Topologically-aware electric vertical takeoff and landing aircraft control method, system, and electric vertical takeoff and landing aircraft

Abstract

The application belongs to the technical field of aircraft control, and specifically discloses an electric vertical take-off and landing aircraft control method and system based on topology perception and an aircraft. The application explicitly incorporates a motor-battery power supply topology matrix into control distribution design, realizes strict decoupling of peak current regulation and main flight control tasks by superimposing a zero space rotational speed correction amount on a nominal control amount, and does not affect the lift and attitude control accuracy of the aircraft. A composite cost function is constructed by using a smooth differentiable maximum value approximation function, and asymptotic convergence in the Lyapunov sense is performed along the zero space projection gradient direction, so as to reduce the peak current of the motor and the battery. By adjusting the weight factor, the motor current and the battery current are simultaneously optimized, and stress transfer caused by single component optimization is avoided. Under the condition of motor or battery failure, the remaining healthy actuators are automatically adapted by order reduction zero space projection to continue to provide effective current protection. The application significantly improves the thermal safety and fault tolerance capability of the power system.

Description

Technical Field

[0001] This invention relates to the field of aircraft control technology, and in particular to a topology-aware control method, system, and electric vertical takeoff and landing (EVTOL) aircraft. Background Technology

[0002] Electric vertical takeoff and landing (EVTOL) aircraft employ a distributed electric propulsion architecture, utilizing multiple electric propulsion units distributed at different locations on the fuselage to provide lift and control torque. In actual flight, factors such as center of gravity shift, asymmetric gust interference, or decreased efficiency of individual propulsion units can cause some motors to bear excessive current loads. Excessive motor current directly causes Joule heat accumulation, which can lead to motor burnout in severe cases. At the same time, the current surge propagates along the power transmission link to the battery pack, which may trigger thermal runaway or a sudden drop in terminal voltage of a specific battery pack, threatening the stability of the DC bus.

[0003] Existing protection solutions include:

[0004] Battery management systems ensure battery safety by preventing overcharging and over-discharging and balancing the state of the cells; this is a passive protection measure.

[0005] The energy management system optimizes the power flow distribution among multiple energy sources, but mainly operates on the energy supply side, ignoring the interaction between flight control and the physical propulsion system topology;

[0006] Traditional control allocation algorithms optimize energy efficiency or minimum control input, but they are unaware of the topological mapping relationship between motors and batteries. This may inadvertently overload a specific motor or battery pack while satisfying flight control commands.

[0007] Furthermore, null-space control allocation methods in over-actuated aircraft utilize the additional degrees of freedom provided by the null space of the control performance matrix to optimize secondary control objectives. For example, the paper "Null-space control allocation for generalized load reduction in eVTOL aircraft" published online in Aerospace Systems in May 2026 by Zhang J., Lu Z., and Holzapfel F. discloses a control allocation method for reducing generalized load using null-space projection; however, this method does not explicitly incorporate the motor-battery power supply topology into the control allocation design, cannot sense and regulate current distribution at the battery pack level, and focuses on reducing generalized load, lacking a direct targeting mechanism for maximum motor current and maximum battery pack current, and also failing to provide a unified trade-off mechanism for the collaborative protection of multiple components of the motor and battery pack, as well as a de-escalation mechanism for motor or battery failures.

[0008] In summary, existing electrical safety protection schemes, including battery management systems, energy management systems, traditional control distribution algorithms, and existing zero-space control distribution methods, do not explicitly integrate the motor-battery power supply topology in the control distribution stage. They also lack a unified targeted regulation mechanism for the maximum motor current and the maximum battery pack current, and lack the ability to degrade current protection under fault conditions. Therefore, they suffer from problems such as passive protection lag, stress transfer, insufficient fault tolerance, and battery pack overload risk due to topology agnosticism. Thus, there is an urgent need to design a control method and system for electric vertical takeoff and landing (EVTOL) aircraft to solve these technical problems. Summary of the Invention

[0009] The purpose of this invention is to address the shortcomings of existing technologies by proposing a topology-aware control method, system, and electric vertical takeoff and landing (EVTOL) aircraft. By explicitly incorporating the motor-battery power supply topology into the control allocation process and utilizing a null-space cumulative projection mechanism, the invention actively suppresses peak currents in the motor windings and battery packs without affecting the main flight control performance, thereby reducing the risk of thermal runaway and significantly improving the safety and reliability of the EVTOL aircraft, thus solving the aforementioned problems.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: a topology-aware control method for an electric vertical takeoff and landing (EVTOL) aircraft, wherein the EVTOL aircraft has... Distributed electric propulsion unit and Each independent battery pack is involved in the attitude and altitude control of the aircraft. There are 1 virtual control variable, and the overdrive condition is satisfied. The m electric propulsion units and n battery packs are connected via a power supply topology mapping relationship; the control method includes the main flight controller generating virtual control commands for maintaining the attitude and altitude of the aircraft based on outer loop commands and aircraft state feedback, and calculating the nominal control quantity corresponding to the virtual control commands through a basic control allocation algorithm; the control method further includes the following steps:

[0011] Step a: Establish a physical mapping model from control input to electrical load: Establish a model based on the square of the rotational speed of each rotor as the control input. The heating current of each motor winding and the discharge current of each battery pack The physical mapping relationship The control input component for the i-th actuator is given in units of . , Let be the rotational speed of the i-th rotor; the physical mapping relationship includes a motor-battery topology matrix. The elements of the topological matrix Characterizing the first The battery pack is the first The power supply relationship of each motor, and the discharge current of each battery pack. satisfy ,in This represents the DC power supply current vector for each motor branch;

[0012] Step b: Construct the composite cost function of peak current: Construct the sub-cost functions of motor current using smooth and differentiable maximum approximation functions. and battery current quantum cost function and with adjustable weighting factors and Weighted summation forms a composite cost function The composite cost function is used to simultaneously suppress peak currents in both the motor and battery electrical components. Regarding the control input vector Differentiable everywhere; calculate the rate of change gradient of the composite cost function with the square of the rotor speed, and determine the contribution weight of each motor current and each battery pack current to the current peak based on the rate of change gradient;

[0013] Step c: Achieve current suppression through cumulative null-space projection superposition: based on the control effectiveness matrix determined by the aircraft dynamics. Constructing the null space projection operator , Control effectiveness matrix This reflects the impact of changes in rotor speed on the aircraft's vertical and angular acceleration, where the null projection matrix... Dimensionless for Moore-Penrose pseudo-reverse, for An identity matrix of order 1. The order is the number of actuators; and the following steps are executed sequentially:

[0014] Step c1: Project the negative direction of the rate of change gradient into the null space, and calculate the single-step null space rotational speed increment without changing the main flight control effect. :

[0015] ;

[0016] in, This is the gain coefficient, in units of... And the gain coefficient The value of makes the composite cost function The solution asymptotically converges to a local optimum along the Lyapunov direction of the null space projection gradient.

[0017] Step c2: Accumulate the single-step zero-space rotational speed increment over the time dimension to obtain the accumulated zero-space rotational speed correction. :

[0018] , ;

[0019] Step c3, the cumulative zero-space rotational speed correction amount satisfies This does not change the forces and torques expected to be generated by the main flight controller;

[0020] Step c4: Calculate the cumulative zero-space rotational speed correction amount. The value is superimposed on the nominal control quantity output by the main flight controller to obtain the final rotor speed command, which actively reduces the peak current of the power system without affecting the attitude control accuracy; the superposition method is determined according to the specific form of the main flight controller.

