Multi-rotor unmanned aerial vehicle anti-winding parachute landing method based on phase plane safety corridor and attitude synchronous ejection

Abstract

The invention provides a multi-rotor unmanned aerial vehicle anti-winding parachute landing method based on a phase plane safety corridor and attitude synchronous ejection. The method comprises the steps that the attitude angle, the angular rate and the change rate of an inertial measurement unit are obtained and fused for calculation; under the constraint of the residual control moment and the disturbance intensity, a phase plane safety corridor which dynamically shrinks along with the height is constructed based on the Lyapunov stability criterion, the corridor margin is calculated, and a parachute landing demand sign is generated; the method comprises the following steps: establishing a paddle disc forbidding model and generating a catapulting trajectory envelope when required to be triggered, calculating a minimum interval and screening a clearance window meeting clearance and throwing threshold values by combining attitude / rolling phase prediction and a centrifugal throwing index, and performing time delay of a back-pushing execution mechanism and determining a triggering moment through a cost function; a parachute landing instruction including motor cut-off and ejection triggering is output, and a parachute landing mode locking instruction is generated according to the acceleration change and the parachute cord stress signal; the method is used for improving the parachute landing triggering time sequence and clearance matching performance, reducing the winding risk and improving the parachute landing reliability.

Description

Technical Field

[0001] This invention relates to the field of parachute safety control technology for multi-rotor UAVs, and in particular to a method for preventing entanglement during parachute landing of multi-rotor UAVs based on a phase-plane safety corridor and attitude-synchronized ejection. Background Technology

[0002] Multirotor drones are widely used for low-altitude missions such as inspection, surveying, and logistics due to their compact structure and flexible operation. However, in situations such as motor failure, blade damage, strong gusts of wind, or control malfunctions, the drone may experience significant attitude deviations and rapid tumbling, posing a risk of crashing. To reduce harm to personnel and property on the ground, emergency parachute devices are often installed in engineering applications. These devices eject or drop the parachute from the drone and pull out the parachute lines to achieve deceleration. However, multirotors have multiple high-speed rotating rotor discs, and the spatial position of the parachute lines and parachute at the moment of triggering, as well as the drone's attitude and motor status, are closely related. If the ejection trajectory interferes with the rotor discs or drone components, entanglement can easily occur, leading to parachute failure. Therefore, a triggering and execution strategy that matches the flight conditions is required.

[0003] Existing parachute systems typically consist of a controller, an inertial measurement unit (IMU), a barometric pressure / range altitude sensor, an electronic speed controller (ESC), and an ejector. Triggering criteria often employ a combination of thresholds based on attitude angle, angular velocity, descent rate, or communication failure. Once a dangerous state is detected, the controller outputs a parachute descent command, encompassing both motor cutoff and ejection triggering. Regarding execution timing, common methods include "cutoff before ejection" or "cutoff and ejection simultaneously," with a configurable fixed delay to allow the propeller disk to decelerate. To minimize interference, some solutions employ lateral / downward ejection structures with a fixed installation orientation, or use simplified geometric models to create restricted areas around the propeller disk, constraining the ejection direction and installation position. After the parachute descent is triggered, some systems monitor sudden acceleration changes or parachute line tension signals to record descent events or perform mode switching and mission termination. The overall implementation relies primarily on rule-based logic and preset parameters.

[0004] The aforementioned solutions often struggle to simultaneously address both "mandatory triggering" and "safe ejection" in uncontrolled rollover scenarios. Firstly, criteria based on a single attitude / angular rate threshold fail to reflect the impact of remaining control torque and external disturbance intensity on recoverability, and lack dynamic convergence constraints on altitude changes, leading to overly conservative or delayed triggering timing. Secondly, ejection and disengagement rely heavily on fixed delays and directions, lacking prediction of short-term attitude changes and moment-by-moment assessment of the ejection trajectory relative to the propeller disk clearance. This means the triggering time may fall within the propeller disk obstruction phase, increasing the probability of parachute lines / packets interfering with and becoming entangled with the propeller disk. Thirdly, insufficient closed-loop locking of actuator delays and parachute descent execution status can easily lead to triggering timing deviations, repeated triggering, and mode rollback risks.

[0005] Therefore, those skilled in the art urgently need a method for preventing multi-rotor drones from becoming entangled during parachute landing. Summary of the Invention

[0006] One objective of this invention is to propose an anti-entanglement parachute landing method for multi-rotor UAVs based on a phase-plane safety corridor and attitude-synchronized ejection. The core technical problem to be solved by this application is: in emergency parachute landing scenarios where multi-rotor UAVs experience attitude instability and may tumble, and the motor status and altitude are constantly changing, how to determine and execute the parachute landing trigger moment (including motor cut-off and ejection sequence) that is phase-synchronized with the aircraft's attitude, so that the ejection trajectory of the parachute pack / parachute lines avoids the propeller disk restricted area within the trigger window and meets the ejection conditions, thereby reducing entanglement and ensuring that the parachute landing can be effectively established.

