera image data to identify the relative position and relative speed of the vehicle in front. Simultaneously, the environmental perception module 131 reads the three-axis angular velocity and three-axis acceleration output from the inertial measurement unit at a preset high sampling frequency, and reads the real-time rotational speed of the wheels through the wheel speed encoder, serving as the observation input for subsequent control loops.

[0060] The attitude execution module 132 is configured to calculate the vehicle's real-time three-dimensional attitude based on raw inertial data. The attitude execution module 132 employs an extended Kalman filter (EKF) algorithm, fusing the gravity component from the accelerometer and the dynamic component from the gyroscope. The state variables output by the attitude execution module 132 include the vehicle's pitch angle. Pitch angular velocity Roll angle and roll angular velocity .

[0061] The attitude execution module 132 calculates the overturning moment caused by the center of gravity shift based on the state variables. To eliminate sensor drift, the attitude execution module 132 executes zero-speed correction logic: when the rotational speed fed back by the wheel speed encoder is zero and the resultant acceleration amplitude measured by the inertial measurement unit is within the standard value of gravitational acceleration (… When the tolerance range is within the specified range, the attitude execution module 132 determines that the vehicle is stationary and resets the integral cumulative error of the gyroscope.

[0062] The power drive module 133 includes a high-power-density hub motor and a steering actuator. The power drive module 133 is the physical actuator for the vehicle's longitudinal movement and path tracking. The power drive module 133 is electrically connected to the composite actuation module 135 and receives and responds to the hub motor torque commands issued by the composite actuation module 135. The power drive module 133 integrates a motor controller that uses a vector control (FOC) algorithm to convert torque commands into three-phase current to drive wheel rotation. By precisely adjusting the output torque of the hub motor, the power drive module 133 generates forward or reverse propulsion while simultaneously maintaining the vehicle's longitudinal (pitch) balance using reaction torque. Furthermore, the power drive module 133 also includes a steering servo or differential controller to respond to desired heading commands by adjusting the front wheel angle or the speed difference between the left and right wheels to achieve vehicle steering.

[0063] The composite actuation module 135 is configured to perform decoupled control of longitudinal drive and lateral balance. The composite actuation module 135 receives corrected acceleration commands from the upper-level cooperative control calculation module 123. And decompose it into hub motor torque command Control torque gyroscope (CMG) gimbal angular velocity command .

[0064] For longitudinal balance, the composite actuation module 135 uses an inverted pendulum dynamics model and a linear quadratic regulator (LQR) to calculate the hub motor torque command. This instruction must simultaneously meet the requirements of trajectory tracking and longitudinal attitude stabilization; its calculation formula is as follows: ; in, For the overall vehicle weight; The radius of the wheel; This refers to the actual vehicle speed; For reference speed; This is the gain coefficient.

[0065] Gain coefficient It is obtained by solving the algebraic Riccati equation, and its optimization objective is to minimize the system's energy cost function. : ; in, The preset state deviation penalty matrix; The preset control energy penalty matrix; Indicates time Integrate from 0 (starting time) until... ; This represents the error between the current state of the system and its ideal state. This represents the control commands applied to the vehicle.

[0066] For lateral balance, the composite actuation module 135 utilizes the precession effect generated by the onboard control torque gyroscope to output a lateral disturbance rejection torque. The composite actuation module 135 calculates the required compensation torque based on the current roll angle error. And the precession angular velocity of the CMG frame was obtained by inverse solving. : ; in, The constant angular momentum of the CMG flywheel; This refers to the current angular position of the CMG frame; This indicates the input target. The composite actuation module 135 has built-in singularity avoidance logic; when a singularity is detected... When approaching a 90-degree singular position, or when the required compensation torque When crossing zero, the composite actuation module 135 drives the frame to reset to the zero-degree position to release the travel space of the universal joint.

[0067] The vehicle-mounted communication module 134 is configured to ensure data interaction between the vehicle and the platooning network. The vehicle-mounted communication module 134 adopts a communication protocol compliant with IEEE 802.11p or C-V2X standards. The vehicle-mounted communication module 134 has a priority message queue, marking control-level messages containing attitude instability warnings, emergency braking requests, and platooning synchronization pulses as the highest priority. When sending data packets, the vehicle-mounted communication module 134 writes the current timestamp and sequence number into the frame header. The receiving end uses this information to calculate the transmission delay and perform packet loss compensation to ensure timing synchronization of the control loop.

[0068] See appendix Figure 7 The cloud-based dispatching platform 110 provided by this invention is configured to perform global route planning, multi-vehicle task allocation, and dynamic anomaly handling. The cloud-based dispatching platform 110 represents the operating area by establishing a discretized road network topology model, which is defined as a directed weighted graph. ,in Represents the set of intersection nodes. Represents the set of road segment edges.