[0021] Furthermore, in this invention, the physical mapping relationship in step a is based on the direct-quadrature (DQ) coordinate system model of the permanent magnet synchronous motor, assuming the direct-axis current... Ignoring inductor dynamics, gearbox losses, and frictional torque; the physical mapping relationship includes:

[0022] The motor winding current is directly proportional to the square of the rotor speed, specifically:

[0023] ;

[0024] in, This represents the effective value of the phase current of the i-th motor, in amperes (A), reflecting Joule heat intensity and the motor current constant. The unit is , This is the propeller torque coefficient. The reduction ratio of the gearbox. This represents the number of pole pairs of the motor. For permanent magnet flux linkage;

[0025] The DC power supply current of the motor branch is determined by both the electromechanical power and copper losses, specifically:

[0026] ;

[0027] in, Let be the DC supply current of the i-th motor branch. For inverter efficiency, This is the DC bus voltage. The resistance of each phase of the motor;

[0028] Finally, the discharge point current of the battery pack is calculated and the branch pairs of current are aggregated through the topology matrix: ;

[0029] Furthermore, the topology matrix The system adopts a centrally symmetrical isolated power supply architecture, whose elements belong to {0,1}, meaning that each motor is exclusively powered by a specific battery pack. The centrally symmetrical isolated power supply architecture further ensures that the motors connected to each battery pack are centrally symmetrically distributed with respect to the center of gravity of the aircraft. This ensures that when any battery pack fails, the motor that loses power still maintains central symmetry with respect to the center of gravity of the aircraft and does not generate additional unbalanced torque.

[0030] Furthermore, in this invention, the composite cost function in step b... for:

[0031] ;

[0032] Among them, the composite cost function Dimensionless and It is an adjustable weighting factor. This is the normalized reference value for the motor current. This is a normalized reference value for the battery current. and These are smooth maximum approximation functions for the motor current and the battery current, respectively;

[0033] The motor current sub-cost function With battery current quantum cost function Using smooth, differentiable maximum approximation functions respectively Acting on the motor current vector With battery pack current vector The structures are respectively ;

[0034] The smooth, differentiable maximum approximation function In order to be able to convert vectors maximum value The accuracy of an arbitrary function that is approximated as a scalar in a smooth and differentiable manner can be determined by the sharpness parameter. The smooth maximum approximation function includes, but is not limited to, the Kreisselmeier-Steinhauser (KS) function, the LogSumExp (LSE) function, and the q-norm function.

[0035] The composite cost function For control input vector The gradient is given by the chain rule, and the specific process for calculating the gradient of the rate of change is as follows:

[0036] Composite cost function on control input vector gradient:

[0037] ;

[0038] in: The composite cost function is applied to the control input vector. gradient, and Based on the smooth maximum approximation function used The analytical form corresponds to the calculation.

[0039] Furthermore, this invention uses the Kreisselmeier-Steinhauser function to adjust the smooth, differentiable maximum approximation function. The approximate accuracy is as follows:

[0040] , ;

[0041] Alternatively, the smooth, differentiable maximum approximation function can be adjusted using the LSE function (LogSumExp). The approximate accuracy is as follows:

[0042] ;

[0043] Alternatively, the function can be approximated by the q-norm function at its smoothly differentiable maximum value. The approximate accuracy is as follows:

[0044]

[0045] When using the KS function or the LSE function, the gradient is specifically expressed as follows:

[0046] , ;

[0047] in and These are the Softmax weight vectors for the motor current and the battery current, respectively, and their elements are:

[0048] , ;

[0049] When other smooth maximum approximation functions are used, the gradient is calculated by chain expansion of the analytical partial derivatives of the function with respect to its input vector.

[0050] Furthermore, this invention also includes a downgrading process for motor or battery failures:

[0051] After detecting a fault in any motor or battery pack, determine the set of remaining healthy actuators that are operating normally. From the control performance matrix Extract the index set The corresponding columns constitute the reduced-order control effectiveness matrix. ;

[0052] Calculate the reduced-order null projection matrix :

[0053] ;

[0054] Calculate the reduced-order single-step zero-space rotational speed increment :

[0055] ;

[0056] The reduced-order single-step zero-space rotational speed increment By index set Mapping back to full-dimensional vector The component that does not correspond to the health actuator is set to zero; the subsequent accumulation and superposition process is the same as steps c2 to c4.

[0057] Furthermore, in this invention, the main flight controller is an arbitrary controller that generates the virtual control commands based on the aircraft state feedback and outer loop commands, and the controller is:

[0058] The incremental nonlinear dynamic inverse controller generates the desired output dynamic through a reference model, obtains the desired pseudo-control quantity by combining feedforward and output feedback, and obtains the virtual control incremental command by subtracting the current virtual control feedback estimate.

[0059] Alternatively, it can be a non-incremental nonlinear dynamic inverse controller, in which the virtual control command is calculated directly by inverting the model.

[0060] Alternatively, for a model prediction controller, the composite cost function J(u) can be used as an additional term of the prediction cost function;

[0061] Alternatively, it could be a proportional-integral-derivative controller;

[0062] Alternatively, it could be a controller within a robust or adaptive control framework.

[0063] Furthermore, in this invention, the basic control allocation algorithm is an algorithm capable of calculating the nominal control quantity that satisfies the physical constraints of the actuator based on the virtual control instruction. Specifically, the basic control allocation algorithm is as follows:

[0064] The pseudo-inverse algorithm for redistribution, through Calculate the nominal speed increment, and overflow the saturated input from the active set and redistribute the residual error to the remaining input;

[0065] Alternatively, the weighted least squares allocation algorithm;

[0066] Alternatively, a direct allocation algorithm can be used to solve for the maximum achievable torque set within the convex hull spanned by the column vectors of the control performance matrix.

[0067] Alternatively, an allocation algorithm based on linear programming or quadratic programming.

[0068] Furthermore, in this invention, the cumulative zero-space rotational speed correction amount The superposition of the nominal control quantity output by the main flight controller is implemented in the following manner, depending on the form of the main flight controller:

[0069] Incremental superposition: The main flight controller outputs the nominal rotational speed increment. The final rotor speed command is That is, the cumulative zero-space correction amount passes through historical working points Implicit accumulation;

[0070] Alternatively, non-incremental superposition: the absolute value of the nominal rotational speed output by the main flight controller. The final rotor speed command is That is, the cumulative zero-space correction is maintained through explicit maintenance. accumulation.

[0071] This invention also provides a topology-aware electric vertical takeoff and landing (EVTOL) aircraft control system, which executes the aforementioned topology-aware EVTOL aircraft control method, wherein the EVTOL aircraft has... Distributed electric propulsion unit, Individual battery packs and There are 1 virtual control variable, and the overdrive condition is satisfied. The control system includes the following modules:

[0072] The aircraft status sensing module is used to collect real-time data on the aircraft's speed, angular rate, attitude angle, and... The current rotational speed of each electric propulsion unit;

[0073] The dynamic system state estimation module has the topology matrix built in. Based on the physical mapping relationship, the rotor speed collected by the aircraft state sensing module is used to estimate the current of each motor winding and the discharge current of each battery pack in real time.

[0074] The main flight control module generates virtual control commands based on outer loop commands and aircraft status feedback;

[0075] The basic control allocation module calculates the nominal control quantity that meets the main flight control requirements based on the virtual control command;

[0076] The zero-space current regulation module receives the outputs of the power system state estimation module and the basic control allocation module, and constructs the composite cost function. Calculate the gradient Through the null space projection operator Generate single-step zero-space rotational speed increment And maintain the cumulative zero-space speed correction amount ;

[0077] The fault detection and management module monitors motor or battery pack faults and provides a set of healthy actuator indexes to the zero-space current regulation module. This triggers the downgrade processing step;

[0078] The actuator instruction synthesis module will convert the accumulated zero-space rotational speed correction amount into... The nominal control quantity output by the basic control distribution module is superimposed to form the final rotor speed command, which is then sent to each motor controller.

[0079] The present invention also provides an electric vertical takeoff and landing aircraft, including... Distributed electric propulsion unit, Individual battery packs and There are 1 virtual control variable, and the overdrive condition is satisfied. Each battery pack is connected to an electric propulsion unit that is symmetrically distributed with respect to the center of gravity of the aircraft, and also includes a topology-aware electric vertical take-off and landing aircraft control system.