[0007] The anti-entanglement parachute landing method for multi-rotor UAVs based on phase-plane safety corridor and attitude-synchronized ejection according to embodiments of the present invention includes:

[0008] S1. Obtain the attitude angle, angular rate, and rate of change of angular rate of the inertial measurement unit, obtain the height and motor status, and perform fusion calculation on the attitude angle and angular rate to obtain the attitude calculation quantity and angular velocity calculation quantity.

[0009] S2. Construct phase plane state variables containing attitude angle and angular rate based on attitude and angular rate calculations. Determine the remaining control torque based on motor state and the disturbance intensity based on angular rate change rate. Under the constraints of remaining control torque and disturbance intensity, determine the dynamic shrinkage safety corridor boundary that varies with altitude based on Lyapunov stability criterion. Calculate the corridor margin of phase plane state variables relative to the dynamic shrinkage safety corridor boundary and determine the parachute demand indicator based on the comparison between corridor margin and corridor threshold.

[0010] S3. Determine the propeller rotation plane by calling structural parameters according to the parachute drop requirement flag, generate the propeller disk no-entry model, and generate the catapult trajectory envelope according to the catapult installation direction and the initial catapult velocity.

[0011] S4. Based on the attitude calculation and angular velocity calculation, perform rotation vector prediction to obtain the roll phase prediction result. Calculate the centrifugal ejection index based on the angular velocity calculation and structural parameters. Map the ejection trajectory envelope to the propeller disk no-entry model under the constraint of the roll phase prediction result. Calculate the minimum interval and filter the clearance windows that meet the clearance threshold and the centrifugal ejection index meets the ejection threshold. Backtrack the clearance window based on the actuator delay parameter and determine the trigger time in combination with the corridor margin.

[0012] S5. Generate a parachute trigger command based on the parachute demand flag, trigger time, and corridor margin. The parachute trigger command includes an ejection trigger signal and a motor cutoff signal.

[0013] S6. Acquire the changes in the body acceleration and the force signals on the parachute lines to form the parachute descent execution state quantity, and generate the parachute descent mode lock command based on the parachute descent execution state quantity.

[0014] Optionally, S1 is as follows:

[0015] The attitude angle and angular rate output by the inertial measurement unit are read at the same sampling time, and the angular rate is zero-biased to form a compensated angular rate;

[0016] The rate of change of angular rate is obtained by performing differential operation on the compensated angular rate, and the attitude angle, compensated angular rate, and rate of change of angular rate are unified into the machine system coordinate expression.

[0017] Obtain the height and motor status, where the motor status includes the motor speed and the motor enable status;

[0018] Using attitude angle and compensated angular rate as fusion inputs, state filtering is used to fuse and calculate the attitude and angular velocity solutions.

[0019] Optionally, S2 is as follows:

[0020] The attitude calculation and angular velocity calculation are paired along the same attitude axis to construct a phase plane state variable that includes attitude angular components and angular velocity components. The phase plane state variable is then used as the state input for subsequent stability determination.

[0021] The motor enable state and motor speed are extracted based on the motor state. The remaining control torque is determined by mapping the motor state to the torque availability. The remaining control torque is then used as the control constraint input for the recoverable domain calculation.

[0022] The amplitude of the rate of change of angular rate is calculated and the perturbation intensity is formed by using the amplitude limiting rule. The perturbation intensity is then used as the external perturbation constraint input for the calculation of the recoverable domain.

[0023] A contraction coefficient that monotonically varies with height is generated based on the height, and the contraction coefficient is used to adjust the contraction of the safety corridor boundary to form a dynamically contracting safety corridor boundary that varies with height.

[0024] Under the constraints of residual control torque and disturbance intensity, the Lyapunov stability criterion is used to determine the stable domain of the phase plane state variables, and the boundary parameters of the dynamically shrinking safety corridor boundary are output to limit the allowable range of the phase plane state variables.

[0025] The distance between the phase plane state quantity and the boundary parameters of the dynamic shrinking safety corridor boundary is calculated to obtain the corridor margin characterizing the phase plane state quantity relative to the dynamic shrinking safety corridor boundary, and the corridor margin is kept as a continuous quantity for subsequent triggering time determination.

[0026] The corridor margin is compared with the corridor threshold. When the corridor margin is less than or equal to the corridor threshold, a parachute demand flag is generated. When the corridor margin is greater than the corridor threshold, the parachute demand flag is kept in a non-triggered state. The corridor margin and the parachute demand flag are then output to the subsequent clearance window filtering and trigger time determination process.

[0027] Optionally, in the process of determining the stable domain of the phase plane state variables and outputting the boundary parameters of the dynamically shrinking safety corridor boundary using the Lyapunov stability criterion under the residual control torque constraint and disturbance intensity constraint, the boundary budget corresponding to the dynamically shrinking safety corridor boundary is determined by the boundary budget formula:

[0028] ;

[0029] in, Budget for the Lyapunov border, As the baseline boundary budget constant, This is the torque availability coefficient. For the disturbance intensity, This is a reference value for disturbance intensity. This is the shrinkage coefficient.