[0069] To address the physical characteristics of self-balancing two-wheeled vehicles—their sensitivity to slope and limited battery capacity—the cloud-based scheduling platform 110 employs a multi-objective weighted cost function to calculate the globally optimal path. Each edge in Assign weights. Edge The cost of passage The calculation formula is as follows: ; in, For nodes To the node The physical length of the road segment; This is the slope function along the road segment; The absolute value of the accumulated gravitational potential energy change in this road section is used to quantify energy consumption demand. The risk smoothness coefficient of the road section; Weighted by distance (efficiency); Energy consumption (slope) weighting; Risk (smoothness) weight.

[0070] Risk flatness coefficient Pre-stored in a high-precision map database, the road surface is discretized and assigned a grade value based on road material, pothole density, and the number of speed bumps (e.g., smooth asphalt road surface is assigned a value of 0, and masonry road surface is assigned a value of 5). If the average gradient of a road segment exceeds the maximum climbing angle threshold of the self-balancing two-wheeled vehicle, the cloud-based dispatch platform 110 will adjust the road segment's gradient accordingly. By setting it to infinity, the path is eliminated at the planning level, ensuring physical accessibility and safety.

[0071] The cloud-based dispatch platform 110 is also equipped with dynamic replacement logic during platooning operations. When it detects that a vehicle in the platoon is offline due to a malfunction or depleted battery, the platform generates a dynamic integration command. The platform then determines the integration command based on the platoon's current speed vector. And the remaining distance, predicting the formation in future consecutive moments. trajectory position .

[0072] The cloud-based dispatch platform 110 calculates the fastest possible arrival time based on the current location of the backup vehicle. travel time And solve for the condition by using an iterative search algorithm. minimum time point The corresponding location is designated as the merging point. This dynamic merging mechanism reduces the time the entire formation spends waiting for backup vehicles, maintaining mission continuity.

[0073] At the data interaction level, the cloud-based dispatch platform 110 and the vehicle terminal communicate via the MQTT protocol for message publishing and subscription. The issued task data packets are encapsulated in JSON format, and the specific fields of the data packets include: task ID, waypoint sequence array, road segment speed limit value, and platoon communication key. The standard network communication handshake process, database storage process, and server load balancing configuration are well-known techniques to those skilled in the art and will not be elaborated upon here.

Claims

1. A self-balancing two-wheeled vehicle platooning control system, characterized in that, include: The cloud-based scheduling platform responds to task requests through a wide area communication network, calculates and generates a fleet scheduling task set that includes a vehicle ID allocation list and global path planning instructions, and issues dynamic replanning instructions. The formation control unit is used to receive the formation scheduling task set, dynamic replanning instructions and real-time status information fed back by vehicles, establish a communication link topology, and solve cooperative control instructions. The vehicle-side execution unit is a physical entity that performs specific motion control and maintains its own balance. It is used to collect local environmental data and respond to the cooperative control commands to drive the vehicle to move, while prioritizing the maintenance of the vehicle's attitude balance during the execution process.

2. The self-balancing two-wheeled vehicle platooning control system according to claim 1, characterized in that, The cloud-based scheduling platform has a built-in hierarchical path planning engine and a high-precision map database. When generating the global path planning instruction, the hierarchical path planning engine retrieves the road attribute information stored in the high-precision map database, performs multi-objective weighted cost calculation, and assigns weights to the edges in the road network. The factors considered in the weight assignment calculation include the physical length of the road segment, the absolute value of the cumulative gravitational potential energy change calculated based on the slope function along the road segment, and the risk smoothness coefficient. The hierarchical path planning engine also marks road sections with average slopes exceeding a preset safety threshold and road surface smoothness levels below a preset value as impassable areas.

3. The self-balancing two-wheeled vehicle platooning control system according to claim 1, characterized in that, The cloud-based scheduling platform is configured with formation task generation logic and dynamic replacement logic; The formation task generation logic reads the location coordinates and remaining battery power of idle vehicles in real time, calculates the matching degree between idle vehicles and the current task using a cost function minimization strategy, and generates the formation scheduling task set containing the vehicle ID allocation list and global path planning instructions. The matching degree is calculated based on the Euclidean distance between the vehicle's current position and the mission starting point, and the reciprocal of the vehicle's remaining battery power. When the dynamic replacement logic detects that a vehicle in the formation is offline, it predicts the trajectory position of the formation in future consecutive moments based on the current travel speed vector and remaining distance of the formation. It sets the current position of the replacement vehicle as the calculation starting point, calculates the dynamic merging point that can reach the trajectory position the fastest, and generates the dynamic replanning instruction.