[0080] This invention differs from existing zero-space control allocation methods by using a motor-battery powered topology matrix. By explicitly incorporating control distribution design and employing a smooth, differentiable maximum approximation function to directly target the maximum motor current and maximum battery pack current, this invention enables the sensing and regulation of current distribution at the battery pack level, providing direct suppression of peak currents. Compared to existing technologies, the advantages and positive effects of this invention are:

[0081] (1) The present invention achieves strict decoupling between peak current regulation and main flight control mission by superimposing the cumulative zero space control increment to the nominal control quantity, without affecting the lift and attitude control accuracy of the aircraft;

[0082] (2) The present invention utilizes the composite cost function and the null-space projection gradient descent method to significantly reduce the peak current of the motor and battery, and the effect is better than the traditional null-space method which aims to minimize the control input norm;

[0083] (3) The present invention optimizes the motor current and battery current simultaneously by adjusting the weighting factor, effectively avoiding stress transfer caused by the optimization of a single component;

[0084] (4) The present invention can automatically adapt the remaining healthy actuators through reduced-order zero-space projection under the condition of motor or battery failure, continue to provide effective current protection, and reduce the overload risk of the remaining components after the failure.

[0085] (5) Based on Lyapunov stability theory, the cost function is monotonically decreasing under static conditions to ensure asymptotic convergence to local optimum. Under dynamic conditions, appropriate selection of gain can make the correction term dominate the disturbance term and ensure current reduction. Attached Figure Description

[0086] Figure 1 This is a flowchart of a topology-aware electric vertical takeoff and landing (EVTOL) control method according to the present invention.

[0087] Figure 2 This is a logical architecture diagram of a topology-aware electric vertical takeoff and landing aircraft control system according to the present invention.

[0088] Figure 3 The motor-battery powered topology matrix of the topology-aware electric vertical takeoff and landing aircraft control method in this invention. Schematic diagram of a centrally symmetrical isolated power supply architecture.

[0089] Figure 4 This is a schematic diagram of the fault reduction process in the topology-aware electric vertical takeoff and landing aircraft control method of the present invention. Detailed Implementation

[0090] The technical solutions of 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.

[0091] Example 1

[0092] like Figures 1 to 4 As shown, the topology-aware electric vertical takeoff and landing (EVTOL) control method of this embodiment has Distributed electric propulsion unit, Individual battery packs A virtual control variable, satisfying the overdrive condition. Taking an electric vertical takeoff and landing (EVTOL) aircraft as an example, the virtual control variables include vertical acceleration. With triaxial angular acceleration In accordance with the isolated power supply requirements stipulated by aviation safety regulations, the eight distributed motors are each connected to four independent battery packs, with each battery pack supplying power to only two specific motors; specifically:

[0093] When the rotor rotates, the air generates a drag torque. The motor must output torque to overcome this drag torque, and the greater the motor's output torque, the greater the current required. Therefore, the higher the rotor speed, the greater the drag torque and the greater the current. Combining this with the fluid dynamics principle that the propeller's drag torque is proportional to the square of its rotational speed, a physical mapping relationship is established from the control input of the square of each rotor's rotational speed to the heating current of each permanent magnet synchronous motor winding and the discharge current of each battery pack. This physical mapping relationship includes a motor-battery topology matrix. The elements of the topological matrix Characterizing the first The battery pack is the first The power supply relationship of the motors, as described in the topology matrix in this embodiment. It adopts a binary isolation power supply form, which is a symmetrical isolation power supply architecture, that is Each motor is exclusively powered by a single, dedicated battery pack, specifically:

[0094] ;

[0095] In this system, the rows of the matrix correspond to four battery packs, and the columns correspond to eight motors. The first battery pack connects to the first and eighth motors, the second battery pack connects to the second and seventh motors, the third battery pack connects to the third and sixth motors, and the fourth battery pack connects to the fourth and fifth motors. The centrally symmetrical isolated power supply architecture further ensures that the motors connected to each battery pack are centrally symmetrically distributed relative to the center of gravity of the aircraft. This ensures that when any battery pack fails, the motor that loses power remains centrally symmetrical relative to the center of gravity of the aircraft, without generating additional unbalanced torque. The system also reduces the current impact of each control action on each battery pack, preventing the control distributor from misunderstanding the topology and thus overloading any particular battery pack.

[0096] Based on this, the control method includes the following steps:

[0097] S1. Constructing physical mapping relationships, specifically including:

[0098] S1.1 Calculate the motor winding current:

[0099] ;

[0100] in:

[0101] For the first The effective value of the phase current of each motor reflects the Joule heat intensity, that is, directly reflects the heat intensity of the motor, and the unit is A;

[0102] For the first The square of the rotor speed, , For the first Since both drag torque and thrust are proportional to the square of the rotor speed, the square of the rotor speed is chosen to simplify subsequent calculations. The unit is ;

[0103] Let be the motor current constant. It integrates data such as propeller torque coefficient, gearbox reduction ratio, motor pole pair number, and permanent magnet flux linkage, which can be obtained directly from the product manual. As long as the motor and propeller are not replaced, For fixed, among which: This is the propeller torque coefficient, in units of... ; The reduction ratio of the gearbox is dimensionless. is the number of pole pairs of the motor, and is dimensionless; For permanent magnet flux linkage, the unit is 1. The propeller torque coefficient, gearbox reduction ratio, motor pole pair number, and permanent magnet flux linkage can all be obtained directly from the product manual without the need for additional measurements.

[0104] In the above content, as long as the rotor speed is known, the heating current of the motor can be calculated immediately, which provides a basis for controlling the current by adjusting the speed;

[0105] S1.2 After the motor draws power from the battery, in addition to providing the mechanical power for rotating the propeller, it also consumes some electrical energy as heat. Therefore, the current flowing from the battery into the motor controller is larger than the current corresponding to pure mechanical power. The DC power supply current of the motor branch is determined by both the electromechanical power and copper losses, specifically:

[0106] ;

[0107] in:

[0108] For the first The DC power supply current of each motor branch, in A;

[0109] Let be the inverter efficiency, and be dimensionless.

[0110] This is the DC bus voltage, in V.

[0111] The resistance of each phase of the motor is expressed in Ω.

[0112] The above content demonstrates that the higher the rotor speed, the faster the battery output current increases not only due to the increased mechanical load, but also due to the increased heat generation, providing a basis for suppressing peak current to protect the battery.

[0113] S1.3 Finally, the battery pack discharge current is calculated by aggregating the branch currents through the topology matrix, specifically as follows:

[0114] ;

[0115] in:

[0116] The motor-battery topology matrix is ​​dimensionless.

[0117] This is the DC power supply current vector of the motor branch. The unit is A;

[0118] This is the battery pack discharge current. The unit is A;

[0119] The above content quantifies the current impact of each control action on each battery pack, avoiding the situation where the control distributor does not understand the topology and thus overloads a certain battery pack.

[0120] S2. The main flight controller generates virtual control commands to maintain the attitude and altitude of the aircraft based on the outer loop commands and the aircraft state feedback. The basic control allocation algorithm calculates the corresponding nominal control quantity based on the virtual control commands. The main flight controller adopts an incremental nonlinear dynamic inverse (INDI) control architecture, and the basic control allocation algorithm adopts a redistribution pseudo-inverse (RPI) algorithm. Specifically:

[0121] S2.1 First, the reference model generates the desired output dynamics based on the outer loop instructions:

[0122] ;

[0123] in:

[0124] The time derivative of the reference model output;

[0125] The gain matrix is ​​the reference model.