[0030] Optionally, S3 specifically refers to:

[0031] Based on the parachute drop requirement flag, structural parameters are retrieved, and the three-dimensional coordinates of the center of the propeller rotation plane, the three-dimensional normal of the propeller rotation plane, and the one-dimensional radius of the propeller disk are extracted to determine the propeller rotation plane.

[0032] A conditional no-entry distance network is used to generate a propeller disk no-entry model. The three-dimensional coordinates of the propeller rotation plane center, the three-dimensional normal of the propeller rotation plane, the one-dimensional radius of the propeller disk, the one-dimensional motor state, and the three-dimensional coordinates of the spatial point to be determined in the machine system are used as eleven-dimensional inputs. In the feature generation layer, the spatial point to be determined is transformed into the propeller disk normal distance, the propeller disk radial distance, and the ratio of the radial distance to the propeller disk radius. The motor state is mapped to the no-entry expansion to output a six-dimensional feature vector. The one-dimensional no-entry distance is output through a three-layer fully connected network and a linear rectifier unit.

[0033] The ejection direction vector of the machine system is generated based on the installation direction of the ejector, and the ejection trajectory point set that evolves with time is generated in combination with the initial ejection velocity to form the ejection trajectory envelope;

[0034] Output the propeller disk exclusion model and the ejection trajectory envelope.

[0035] Optionally, S4 specifically refers to:

[0036] The prediction time domain is determined by using attitude and angular velocity calculations as inputs, and the rotation vector sequence is obtained by integrating the angular velocity calculations.

[0037] The rotation vector sequence is converted into a tumble phase prediction result, which includes the change in body attitude at each moment in the prediction time domain.

[0038] Based on the calculated angular velocity and structural parameters, a centrifugal ejection index sequence is generated. The structural parameters include at least the ejection radius parameter and the parachute connection point relative to the body coordinate parameter.

[0039] The ejection trajectory envelope is subjected to attitude transformation under the constraint of the roll phase prediction result to obtain the three-dimensional coordinates of the spatial point to be determined in the machine system at each time in the prediction time domain.

[0040] The coordinates of the spatial point to be determined in the machine system are input into the propeller disk no-entry model in three dimensions to obtain the corresponding no-entry distance, and the no-entry distance at each prediction time is minimized to obtain the minimum interval sequence.

[0041] The minimum interval sequence is compared with the net air threshold to obtain the net air window that meets the net air threshold. The centrifugal separation index sequence is compared with the separation threshold to further filter the net air window and obtain the net air window that meets both the net air threshold and the separation threshold.

[0042] Based on the time delay parameters of the actuator, the clearance window is back-calculated to form a set of candidate triggering times, and the set of candidate triggering times is correlated with the corridor margin to determine the triggering time.

[0043] Optionally, in the process of generating a candidate triggering time set by back-calculating the clearance window based on the actuator delay parameters, and then correlating and filtering the candidate triggering time set with the corridor margin to determine the triggering time, a cost function is used to correlate and filter the candidate triggering times in the candidate triggering time set, and the candidate triggering time with the lowest cost is determined as the triggering time. The cost function is:

[0044] ;

[0045] in, For the trigger time, The operation is performed to select the candidate time step that minimizes the value within the parentheses. Candidate trigger time, For the set of candidate trigger times, As the weight of the corridor margin cost, For the corridor threshold, For corridor margin, Net short cost weighting, The airspace threshold, To be according to The minimum interval value read after mapping to the predicted time series. In order to get rid of the cost weight, To escape the threshold, To be according to The centrifugal shedding index values ​​are read after mapping to the predicted time series. To push back the delay, As time priority weight, At the current sampling time, To predict the length of the time domain, This is an operation that takes the larger of zero and the value inside the parentheses.

[0046] Optional, S5 specifically includes:

[0047] When the parachute demand flag is in the triggered state, the trigger time and corridor margin are read as inputs for instruction generation.

[0048] The corridor margin is compared with the linkage threshold. When the corridor margin is not greater than the linkage threshold, the ejection trigger signal and the motor cut-off signal are determined to be output at the same trigger time. When the corridor margin is greater than the linkage threshold, the motor cut-off signal is determined to be output before the ejection trigger signal and a preset interval time is determined.

[0049] A motor cut-off signal is generated based on the trigger time. The motor cut-off signal includes enable and disable commands for all motors.

[0050] A launch trigger signal is generated based on the trigger time and a preset interval, and the launch trigger signal and the motor cut-off signal are encapsulated into a parachute trigger command.

[0051] Optionally, step S6 specifically includes:

[0052] The body acceleration output by the inertial measurement unit is obtained, and the body acceleration change is obtained by differential calculation.

[0053] Acquire the paracord force signal output by the paracord force sensor, and determine the validity of the paracord force signal to obtain a valid paracord force signal;

[0054] By correlating the changes in body acceleration with the effective parachute rope force signal in a time series, a parachute descent execution state quantity containing acceleration impact index and force duration index is constructed.

[0055] The parachute execution status quantity is compared with the locking threshold. When the parachute execution status quantity meets the locking threshold, a parachute mode locking command is generated. The parachute mode locking command is used to maintain the parachute mode and prevent the motor cut-off signal cancellation and the regeneration of the parachute trigger command.