4. The self-balancing two-wheeled vehicle platooning control system according to claim 1, characterized in that, The formation control unit specifically includes: The queue topology management module constructs a communication topology with a directed graph structure based on the vehicle ID allocation list in the formation scheduling task set, and determines the navigator or preceding vehicle tracking object for each vehicle according to role attributes. The formation planning module selects a target geometric configuration based on road environment parameters and calculates the expected position offset of each vehicle relative to the formation reference point based on the target geometric configuration. The collaborative control calculation module selects data interaction objects based on the communication topology to obtain the status information of adjacent vehicles or the navigator, and constructs the formation tracking error by combining the expected position offset with the real-time status information fed back by the vehicles. The cooperative control calculation module uses an attitude and position coupling control algorithm to eliminate the formation tracking error, thereby calculating the cooperative control command for each vehicle.

5. A self-balancing two-wheeled vehicle platooning control system according to claim 4, characterized in that, The formation planning module stores a formation geometry parameter library, defining basic configurations including single-column, double-column, and V-shaped formations; The formation planning module executes automatic switching logic based on environmental perception. When the effective road width data fed back by the vehicle is less than the minimum lateral width required for the current formation, the target formation parameters are switched to single-column column mode. The formation planning module generates a smooth transition trajectory during the switching process and updates the desired position offset in real time based on the smooth transition trajectory. The transition smooth trajectory uses a fifth-order polynomial interpolation algorithm with limited curvature derivatives to connect the current position before the switch with the desired position after the switch.

6. A self-balancing two-wheeled vehicle platooning control system according to claim 4, characterized in that, The cooperative control computation module implements the attitude and position coupling control algorithm through the following steps: The formation tracking error is calculated based on the virtual navigator's state vector, the vehicle's actual state vector, and the desired position offset. Based on model predictive control theory, the original expected acceleration required to eliminate the formation tracking error is obtained by solving a quadratic programming cost function that includes prediction error terms and control input terms. An attitude coupling gain function is introduced to dynamically correct the original desired acceleration, resulting in a corrected acceleration command. The attitude coupling gain function is configured to characterize the dynamic margin under the current attitude, and the value of the attitude coupling gain function decreases exponentially as the current pitch angle and roll angle of the vehicle increase.

7. A self-balancing two-wheeled vehicle platooning control system according to claim 1, characterized in that, The collaborative control calculation module is also configured with formation stability prediction logic; The formation stability prediction logic monitors in real time the Lyapunov function and its derivative constructed using the formation tracking error; When the derivative is predicted to be greater than zero and the duration exceeds the preset time window, the formation is determined to be in a divergence risk state. An alarm signal is sent to the cloud scheduling platform, and all vehicles are instructed to enter a degraded operation mode that only performs single-vehicle self-balancing and emergency braking.

8. A self-balancing two-wheeled vehicle platooning control system according to claim 6, characterized in that, The vehicle-mounted execution unit specifically includes: An environmental perception module, comprising at least one of a lidar, a visual sensor, or an ultrasonic sensor, is configured to collect local environmental data around the vehicle. The local environmental data includes road boundaries, the location of static obstacles, and the relative distance to the preceding vehicle; The attitude execution module includes an inertial measurement unit and a control torque gyroscope component, and is configured to perform attitude calculation and balance control; The power drive module includes a hub motor and a steering actuator; The composite actuation module receives the coordinated control command and decomposes it into hub motor torque command and control torque gyroscope universal joint angular velocity command, which are then sent to the power drive module and the attitude execution module respectively to perform decoupled control of longitudinal drive and lateral balance.

9. A self-balancing two-wheeled vehicle platooning control system according to claim 8, characterized in that: The composite actuation module has built-in lateral balance control logic and longitudinal drive control logic: In the lateral balance control logic, the composite actuation module calculates the required compensation torque based on the current roll angle error, and combines the constant angular momentum of the control torque gyroscope flywheel with the current frame angle position to inversely solve the gimbal angular velocity command. The composite actuation module drives the frame to reset when it detects that the frame angle is close to a singular position and the compensation torque is zero. In the longitudinal drive control logic, the composite actuation module is based on the inverted pendulum dynamics model and uses a linear quadratic regulator to calculate the hub motor torque command; The hub motor torque command is composed of a feedforward term based on the modified acceleration command and a feedback term based on vehicle status information. The gain coefficient of the feedback term is obtained by solving the algebraic Riccati equation.

10. A self-balancing two-wheeled vehicle platooning control system according to claim 8, characterized in that, The attitude execution module is also configured with attitude calculation logic: The attitude execution module uses the extended Kalman filter algorithm to fuse the three-axis acceleration and three-axis angular velocity output by the inertial measurement unit to calculate the real-time three-dimensional attitude of the vehicle. The real-time three-dimensional attitude is fed back to the formation control unit as the real-time status information; The attitude execution module executes zero-speed correction logic. When the rotational speed fed back by the wheel speed encoder is zero and the resultant acceleration amplitude measured by the inertial measurement unit is within the preset tolerance range of the standard value of gravitational acceleration, the vehicle is determined to be stationary and the integral cumulative error of the gyroscope is reset.

Images