[0126] This is an outer loop instruction;

[0127] Output for reference model;

[0128] S2.2 Then, combining the feedforward and output feedback, the desired pseudo-control quantity is obtained:

[0129] ;

[0130] in:

[0131] This is the desired pseudo-control quantity;

[0132] The tracking error feedback gain matrix;

[0133] To output the estimated measurement value;

[0134] S2.3, Subtract the current acceleration estimate to obtain the virtual control increment command:

[0135] ;

[0136] For virtual control incremental instructions;

[0137] This is an estimate of the acceleration (pseudo-control variable) at the current operating point;

[0138] S2.4 Iteratively solve for the nominal speed increment using the redistribution pseudo-inverse algorithm. The specific iterative process is as follows:

[0139] S2.4.1 Initialize the active set All executors are initially considered active, and the initial solution is set to... ;

[0140] S2.4.2, In the currently active set Using pseudo-inverse to find residuals Corresponding minimum norm increment ;

[0141] S2.4.3, Check the updated Whether the actuator position or rate constraint is violated, i.e. ,in For the embedding mapping of active sets to full-dimensional space, Each component is filled in according to its active set index. The corresponding position of the full-dimensional vector is set to zero at the inactive set index; if a violation occurs, the corresponding executor is removed from the active set and saturated to the nearest constraint boundary, and the current constraint is updated. ;

[0142] S2.4.4, Repeat steps S2.4.2-S2.4.3 until all actuators satisfy the constraints or the active set is empty; the final solution is denoted as... ;

[0143] Through the above iterative process, the reallocation pseudo-inverse algorithm, under the premise of satisfying the physical constraints of the actuator, will convert the virtual control incremental instructions... Converted into an executable nominal rotor speed increment ;

[0144] Further:

[0145] 1) The main flight controller is any controller that generates the virtual control commands based on aircraft state feedback and outer loop commands, including but not limited to:

[0146] The incremental nonlinear dynamic inverse controller generates the desired output dynamic through a reference model, obtains the desired pseudo-control quantity by combining feedforward and output feedback, and obtains the virtual control incremental command by subtracting the current virtual control feedback estimate.

[0147] The non-incremental nonlinear dynamic inverse controller calculates the virtual control command directly by inverting the model.

[0148] The model predictive controller will use the composite cost function As an additional term to the prediction cost function;

[0149] Proportional-integral-derivative controller;

[0150] Controllers within a robust or adaptive control framework;

[0151] 2) The basic control allocation algorithm is any algorithm capable of calculating the nominal control quantity that satisfies the physical constraints of the actuator based on the virtual control instruction, including but not limited to:

[0152] The pseudo-inverse algorithm for redistribution, through Calculate the nominal speed increment, and overflow the saturated input from the active set and redistribute the residual error to the remaining input;

[0153] Weighted least squares allocation algorithm;

[0154] The direct allocation algorithm solves for the maximum achievable torque set within the convex hull spanned by the column vectors of the control performance matrix;

[0155] Allocation algorithms based on linear programming or quadratic programming.

[0156] S3. Construct the motor current sub-cost function and the battery current sub-cost function using a smooth and differentiable maximum approximation function, respectively. Then, weightedly sum the motor current sub-cost function and the battery current sub-cost function to form a composite cost function, which is used to simultaneously suppress the peak current of the motor and battery electrical components. The composite cost function is as follows:

[0157] ;

[0158] in:

[0159] The composite cost function is dimensionless.

[0160] and It is an adjustable weighting factor used to balance motor protection and battery protection, and is dimensionless;

[0161] This is a normalized reference value for the motor current, in A.

[0162] This is a normalized reference value for battery current, in A.

[0163] It is a smooth maximum value approximation function for the motor current;

[0164] It is a smooth maximum approximation function for the battery current.

[0165] In this embodiment, the Kreisselmeier-Steinhauser (KS) function is selected as the smooth maximum approximation function, then:

[0166] ;

[0167] ;

[0168] in:

[0169] The current maximum motor current The unit is A;

[0170] This is the current maximum battery current. The unit is A;

[0171] This is the sharpness parameter for smooth approximation; a larger value results in higher approximation accuracy. ;

[0172] It is a natural exponential function; It is the natural logarithm function;

[0173] By using a composite cost function to reflect the peak current of the motor and battery, the specific parameters of the peak current of the motor and battery are quantified, providing a basis for subsequent reduction of the peak current of the motor and battery.

[0174] S4. Calculate the rate of change gradient of the composite cost function as a function of the square of the rotor speed, and determine the contribution weights of each motor current and each battery pack current to the current peak value based on the rate of change gradient. The specific process for calculating the rate of change gradient is as follows:

[0175] ;

[0176] in:

[0177] The composite cost function is applied to the control input vector. The gradient;

[0178] Let be the gradient of the cost function of the motor current. ,in: This is the Softmax weight vector of the motor current. , is dimensionless;

[0179] The gradient of the battery current cost function. ,in: The softmax weight vector of the battery current. , is dimensionless; The Jacobian matrix of the branch power supply current versus the control input;

[0180] In this embodiment, specifically, the elements of the Softmax weight vector are:

[0181] , ;

[0182] in:

[0183] For the first The Softmax weights of each motor;

[0184] For the first The Softmax weights of each battery pack;

[0185] and Automatically concentrate the gradient on the component with the highest current.

[0186] The gradient formula was used to calculate the adjustment of the rotor speed for each rotor when reducing the most dangerous peak current.

[0187] S5. Construct a null-space projection operator based on the control effectiveness matrix determined by aircraft dynamics, projecting the negative direction of the rate of change gradient into the null space to generate a null-space rotational speed increment that does not change the main flight control effect. The null-space projection operator is:

[0188] ;

[0189] in:

[0190] is the null space projection matrix, which is dimensionless and can extract the portion of any vector that lies in the null space;

[0191] for The identity matrix, In this embodiment, the number of actuators is... ;

[0192] To control the performance matrix, reflecting the impact of changes in rotor speed on the aircraft's vertical and angular acceleration, In this embodiment, p=4, that is ;

[0193] for Moore-Penrose pseudo-inverse;

[0194] In this embodiment, specifically, the single-step zero-space rotational speed increment is:

[0195] ;

[0196] in:

[0197] For the first The single-step zero-space control speed vector for each control cycle, in units of... ;

[0198] This is the gain coefficient, used to adjust the current suppression rate. , If it's too small, the current will decrease slowly. Too high a value may cause sudden changes in rotation speed, so a moderate value should be chosen to allow the current to decrease smoothly within 1-2 seconds. (Unit: ...) ;

[0199] The single-step zero-space rotational speed increment satisfies It does not generate additional force or torque;

[0200] In this embodiment, specifically, the single-step zero-space rotational speed increment Accumulated over the time dimension, the accumulated zero-space rotational speed correction is obtained:

[0201] , ;

[0202] Due to the null space projection matrix satisfy Therefore, the increment at each step and the correction amount formed by its accumulation All meet That is, the cumulative zero-space rotational speed correction does not change the force and torque expected to be generated by the main flight controller.

[0203] In the above, "current regulation without affecting flight" is achieved through zero-space projection. The control commands issued by the pilot or autopilot are still executed by the basic distributor. The design superimposes a fine adjustment in zero space, and the two effects do not interfere with each other.

[0204] S6. Adjust the cumulative zero-space rotational speed correction amount. The nominal control input from the main flight controller is added together to output the final rotor speed command. The peak current of the power system is actively reduced without affecting attitude control accuracy. Depending on the type of main flight controller used, the cumulative zero-space rotational speed correction is... The superposition of the nominal control quantity output by the main flight controller can be implemented in any of the following ways:

[0205] Incremental overlay (applicable to incremental flight control architectures, such as INDI): The main flight controller outputs the nominal speed increment. The final rotor speed command is That is, the cumulative zero-space correction amount passes through historical working points Implicit accumulation.

[0206] Non-incremental superposition (applicable to non-incremental flight control architectures such as NDI, MPC, PID): The main flight controller outputs the absolute value of the nominal rotational speed. The final rotor speed command is That is, the cumulative zero-space correction is maintained through explicit maintenance. accumulation.

[0207] Due to control variables It is the square of the rotor speed. The final rotor speed command needs to be... Rotor speed conversion command The speed command is then limited to between the physically permissible maximum and minimum values; finally, the speed command is sent to the motor controller of each rotor for execution via pulse width modulation or serial communication.