[0056] The beneficial effects of this invention are:

[0057] (1) This proposal proposes an improved method for determining parachute demand based on a phase-plane safety corridor. By pairing attitude angles and angular velocities along axes to form phase-plane state variables, and introducing the Lyapunov stability criterion under the constraints of residual control torque and disturbance intensity, the stable domain boundary is calculated, thus replacing the triggering logic that relies solely on attitude thresholds or angular velocity thresholds. This scheme further maps the motor state to a torque availability coefficient, converts the rate of change of angular velocity into the external disturbance intensity by combining the moment of inertia, and uses an altitude contraction coefficient to dynamically contract the boundary, making the safety corridor more conservative as the altitude decreases, avoiding trigger lag or inconsistency in triggering basis caused by using the high-altitude criterion at low altitudes. At the same time, the corridor margin is used as a continuous output quantity and runs through the selection of subsequent triggering times, so that the parachute triggering changes from a discrete judgment of "whether it exceeds the limit" to "margin-driven from the distance to the instability boundary", which is closer to the recoverability boundary in the scenario of power loss and external disturbance change, and reduces the triggering opportunities that are not compatible with the motor controllability.

[0058] (2) This proposal puts forward a novel attitude synchronization ejection method based on roll phase prediction, propeller disk no-entry distance regression, and ejection trajectory envelope mapping. By integrating the angular velocity solution, the rotation vector sequence is obtained and the attitude change in the prediction time domain is generated. The ejection trajectory envelope is transformed to the machine system under each predicted attitude. The minimum interval between the ejection trajectory and the propeller disk no-entry model is calculated at each time step, thereby forming a clearance window that meets the clearance threshold. Unlike the existing "fixed delay + fixed direction" approach, this scheme explicitly incorporates the "future short-term attitude phase" into the ejection feasibility assessment. By using a conditional no-entry distance network, the propeller disk geometric parameters and motor state are taken as input to obtain the no-entry distance that changes with the power state, so that the clearance judgment can reflect the change of the propeller disk danger zone with the speed / enabling state. In addition, this scheme introduces a centrifugal ejection index calculated from angular velocity and structural parameters to perform secondary screening of the clearance window, avoiding trigger points where only the geometric clearance is satisfied but the parachute ropes cannot be effectively ejected. Then, the window is backtracked in time by combining the actuator delay, and the optimal triggering time is selected from the candidate triggering time using a cost function that includes corridor margin, clearance margin and ejection margin. This ensures that the final triggering time simultaneously satisfies the recoverability constraint, geometric clearance constraint and execution timing constraint, thereby reducing the possibility of entanglement caused by the instantaneous intersection of the ejection and the propeller disk.

[0059] (3) This proposal proposes a parachute drop command generation and execution locking method based on corridor margin linkage. The relative timing of "motor cut-off - ejection trigger" is used as an adjustable control quantity rather than a fixed configuration: when the corridor margin is lower than the linkage threshold, simultaneous triggering is used to shorten the decision link. When the margin is large, motor cut-off is allowed first, and the interval is set by combining the stop time and the delay of the actuator, so as to take into account the propeller deceleration and ejection timing matching when the safety margin allows. After triggering, the scheme constructs the parachute drop execution state quantity through the change of body acceleration and the parachute rope force signal, and generates a parachute drop mode locking command when the impact and continuous force locking thresholds are met, prohibiting motor cut-off cancellation and repeated triggering, reducing the risk of secondary actions caused by sensor transients or state machine backtracking. The above closed-loop link expands the parachute drop from "triggering one action" to a consistent process of "judgment - window selection - pushback - linkage execution - state confirmation and locking", which helps to achieve ejection execution synchronized with attitude phase and suppress entanglement-related failure modes in uncontrolled rollover and power degradation scenarios. Attached Figure Description

[0060] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0061] Figure 1 This is a flowchart of a method for preventing entanglement of a multi-rotor UAV using a phase-plane safety corridor and attitude-synchronized ejection, as proposed in this invention.

[0062] Figure 2 This is a flowchart of attitude and dynamic state acquisition and fusion calculation for a multi-rotor UAV anti-entanglement parachute landing method based on phase plane safety corridor and attitude synchronous ejection proposed in this invention.

[0063] Figure 3 This is a flowchart of the dynamic shrinking safety corridor boundary and parachute requirement determination process for a multi-rotor UAV anti-entanglement parachute landing method based on phase plane safety corridor and attitude synchronous ejection proposed in this invention.

[0064] Figure 4 This is a flowchart of the propeller disk exclusion model and ejection trajectory envelope generation process for a multi-rotor UAV anti-entanglement parachute landing method based on phase plane safety corridor and attitude synchronous ejection proposed in this invention. Detailed Implementation

[0065] In Example 1, reference Figure 1 as well as Figure 4 A method for preventing entanglement during parachute landing of a multi-rotor UAV based on a phase-plane safety corridor and attitude-synchronized ejection includes:

[0066] S1. Obtain the attitude angle, angular rate, and rate of change of angular rate of the inertial measurement unit, obtain the height and motor status, and perform fusion calculation on the attitude angle and angular rate to obtain the attitude calculation quantity and angular velocity calculation quantity.