[0208] This embodiment preferably also includes handling for motor or battery malfunctions, such as... Figure 4 As shown, it specifically includes:

[0209] After detecting the faulty components, determine the set of healthy actuator indexes that are still functioning normally. Then from the original control effectiveness matrix Extract the corresponding columns to construct a reduced-order matrix:

[0210] ;

[0211] in:

[0212] The reduced-order null space projection matrix has dimensions of ;

[0213] for 3D identity matrix;

[0214] Then calculate the reduced-order single-step zero-space rotational speed increment:

[0215] ;

[0216] in:

[0217] For order reduction, single-step zero-space rotational speed increment;

[0218] For the gradient vector corresponding to The amount;

[0219] Finally, the reduced-order single-step zero-space rotational speed increment will be... Mapping back to full-dimensional vector The component not corresponding to the healthy actuator is set to zero; the subsequent accumulation and superposition process is the same as steps S5 to S6 in this embodiment, that is, the single-step zero-space rotational speed increment. Press the same way The final rotor speed command is accumulated and synthesized by superposition in step S6; during a fault, the method can still use the remaining healthy actuators to continuously suppress the peak current.

[0220] Since damage to the motor or battery pack can lead to a reduction in zero space, the system first searches for healthy actuators, then reduces the order of the zero space projection matrix, recalculates the reduced-order zero space projection matrix, calculates the adjustment amount for the reduced order, and then fills it back into all dimensions. This allows the remaining available rotors to optimize the current distribution within a certain range, avoiding overloading of the remaining motors.

[0221] In this embodiment, by establishing a physical mapping from rotor speed to motor current and battery current, the requirement to suppress peak current is expressed as a smooth and differentiable composite cost function, and its gradient is calculated to indicate the most effective speed adjustment direction. Then, using the null space of the aircraft's overdrive system, this gradient direction is projected into the null space to generate a single-step null space speed increment that is completely transparent to the main flight control. This increment is accumulated over time to form a cumulative null space speed correction, which is then superimposed on the nominal control quantity. This achieves a continuous reduction in the maximum current in each motor and battery pack without affecting any attitude control or altitude tracking accuracy, thereby actively mitigating Joule heat accumulation and reducing the risk of thermal runaway. At the same time, by adjusting the weights to balance the protection of the motor and battery, stress transfer is effectively avoided. After a motor or battery failure, the remaining healthy actuators are automatically adapted through reduced-order null space projection, and the remaining healthy redundancy continues to provide current protection, greatly improving the safety and reliability of the power system.

[0222] Example 2

[0223] The technical difference between the topology-aware electric vertical takeoff and landing (EVTOL) control method in this embodiment and the method in Embodiment 1 lies in the following: In this embodiment, when constructing the motor current sub-cost function and the battery current sub-cost function, an implementation form other than the KS function is adopted in the smooth, differentiable maximum approximation function, including but not limited to:

[0224] 1) Construct the cost function using the LogSumExp function:

[0225] ;

[0226] ;

[0227] in The sharpness parameter is also infinitely differentiable, and its gradient is in the form of Softmax.

[0228] 2) Approximation using the q-norm:

[0229] ;

[0230] ;

[0231] in For integers greater than 1, when Time approaching The gradient of the q-norm approximation is This also achieves the effect of concentrating the gradient on the high current component.

[0232] The other technical features of this embodiment are exactly the same as those in Embodiment 1. Due to the use of different smooth maximum approximation functions, the numerical stability of this embodiment may be slightly different, but it can still achieve the technical effect of reducing the peak current of the motor and battery without affecting the main flight control.

[0233] Example 3

[0234] The technical difference between the topology-aware electric vertical takeoff and landing (EVTOL) control method of this embodiment and the method in Embodiment 1 lies in the following: In this embodiment, when calculating the nominal control quantity, a basic control allocation algorithm other than the redistribution pseudo-inverse algorithm is used, including but not limited to:

[0235] Weighted Least Squares Allocator: Solving composite optimization problems, specifically:

[0236] ;

[0237] in: This is the virtual control error weight matrix; To control the input weight matrix; To control the scalar weights, ;

[0238] Direct allocation method: The maximum reachable torque set is directly solved within the convex hull spanned by the column vectors of the control performance matrix, and then the minimum control quantity is determined through iteration;

[0239] Assignors based on linear or quadratic programming: The control assignment problem is expressed as a linear or quadratic programming problem, and actuator saturation constraints are handled explicitly.

[0240] The other technical features of this embodiment are exactly the same as those in Embodiment 1. Regardless of which basic allocation algorithm is used, the output is a nominal control quantity that satisfies the main control objective and obeys the physical constraints of the actuator. Then, the zero-space current regulation module still superimposes the cumulative zero-space speed correction quantity based on this nominal solution. This allows peak current to be suppressed without affecting the main control task.

[0241] This embodiment is applicable to flight controllers with different computing resources and real-time requirements, and is especially suitable for system integration with existing specific control allocation algorithm libraries.

[0242] Example 4

[0243] The key difference between the topology-aware electric vertical takeoff and landing (EVTOL) control method in this embodiment and the method in Embodiment 1 lies in the following: In this embodiment, the weighting factor in the composite cost function... and The ways in which values ​​are obtained are different, such as:

[0244] Motor priority configuration: Settings , At this point, the composite cost function only includes the motor current sub-cost. It is suitable for scenarios where motor thermal limitation is the system bottleneck;

[0245] Battery preference settings: , At this point, the composite cost function only includes the battery current sub-cost. It is suitable for scenarios where battery thermal management is a critical constraint;

[0246] Adaptive weight configuration: Defines motor current margin With battery current margin ,in and These are the rated currents of the motor and the battery pack, respectively. and Adaptive adjustment as follows:

[0247] , ;

[0248] When a certain type of component approaches its rated limit, its corresponding weight automatically increases, making the control more focused on protecting the most dangerous component.

[0249] The other technical features of this embodiment are exactly the same as those in Embodiment 1. This embodiment expands the application scope of the invention by flexibly configuring weights to adapt to different flight mission phases or different thermal management requirements.

[0250] Example 5

[0251] The key difference between the topology-aware electric vertical takeoff and landing (EVTOL) control method in this embodiment and the method in Embodiment 1 lies in the step size of the null rotation speed increment. Instead of being fixed as a constant, an adaptive gain strategy is adopted, such as:

[0252] Adaptive step size based on line search: In each control cycle, an exact line search is performed with the current gradient direction to find the optimal step size that satisfies the Armijo criterion or the Wolfe condition. This strategy ensures the maximum decrease in cost function at each step, but it involves a large amount of computation.

[0253] Proportional gain based on current deviation: Defines the deviation between the current peak current and the safety threshold. The gain is ,in It is a monotonically increasing function. The more the current exceeds the threshold, the larger the step size becomes to quickly reduce the overload; when the current is below the threshold, the step size decreases to smooth the adjustment.

[0254] Attenuation gain: Setting ,in The number of iterations. The gain gradually decreases as the iteration proceeds, which helps to improve steady-state accuracy and avoid oscillations near the optimal solution.

[0255] The other technical features of this embodiment are exactly the same as those in Embodiment 1. In this embodiment, an appropriate gain strategy can be selected according to the specific task requirements: proportional gain is selected for high dynamic scenes to respond quickly, and attenuation gain is selected for high-precision hovering scenes to adjust smoothly.

[0256] Example 6

[0257] The key difference between the topology-aware electric vertical takeoff and landing (EVTOL) control method in this embodiment and the method in Embodiment 1 is that the aircraft used in this embodiment is not limited to an octocopter, four-battery configuration, but is extended to other configurations and topology connections, such as:

[0258] Hexarotor with three batteries: The aircraft is equipped with six distributed motors and three independent battery packs. Each battery pack powers two symmetrically distributed motors. The topology matrix is ​​as follows. The specific form is determined according to the principle of centroidal symmetry;

[0259] Twelve rotors and six batteries: The aircraft is equipped with twelve motors and six battery packs, with each battery pack powering two motors. The topology matrix is ​​expanded as follows: The upper limit of summation in the other formulas will be adjusted accordingly.