[0067] S2. Construct phase plane state variables containing attitude angle and angular rate based on attitude and angular rate calculations. Determine the remaining control torque based on motor state and the disturbance intensity based on angular rate change rate. Under the constraints of remaining control torque and disturbance intensity, determine the dynamic shrinkage safety corridor boundary that varies with altitude based on Lyapunov stability criterion. Calculate the corridor margin of phase plane state variables relative to the dynamic shrinkage safety corridor boundary and determine the parachute demand indicator based on the comparison between corridor margin and corridor threshold.

[0068] S3. Determine the propeller rotation plane by calling structural parameters according to the parachute drop requirement flag, generate the propeller disk no-entry model, and generate the catapult trajectory envelope according to the catapult installation direction and the initial catapult velocity.

[0069] S4. Based on the attitude calculation and angular velocity calculation, perform rotation vector prediction to obtain the roll phase prediction result. Calculate the centrifugal ejection index based on the angular velocity calculation and structural parameters. Map the ejection trajectory envelope to the propeller disk no-entry model under the constraint of the roll phase prediction result. Calculate the minimum interval and filter the clearance windows that meet the clearance threshold and the centrifugal ejection index meets the ejection threshold. Backtrack the clearance window based on the actuator delay parameter and determine the trigger time in combination with the corridor margin.

[0070] S5. Generate a parachute trigger command based on the parachute demand flag, trigger time, and corridor margin. The parachute trigger command includes an ejection trigger signal and a motor cutoff signal.

[0071] S6. Acquire the changes in the body acceleration and the force signals on the parachute lines to form the parachute descent execution state quantity, and generate the parachute descent mode lock command based on the parachute descent execution state quantity.

[0072] In this embodiment, step S1 specifically includes:

[0073] The controller uses a fixed sampling period Run, and at each sampling time The inertial measurement unit (IMU) performs attitude and dynamic state acquisition and outputs attitude angles. With angular velocity ,in It is an attitude angular quantity composed of three components: roll angle, pitch angle, and yaw angle. The vector is a three-axis angular rate vector. To ensure that the subsequent construction of phase plane state variables has a consistent input reference with the roll phase prediction, the controller reads... Then call the gyroscope zero bias parameter Zero bias compensation is performed to obtain the compensated angular rate. The Obtained from factory calibration, and allows for online correction via state-filtered fusion calculation during operation;

[0074] In obtaining Then, the controller performs differential calculations on the compensated angular velocities at adjacent sampling times, and divides the differential results according to the sampling period. The rate of change of angular velocity was obtained by conversion. To suppress the impact of spikes on disturbance intensity determination, the controller... Amplitude limiting is performed, with the limiting threshold being a preset threshold for the rate of change of angular velocity. Subsequently, the controller invokes the inertial measurement unit to install the calibration matrix. Compensation angular rate With the rate of change of angular velocity The attitude angle is expressed by transforming the sensor coordinate system into the machine coordinate system. The attitude angle definition is consistent with that of the machine system, and is used for interaction with the machine system. , Under the same coordinate semantics, proceed to the subsequent fusion and determination link;

[0075] At the same sampling time The controller reads the height output from the height sensor. The height is used in step S2 to generate a dynamically shrinking safety corridor boundary that varies with height. The controller simultaneously reads the motor speed and motor enable status output by the electronic speed controller, where the motor speed is denoted as... The motor enable state is recorded as The controller first according to Determine the number of motors in the enabled state, and assign corresponding motors to the enabled motors. The average motor speed is calculated from the components. The controller then normalizes the average motor speed according to the rated speed parameter, and combines this with the normalization of the enabled motors according to the total number of motors, to obtain the one-dimensional motor state. ,in The motor status represents the current availability of rotor power. Used in step S2 to determine the remaining control torque, and as an input to the conditional no-entry distance network in step S3 to participate in the generation of no-entry expansion, so that the no-entry distance participates in the clearance window filtering in step S4 according to the motor state.

[0076] After completing the above data acquisition and processing, the controller uses an attitude angle and compensated angular rate As the fusion input, state filtering is used to fuse the output attitude solution. Angular velocity calculation The state filtering fusion solution is performed in the order of prediction and correction: the prediction stage is based on... The attitude state is propagated, and the correction phase is carried out in accordance with the law. The predicted pose is updated, and the zero bias parameter is adjusted during the update process. Perform synchronous correction, angular velocity calculation quantity Depend on The output is obtained by smoothing it through a filter to ensure that its time phase is consistent with... Consistent, ultimately, the controller outputs the attitude calculation at each sampling time. Angular velocity calculation quantity angular rate of change ,high Motor status This data is used as the input data source for steps S2 and S4 to support the continuous decision-making chain of dynamically shrinking safety corridor boundary determination, corridor margin calculation, rollover phase prediction, clearance window screening, and trigger time determination.

[0077] In this embodiment, step S2 specifically includes:

[0078] The controller at each sampling time Received attitude calculation Angular velocity calculation ,in The attitude angle calculation results under the machine system, The controller will calculate the angular rate in the machine system. and Decompose and pair along the same attitude axis to obtain the attitude angular components of each attitude axis. With angular velocity components subscript Indicates the attitude axis identifier, the controller uses and Constructing phase plane state variables The It is a binary state variable containing attitude angular components and angular velocity components, and the attitude axes are... As the state input for subsequent stability determination, when it is necessary to merge multiple attitude axes into a single determination input, the controller uses the minimum value of the corridor margin of each attitude axis as the global corridor margin to enter the subsequent process.

[0079] The controller acquires the motor status at the same sampling time. and in motor status Internal extraction of motor enable state With motor speed ,in This is the enable flag vector for each motor. For the speed vector of each motor, the controller according to Count the number of enabling motors, and... The average motor speed is obtained by averaging the components corresponding to the enabled motors. The controller normalizes the average motor speed according to the rated speed parameter and normalizes the number of enabled motors according to the total number of motors. The two are then weighted and combined according to preset weights to obtain the motor state. The controller calls the preset torque availability mapping. Will Mapped to torque availability coefficient The Implemented using piecewise linear mapping or lookup table mapping, with the input being... The output is a one-dimensional scalar. The controller will set the rated control torque parameter. according to Scaling is performed to obtain the remaining control torque. and will As the control constraint input for recoverable domain calculation;

[0080] The controller acquires the rate of change of angular velocity at the same sampling time. ,right The angular acceleration amplitude is obtained by calculating the Euclidean norm of the three-axis components. and according to the preset angular rate of change threshold Limiting the amplitude yields the limited amplitude value. The controller will limit the amplitude. With respect to the moment of inertia parameter in the structural parameters Combined with the generated disturbance intensity ,in The controller will provide the three-axis rotational inertia parameters of the machine system. The angular acceleration components are obtained by decomposing the components along three axes. These components are then multiplied by the moments of inertia of the corresponding axes to obtain the external disturbance torque components. Finally, the Euclidean norm of the external disturbance torque components is calculated to obtain the one-dimensional disturbance strength. This results in the intensity of the disturbance. It can be used as an external perturbation constraint input for recoverable domain computation;

[0081] Controller gets height Then, call the height-to-shrinkage coefficient mapping. Generate shrinkage coefficient The Adopt the minimum working height With the highest working height Linear interpolation for the boundary is implemented: the controller first... Clip to range Then adjust the height after cutting according to and Normalization yields ,make The height changes monotonically and the value falls within the interval. In obtaining , and Subsequently, the controller converges the three types of constraints into a single Lyapunov boundary budget, and determines the dynamically shrinking safety corridor boundary that varies with altitude. The controller generates function values ​​for each attitude axis using Lyapunov functions. ,in Depend on and The weighted sum of squares is calculated, and the weighting parameters are derived from the Lyapunov weights. Given, the The controller must include at least two values: attitude angle weight and angular rate weight. Then, it calculates the Lyapunov boundary budget corresponding to the dynamically shrinking safety corridor boundary. and will satisfy The contour lines serve as the basis for generating the dynamically shrinking safety corridor boundaries for each attitude axis. Determined using the boundary budget formula:

[0082] ;

[0083] in, Budget for the Lyapunov border, As the baseline boundary budget constant, This is the torque availability coefficient. For the disturbance intensity, This is a reference value for disturbance intensity. The shrinkage coefficient;

[0084] The controller will With Lyapunov weight Together, they are used to generate boundary parameters, outputting the boundary parameters of the dynamically shrinking safety corridor boundary. The The boundaries of each attitude axis are given in the form of parametric curves, including the allowable range parameters of attitude angles and angular rates, and are used to limit the phase plane state variables. Within the allowed range, the controller then assigns the phase plane state variables. Boundary parameters of the dynamically shrinking safety corridor boundary Distance calculation is performed by discretizing the boundary parameterized curve to form a boundary point set, and then calculating... The corridor margin is obtained by finding the minimum Euclidean distance to the boundary point set. and stipulate Positive values ​​are applied inside the boundary, and negative values ​​are applied outside the boundary. The controller will assign values ​​to each attitude axis. Find the minimum value to obtain the corridor margin. and maintain corridor slack. This is a continuous quantity used to determine the subsequent triggering time;

[0085] The controller will adjust the corridor margin. Corridor threshold When comparing, Less than or equal to At that time, the controller generates a parachute drop request flag. For the triggered state, when Greater than At that time, the controller maintains the parachute drop demand flag. In the non-triggered state, the controller will adjust the corridor margin. With parachute demand marker The output is used in the subsequent clearance window filtering and trigger time determination process, so that the subsequent process can use the propeller disk no-entry model and calculate the clearance window. Perform consistent constraint filtering on candidate trigger times.