[0260] Tiltrotor configuration: The motors not only provide vertical lift, but can also tilt to provide forward thrust. In this case, the control efficiency matrix... No longer fixed, but needs to be adjusted according to the current tilt angle. The result is obtained by relinearizing at this tilt angle, and the updated result is... For null projection operator The construction remains unchanged; the null space projection method remains the same;

[0261] Hybrid power configuration: A multi-energy system including fuel cells and battery packs. In this case, the topology matrix needs to be extended to include the fuel cell output power in the energy flow mapping, establishing a complete link from the control input to the output current of each energy source. A composite cost function can increase the fuel cell current sub-cost, for example: ,in: Let the discharge current vector of each fuel cell stack be denoted as . , This refers to the number of fuel cell stacks; For a smooth and differentiable maximum approximation function, the total composite cost function becomes: ,in: This is a weighting factor for the cost of fuel cells; This serves as a normalized reference value for the fuel cell current; the zero-space projection method remains unchanged, with only the fuel cell current term added to the gradient calculation.

[0262] The other technical features of this embodiment are exactly the same as those in Embodiment 1. This embodiment shows that no matter how the number of motors and the battery layout change, as long as a corresponding topology matrix is ​​established... And update the control performance matrix. The zero-space current regulation method proposed in this invention can be directly applied to all dimensions.

[0263] Example 7

[0264] The key difference between the topology-aware electric vertical takeoff and landing (EVTOL) control method in this embodiment and the method in Embodiment 1 is that the main flight controller in step S2 of this embodiment is not limited to incremental nonlinear dynamic inverse, but adopts other control architectures, including but not limited to:

[0265] Non-incremental nonlinear dynamic inverse: directly use the model inverse to calculate the required control quantity. Instead of an incremental form, this architecture requires high model accuracy, but the zero-space current control method is equally effective: it simply uses the control quantity calculated by the non-incremental dynamic inverse as... Then, the cumulative zero-space speed correction is superimposed. ;

[0266] Model predictive control: Zero-space current regulation is added as an additional term to the model predictive control cost function. Specifically, the model predictive control optimization problem is: ,in: The weight of the current suppression term, ; As the composite cost function of the present invention, the optimization problem further includes actuator physical constraints; this form unifies the main control objective and the current suppression objective into a single optimization problem, without the need to separate the nominal solution and the null space solution;

[0267] Traditional PID+ control allocation architecture: The outer loop uses a PID controller to generate force and torque commands, while the inner loop uses the control allocation method of this invention. This architecture is suitable for systems with low real-time requirements or existing mature PID controllers.

[0268] Robust or adaptive control framework: Explicitly consider model uncertainties and parameter perturbations in the controller design, and then integrate a zero-space current regulation module in its inner-loop control distribution section;

[0269] The other technical features of this embodiment are exactly the same as those in Embodiment 1. This embodiment shows that as long as the control system has overdrive redundancy, the control performance matrix can be optimized. Since the zero space is non-trivial, the zero space projection current control method of the present invention can be used in conjunction with it.

[0270] Example 8

[0271] like Figures 2 to 3 As shown, the topology-aware electric vertical takeoff and landing (EVTOL) control system of this embodiment executes the topology-aware EVTOL control method of any of the embodiments 1-7, to achieve the following: Distributed electric propulsion unit, Individual battery packs A virtual control variable, satisfying the overdrive condition. Taking an electric vertical takeoff and landing (EVTOL) aircraft as an example, the virtual control variables include vertical acceleration. With triaxial angular acceleration In accordance with the isolation power supply requirements of aviation safety regulations, the eight distributed motors are connected to four independent battery packs, and each battery pack supplies power to only two specific motors.

[0272] The topology-aware electric vertical takeoff and landing aircraft control system of this embodiment specifically includes the following modules:

[0273] Aircraft Status Sensing Module: The aircraft status sensing module is used to collect the aircraft's speed, angular rate, attitude angle, and the current rotational speed of each rotor in real time. By providing accurate measurement data of the aircraft's current motion state and rotor operating point, it provides the most basic input for subsequent control allocation and current regulation. The aircraft status sensing module can be composed of an inertial measurement unit, a satellite positioning receiver, a barometric altimeter, and a rotor speed sensor. The inertial measurement unit is used to measure the aircraft's three-axis angular rate and linear acceleration. Combined with satellite positioning and the barometric altimeter, the speed, attitude angle, and rate of climb are calculated. The rotor speed sensor is used to provide real-time feedback on the current rotational speed of each rotor. The inertial measurement unit, satellite positioning receiver, barometric altimeter, and rotor speed sensor send data to the control system at a fixed frequency.

[0274] Dynamic system state estimation module: The dynamic system state estimation module has a built-in topology matrix. It is used to calculate the current of each motor winding and the discharge current of each battery pack in real time based on the rotor speed and the physical mapping relationship. The motor winding current and battery pack discharge current, which are difficult to measure directly, are estimated indirectly and in real time through known physical formulas and easily measurable rotor speed. By using the current values ​​of the motor winding current and battery pack discharge current, it can determine which motor or battery pack is currently experiencing excessive current stress, thereby determining the target that needs to be suppressed in the future.

[0275] The main flight control module is used to output the virtual control commands based on the outer loop commands and flight state measurements. The virtual control commands are force and torque-related quantities required for aircraft attitude and altitude control. The main flight control module can be any controller capable of generating the virtual control commands based on aircraft state feedback and outer loop commands, including but not limited to incremental nonlinear dynamic inverse controllers, non-incremental nonlinear dynamic inverse controllers, model predictive controllers, proportional-integral-derivative controllers, etc. In this embodiment, the main flight control module uses an incremental nonlinear dynamic inverse control law to calculate the virtual control increment. Specifically, the flight control module receives outer-loop commands and aircraft state feedback, generates the desired smooth dynamics using a reference model, and then combines proportional-integral-derivative feedback correction to obtain the desired pseudo-control quantity. Then use pseudo-control quantity Subtract the currently measured actual acceleration To obtain virtual control increment ;

[0276] Basic control allocation module: The basic control allocation module is used to calculate the nominal control quantity that meets the main flight control requirements according to the virtual control command. The algorithm used by the basic control allocation module is any algorithm that can calculate the nominal control quantity that meets the actuator physical constraints according to the virtual control command, including but not limited to the redistribution pseudo-inverse algorithm, weighted least squares allocation algorithm, direct allocation algorithm, allocation algorithm based on linear programming or quadratic programming, cascaded generalized inverse method, etc. In this embodiment, the basic control allocation module receives the virtual control increment from the main flight control module. and the current control input working points of each rotor. The nominal speed increment is solved iteratively using a redistribution pseudo-inverse algorithm. This results in a physically feasible system that meets the requirements of primary flight control. This provides a starting point for subsequent zero-space optimization;

[0277] Zero-space current regulation module: The zero-space current regulation module receives the outputs of the power system state estimation module and the basic control allocation module, calculates and generates a single-step zero-space speed increment, and maintains the cumulative zero-space speed correction to suppress peak current; specifically, the zero-space current regulation module obtains the current of each motor from the power system state estimation module. and the current of each battery pack Obtain the current working point from the basic control allocation module. And the nominal control quantity; then construct the composite cost function using a smooth, differentiable maximum approximation function. Then calculate the gradient and the corresponding Softmax weights; then based on the control performance matrix and the set of health actuators provided by the fault detection and management module. Construct the null projection matrix or reduced-order projection matrix This leads to the generation of a single-step zero-space rotational speed increment. and through (Initial value) Accumulated zero-space speed correction amount Because the null projection satisfies The cumulative zero-space rotational speed correction does not change the force and torque expected to be generated by the main flight controller; due to the differentiability of the smooth maximum approximation function and the concentrated characteristics of the Softmax weight, the rotor speed adjustment process is smooth and oscillating, and can handle multiple components close to the peak value at the same time, avoiding stress transfer.