[0086] In this embodiment, step S3 specifically includes:

[0087] The controller at the sampling time Receive parachute request sign ,when When the trigger state is reached, the structural parameters are called and step S3 is initiated. The controller extracts the three-dimensional coordinates of the propeller rotation plane center from the structural parameters. , Propeller rotation plane normal three-dimensional One dimension with the propeller disk radius ,in Let be the geometric center position vector of the propeller's plane of rotation in the machine system. Let be the unit normal vector of the propeller's plane of rotation in the engine system. The controller uses the corresponding disk radius value for the propeller. and Determine the propeller rotation plane and use it as the geometric reference for subsequent no-entry distance regression;

[0088] When generating the propeller disk no-entry model, the controller employs a conditional no-entry distance network, and the controller constructs an eleven-dimensional input vector. , The three-dimensional coordinates of the center of the propeller rotation plane , Propeller rotation plane normal three-dimensional One-dimensional propeller disk radius One-dimensional motor status The coordinates of the spatial point to be determined in the three-dimensional coordinate system It is assembled in a fixed order, among which The coordinates of the machine space points obtained after discretizing the ejection trajectory envelope into a point set in step S4 are used in the feature generation layer. , , , and As input, first calculate the relative displacement vector. The for and The difference vector, the feature generation layer will exist The projection in the direction is used as the normal distance of the propeller disk. ,in For a one-dimensional directed distance, the sign is determined by... Once the target is determined, the feature generation layer will... The remaining components after removing the normal projection are taken as in-plane components, and the radial distance of the propeller disk is obtained by calculating the length of the in-plane components. The feature generation layer then uses and Calculate the ratio of radial distance to propeller disk radius. and with and Calculate the difference between the radial distance and the propeller disk radius. ,in Used to characterize the radially normalized position Used to characterize the radial margin relative to the propeller disk boundary;

[0089] The feature generation layer will generate the motor state Input motor state to forbidden expansion mapping Obtain the forbidden expansion amount ,in Implemented using lookup table mapping or piecewise linear mapping, with the input being... The output is a one-dimensional scalar. In one implementation method, Motor state threshold Divide the work area, when Not less than Output working expansion amount when Less than The non-working expansion amount is output in real time, so that the no-entry boundary can participate in the subsequent clearance window selection according to the motor state. The feature generation layer outputs a six-dimensional feature vector. The In order from , , , , and This component is used to compress the key geometric relationships between spatial points and the propeller disk, along with motor state condition information, into a regressible input.

[0090] The middle layer uses a three-layer fully connected network. Regression calculations are performed using 32 neurons in the first hidden layer, 32 neurons in the second hidden layer, and 16 neurons in the third hidden layer. Activation is achieved using linear rectified units, and the output layer outputs a one-dimensional forbidden distance. , Defined as the signed distance from the spatial point to be determined to the propeller disk no-entry boundary, the sign rule is positive outside the no-entry boundary and negative inside the no-entry boundary. The controller constructs the conditional no-entry distance network together with its input and output interfaces into a propeller disk no-entry model, and provides the propeller disk no-entry model to step S4 for minimum interval calculation and clearance window filtering.

[0091] When generating the ejection trajectory envelope, the controller calls the ejector installation direction parameter in the structural parameters to generate the ejection direction vector of the machine system. ,in The controller obtains the initial launch velocity by using the unit direction vector within the machine system. And call the launch start coordinates in the structure parameters. As the initial position of the trajectory, the controller sets the trajectory time domain length. With discrete time step ,according to Generate a discrete time sequence, and the controller calculates the trajectory center point for each discrete time. The By along Direction The advancement is achieved, and the controller will... The ejection trajectory point set is composed of the envelope radius in the structural parameters. Expand outwards for each trajectory center point: the controller in conjunction with... Generate surroundings within orthogonal cross-sections The envelope sampling point set, so that the envelope sampling points to The distance is The union of the envelope sampling points at all discrete moments is defined as the ejection trajectory envelope. The controller outputs the propeller disk no-entry model and the ejection trajectory envelope to step S4, which is used to complete the generation of the spatial points to be determined, the calculation of the no-entry distance, and the screening of the clearance window under the constraint of the roll phase prediction result.

[0092] In this embodiment, step S4 specifically includes:

[0093] The controller at the sampling time Received attitude calculation Angular velocity calculation quantity Corridor margin ejection trajectory envelope Propeller disk no-entry model With actuator delay parameters ,in Including motor cut-off delay With ejection trigger delay The controller takes first and The larger one is used as the pushback delay The controller then and Jointly determine the prediction time domain length :

[0094] The controller calculates the attitude deviation amplitude. , for The controller calculates the angular velocity amplitude based on the maximum absolute value of the three attitude angular components. , for The Euclidean norm of the three-axis angular rate components, and the controller's preset attitude threshold. Angular velocity threshold and And pre-set candidate values ​​for prediction in the time domain. , and ,when Not less than or Not less than Time to take ,when In and Take between ,when Less than and Less than Time to take The controller then right Apply a lower bound constraint so that Not less than The controller is set to predict the distance from the walk. , and according to and Generate from Start, length is Predicted time series of steps ,in Depend on and Obtained by dividing by the integer part;