[0278] Fault Detection and Management Module: This module monitors motor or battery faults and provides a set of healthy actuator indexes to the zero-space current regulation module. This is to enable the execution of the downgrade processing steps; the fault detection and management module continuously monitors the health status of each motor and each battery pack, including comparing the actual motor current with the expected current, monitoring the battery pack voltage and temperature, and receiving fault signals from the battery management system; once a fault is detected, the fault detection and management module immediately generates the set of healthy actuator indexes. The data is transmitted in real time to the zero-space current control module; simultaneously, the fault detection and management module will issue an alarm to the pilot or ground station and execute safety protection strategies when necessary; when a fault occurs, the zero-space current control module can still rely on the healthy actuator set. Recalculate the reduced-order null projection matrix And gradient, thereby continuing to provide current optimization capabilities in reduced-order mode;

[0279] Actuator command synthesis module: The actuator command synthesis module is used to adjust the cumulative zero-space rotational speed according to the cumulative zero-space rotational speed correction amount. The nominal control quantity output by the basic control distribution module is superimposed on the nominal control quantity output by the main flight controller and the cumulative zero-space speed correction quantity output by the zero-space current regulation module. The actuator command synthesis module synthesizes and outputs the final rotor speed command. Depending on the type of main flight controller used, the module supports two superposition methods: incremental superposition (suitable for incremental flight control architectures such as INDI) and non-incremental superposition (suitable for non-incremental flight control architectures).

[0280] Incremental superposition of nominal speed increments output by the basic control distribution module Single-step zero-space speed increment output by the zero-space current control module Calculate the new control input operating point:

[0281] ;

[0282] The absolute value of the nominal speed output by the non-incremental superposition receiving basic control distribution module. The cumulative zero-space speed correction maintained by the zero-space current control module Directly superimposed:

[0283] ;

[0284] Then due to the control variable It is the square of the rotor speed. The above needs to be Rotor speed conversion command The speed command is then limited to the physical maximum and minimum values. Finally, the speed command is sent to the motor controller of each rotor in the form of pulse width modulation or serial communication for execution. The actuator command synthesis module completes the final conversion from abstract control increments to actual motor executable commands.

[0285] In this embodiment, the aircraft state sensing module collects the aircraft's speed, angular rate, attitude angle, and current rotor speed through an inertial measurement unit, satellite positioning, and speed sensors, providing a real-time motion state reference for subsequent control. The power system state estimation module has a built-in topology matrix describing the connection relationship between the motor and the battery pack, and calculates the current of each motor winding and the discharge current of each battery pack in real time based on the analytical mapping formula between the square of the rotor speed and the motor current and the branch power supply current, enabling the control system to understand the current electrical load distribution. The main flight control module uses a main flight controller (preferred in this embodiment, an incremental nonlinear dynamic inverse control law) to generate virtual control increments based on outer loop commands and state feedback. The virtual control increments represent the force and torque changes required by the aircraft, but do not specify specific rotor actions. The basic control allocation module converts the received virtual control commands into nominal control quantities through a basic control allocation algorithm (preferred in this embodiment, a redistribution pseudo-inverse algorithm). The nominal control quantities are optimized under the premise of satisfying the physical constraints of each rotor. First, ensure the achievement of the main flight control task; the zero-space current regulation module receives the power system state estimation results and basic control allocation results, calculates the gradient of the composite cost function with respect to the control input, and projects the negative direction of the gradient into the zero space of the control effectiveness matrix to obtain the single-step zero-space speed increment, and forms the cumulative zero-space speed correction through time accumulation. The cumulative zero-space speed correction does not generate additional force or torque, thus not affecting the flight attitude, and can continuously reduce the peak current of the motor and battery; the fault detection and management module monitors the fault status of the motor or battery, provides the healthy actuator index to the zero-space current regulation module, so that the fault detection and management module automatically switches to the reduced-order zero-space projection mode when a fault occurs, and continues to implement current suppression using the remaining healthy rotors; the actuator command synthesis module synthesizes the nominal control quantity and the cumulative zero-space speed correction quantity in the superposition method of the cumulative zero-space speed correction quantity and the nominal control quantity output by the main flight controller, and converts it into the actual speed command output of each rotor.

[0286] Example 9

[0287] The electric vertical takeoff and landing aircraft in this embodiment includes: Distributed electric propulsion unit, Individual battery packs and There are 1 virtual control variable, and the overdrive condition is satisfied. Each battery pack is connected to an electric propulsion unit that is symmetrically distributed with respect to the center of gravity of the aircraft. It also includes the topology-aware electric vertical take-off and landing aircraft control system in embodiment 8.

[0288] In this embodiment, specifically, the aircraft is equipped with a flight controller, which includes:

[0289] Memory: Used to store computer programs and the topology matrix required for their operation. Control effectiveness matrix Motor current constant Normalized reference values and Gain coefficient Parameters, etc.; the memory includes, but is not limited to, DDR4 SDRAM or equivalent random access memory;

[0290] Sensor interfaces include, but are not limited to, interfaces such as CAN bus, RS-485 serial port, SPI bus, and I2C bus, used to receive flight status signals output by inertial measurement unit, satellite positioning receiver, barometric altimeter, and rotor speed sensor;

[0291] Actuator interfaces include, but are not limited to, pulse width modulation (PWM) output channels, CAN bus, DroneCAN bus, etc., which are used to send the final rotor speed command to each motor controller.

[0292] In this embodiment, the flight controller can be an integrated form of an independent flight control board, or it can be a software porting form based on a general embedded computing platform.

[0293] Specifically, in this embodiment, the computer-readable storage medium includes, but is not limited to, read-only memory, flash memory, solid-state drive, hard disk drive, optical disk, and cloud-distributed storage media. When the computer program is loaded into the memory of the flight controller and executed by the processor, all steps of the topology-aware electric vertical takeoff and landing aircraft control method in any of the embodiments 1-7 are implemented.

[0294] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A topology-aware control method for an electric vertical takeoff and landing (EVTOL) aircraft, wherein the EVTOL aircraft has... Distributed electric propulsion unit and Each independent battery pack is involved in the attitude and altitude control of the aircraft. A virtual control variable, comprising vertical acceleration and triaxial angular acceleration, wherein the control method satisfies the overdrive condition. The m electric propulsion units and n battery packs are connected via a power supply topology mapping relationship; characterized in that... The control method includes generating virtual control commands by the main flight controller based on outer loop commands and aircraft state feedback, and calculating the nominal control quantity corresponding to the virtual control commands using a basic control allocation algorithm. The control method also includes the following steps: Step a: Establish a physical mapping model from control input to electrical load: Establish a model based on the square of the rotational speed of each rotor as the control input. To the heating current of each motor winding and the discharge current of each battery pack The physical mapping relationship For the control input component of the i-th actuator, Let be the rotational speed of the i-th rotor; the physical mapping relationship includes a motor-battery topology matrix. The elements of the topological matrix Characterizing the first The battery pack is the first The power supply relationship of each motor, and the discharge current of each battery pack. satisfy ,in This represents the DC power supply current vector for each motor branch; Step b: Construct the composite cost function of peak current: Construct the sub-cost functions of motor current using smooth and differentiable maximum approximation functions. and battery current quantum cost function and with adjustable weighting factors and Weighted summation forms a composite cost function The composite cost function Regarding the control input vector Every detail matters; Step c: Achieve current suppression through cumulative null-space projection superposition: based on the control effectiveness matrix determined by the aircraft dynamics. Constructing the null space projection operator , ,in, for Moore-Penrose pseudo-reverse, for A unit matrix of order 1; and then perform the following steps in sequence: Step cl, calculate single step null space rotational velocity increment : ,in, For the gain coefficient, and the gain coefficient The value of makes the composite cost function The solution asymptotically converges to a local optimum along the Lyapunov direction of the null space projection gradient. Step c2, accumulating single-step zero-space rotational speed increment in time dimension to obtain accumulated zero-space rotational speed correction amount : ， ； Step c3, the cumulative zero-space rotational speed correction amount satisfies without changing the force and moment desired to be generated by the primary flight controller Step c4, the cumulative null-space rotational speed correction amount The final rotor rotational speed command is obtained by superimposing the nominal control amount on the main flight controller output.

2. The topology-aware based control method for electric vertical take-off and landing aircraft according to claim 1, wherein: The physical mapping relationship in step a is based on a direct-quadrature coordinate system model of the permanent magnet synchronous motor, the direct-axis current ignoring inductance dynamics, ignoring gear box losses and friction torque; the physical mapping relationship comprises: The motor winding current is proportional to the square of the rotor speed, specifically: ； in, Let be the effective value of the phase current of the i-th motor, and be the motor current constant. , This is the propeller torque coefficient. The reduction ratio of the gearbox. This represents the number of pole pairs of the motor. For permanent magnet flux linkage; The DC power supply current of the motor branch is determined by both the electromechanical power and copper losses, specifically: , The supply current for the i-th motor branch, For inverter efficiency, This is the DC bus voltage. The resistance of each phase of the motor; Furthermore, the topology matrix The system adopts a centrally symmetrical isolated power supply architecture, where the elements belong to {0,1}, meaning that each motor is exclusively powered by only one specific battery pack. The centrally symmetrical isolated power supply architecture further ensures that the motors connected to each battery pack are centrally symmetrically distributed with respect to the center of gravity of the aircraft. This ensures that when any battery pack fails, the motor that loses power still maintains central symmetry with respect to the center of gravity of the aircraft and does not generate additional unbalanced torque.

3. The control method of claim 1, wherein: The composite cost function in step b is: wherein, with are normalized reference values for the motor current and battery current, respectively, such that the composite cost function is dimensionless; The motor current sub-cost function With battery current quantum cost function Using smooth, differentiable maximum approximation functions respectively Acting on the motor current vector With battery pack current vector The structures are respectively ; The smooth, differentiable maximum approximation function In order to be able to convert vectors maximum value The accuracy of an arbitrary function that is approximated as a scalar in a smooth and differentiable manner is determined by the sharpness parameter. adjust; The composite cost function The gradient of the control input vector is given by the chain rule: ； in: With According to the analytical form of the smooth maxima approximation function corresponding calculations.

4. The topology-aware based control method for electric vertical take-off and landing aircraft according to claim 3, wherein: Smoothing differentiable maximum approximation function adjusted by KS function approximation accuracy, specifically: ， ； Alternatively, the approximation accuracy of the smooth differentiable maximum approximation function is adjusted by the LSE function is adjusted by the LSE function, in particular: ； Alternatively, the approximation function is obtained by q-norm function smoothing the maximum value of the differentiable function The approximation accuracy is specifically: When using the KS function or the LSE function, the gradient is specifically expressed as follows: ， ； wherein with are the Softmax weight vectors for the motor current and battery current, respectively, whose elements are: ， ； When other smooth maximum approximation functions are used, the gradient is calculated by chain expansion of the analytical partial derivatives of the function with respect to its input vector.

5. The topology-aware based control method for electric vertical take-off and landing aircraft according to claim 1, wherein: It also includes downgrading steps for motor or battery failures: After detecting a fault in any motor or battery pack, determine the set of remaining healthy actuators that are operating normally. ; extracting a set of indices from the control effectiveness matrix corresponding columns, constituting a reduced control effectiveness matrix ;​ Computing a reduced-order null space projection matrix : ； Computing a reduced-order one-step null-space speed increment : ； said reduced order one-step null-space rotational increment by index set map back to full-dimensional vector , components not corresponding to healthy actuators are zeroed; the subsequent accumulation and superposition process is identical to steps c2 to c4.

6. The topology-aware based control method for electric vertical take-off and landing aircraft according to claim 1, wherein: The main flight controller is any controller that generates the virtual control commands based on the aircraft state feedback and outer loop commands. The controller is: The incremental nonlinear dynamic inverse controller generates the desired output dynamic through a reference model, obtains the desired pseudo-control quantity by combining feedforward and output feedback, and obtains the virtual control incremental command by subtracting the current virtual control feedback estimate. Alternatively, it can be a non-incremental nonlinear dynamic inverse controller, in which the virtual control command is calculated directly by inverting the model. Alternatively, for a model prediction controller, the composite cost function J(u) can be used as an additional term of the prediction cost function; Alternatively, it could be a proportional-integral-derivative controller; Alternatively, it could be a controller within a robust or adaptive control framework.

7. The topology-aware based control method for electric vertical take-off and landing aircraft according to claim 1, wherein: The basic control allocation algorithm is an algorithm capable of calculating the nominal control quantity that satisfies the physical constraints of the actuator based on the virtual control command. Specifically, the basic control allocation algorithm is as follows: The reassignment pseudo-inverse algorithm, by calculating a nominal speed increment and spilling saturated inputs from the active set and reassigned residual errors to the remaining inputs; Alternatively, the weighted least squares allocation algorithm; Alternatively, a direct allocation algorithm can be used to solve for the maximum achievable torque set within the convex hull spanned by the column vectors of the control performance matrix. Alternatively, an allocation algorithm based on linear programming or quadratic programming.

8. The topology-aware based control method for electric vertical take-off and landing aircraft according to claim 1, wherein: The cumulative null space rotational speed correction amount The superposition with the nominal control amount output from the primary flight controller is implemented in accordance with the form of the primary flight controller using the following superposition methods: Incremental superposition: The main flight controller outputs the nominal rotational speed increment. The final rotor speed command is That is, the cumulative zero-space correction amount passes through historical working points Implicit accumulation; Alternatively, non-incremental superposition: the absolute value of the nominal rotational speed output by the main flight controller. The final rotor speed command is That is, the cumulative null space correction is maintained through explicit maintenance. accumulation.

9. A topology-aware based electric vertical take-off and landing aircraft control system, characterized in that, The control system executes the topology-aware electric vertical takeoff and landing (EVTOL) aircraft control method according to any one of claims 1-8, wherein the EVTOL aircraft has Distributed electric propulsion unit, Individual battery packs and There are 1 virtual control variable, and the overdrive condition is satisfied. The control system includes the following modules: An aircraft state sensing module is configured to acquire in real time the speed, angular rate, attitude angle and current rotational speed of the electric propulsion unit. a power system state estimation module, which is built in the topological matrix a physical mapping relationship, each motor winding current and each battery pack discharge current are estimated in real time according to the rotor speed collected by the aircraft state sensing module; The main flight control module generates virtual control commands based on outer loop commands and aircraft status feedback; The basic control allocation module calculates the nominal control quantity that meets the main flight control requirements based on the virtual control command; Null-space current regulation module receives outputs of the power system state estimation module and the base control allocation module, constructs a composite cost function , computes a gradient , generates a single-step null-space speed increment through a null-space projection operator and maintains an accumulated null-space speed correction ;​ a fault detection and management module that monitors motor or battery pack faults, providing a healthy actuator index set to the null-space current regulation module , triggering a reduction step The actuator instruction synthesis module will convert the accumulated zero-space rotational speed correction amount into... The nominal control quantity output by the basic control distribution module is superimposed to form the final rotor speed command, which is then sent to each motor controller.

10. An electric vertical takeoff and landing aircraft, comprising: Distributed electric propulsion unit, Individual battery packs and There are 1 virtual control variable, and the overdrive condition is satisfied. Each battery pack is connected to an electric propulsion unit that is centrally symmetrically distributed with respect to the aircraft's center of gravity. It also includes the topology-aware electric vertical takeoff and landing aircraft control system as described in claim 9.

Images