[0095] controller with The angular velocity solution is integrated and predicted using the initial values ​​as the prediction parameters to obtain the rotation vector sequence. ,in For the three-dimensional rotation vector representing the attitude increment in the machine system, integral prediction is achieved using a step-by-step accumulation method:

[0096] The controller calculates the angular velocity in each prediction step according to... The value is converted into the rotation increment for this step and added to the rotation vector of the previous prediction step to obtain a rotation vector sequence corresponding to the prediction time. The controller then converts the rotation vector sequence into a roll phase prediction result. And the changes in body attitude at each time point in the predicted time domain are denoted as The controller according to Determine the main roll axis , for The unit vector obtained by direction normalization is used by the controller from Extracting from the surrounding The initial attitude angle components corresponding to the attitude axis are used as the initial values ​​of the roll phase. The controller will exist The projection of the direction is used as the roll phase increment and accumulated to obtain and will As The attitude change input is used to provide the attitude change at each prediction time during subsequent attitude transformations.

[0097] The controller generates a sequence of centrifugal ejection indexes based on the calculated angular velocity and structural parameters. Structural parameters include the throw-off radius parameter. Relative coordinate parameters of the parachute rope connection point to the aircraft body ,in This is the equivalent radius of the parachute lines as they are thrown off. Given the position vector of the parachute rope connection point within the system, the controller calculates the angular velocity amplitude at each predicted moment. and based on With the main roll axis Calculate the effective throw radius , for arrive vertical distance and according to Apply upper bound constraints, and the controller will The square of and Multiply to obtain the centrifugal ejection index The centrifugal ejection index sequence is used to determine the ejection threshold.

[0098] The controller will envelop the launch trajectory. In the rolling phase prediction results Under constraints, attitude transformation is performed to obtain the three-dimensional coordinates of the spatial point to be determined in the machine system at each time point in the prediction time domain. ,in As the discrete sampling point index for the ejection trajectory envelope, the controller takes the reference point set of the ejection trajectory envelope as input and, based on... and The rotation operator for the predicted moment is constructed. The controller applies the rotation operator to the reference point set point by point and superimposes the installation posture offset of the catapult in the machine system to obtain the three-dimensional coordinates of the spatial point to be determined in the machine system at the corresponding predicted moment, so that the catapult trajectory envelope and the roll phase prediction result are aligned under the same time reference.

[0099] The controller will determine the coordinates of the spatial point to be decided at each prediction time in the three-dimensional coordinates of the machine system. Input propeller disk no-entry model Obtain the corresponding no-entry distance The minimum interval is obtained by performing a minimum operation on the forbidden distance at the same prediction time. When a drone has multiple propellers, the controller calls the corresponding propeller disk no-entry model for each propeller and calculates the minimum spacing for that propeller. Then, it performs a minimization operation on the minimum spacing of all propellers to obtain the minimum spacing of the entire drone. ,make As a single criterion for screening the clearance window;

[0100] The controller will use the minimum interval sequence With net air threshold Time-by-time comparison, extract the satisfied Not less than The controller then uses a continuous time period as the initial clearance window, and then uses the centrifugal separation index sequence. With the throw-off threshold The comparison is performed time-by-time, and the initial clearance window is further filtered to retain only those elements within the window that simultaneously meet the requirements. Not less than During a continuous time period, the net airspace threshold is simultaneously satisfied. With the throw-off threshold The clearance window, the controller based on the pushback delay Perform time rollback on the clearance window and set the triggerable time after rollback according to... Discretization forms a set of candidate trigger times ;

[0101] Controller set of candidate trigger times Internal corridor margin Determine the trigger time The following cost function is used to correlate and filter candidate trigger times:

[0102] ;

[0103] in, For the trigger time, The operation is performed to select the candidate time step that minimizes the value within the parentheses. Candidate trigger time, For the set of candidate trigger times, As the weight of the corridor margin cost, For the corridor threshold, For corridor margin, Net short cost weighting, The airspace threshold, To be according to The minimum interval value read after mapping to the predicted time series. In order to get rid of the cost weight, To escape the threshold, To be according to The centrifugal shedding index values ​​are read after mapping to the predicted time series. To push back the delay, As time priority weight, At the current sampling time, To predict the length of the time domain, This operation takes the larger of zero and the value within the parentheses.

[0104] The controller is calculating and At that time, according to Mapped to discrete indices in the predicted time series And read the minimum interval corresponding to that index. Centrifugal Ejection Index As input, the determination of the triggering time is simultaneously constrained by corridor margin, clearance window, shake-off condition, and actuator delay.

[0105] In this embodiment, step S5 specifically includes:

[0106] The controller at the sampling time Receive parachute request sign ,when When the state is triggered, the parachute trigger command generation process begins, and the controller reads the trigger time output in step S4. Corridor margin output from step S2 As input for instruction generation, the As the target time reference for the ejection trigger to be executed, the To provide continuous corridor margin, the controller simultaneously reads the preset linkage threshold. With preset interval time ,in The threshold used to characterize the linkage determination between ejection and cutting off Used to characterize the time interval between the motor cut-off signal and the ejection trigger signal;

[0107] The controller will adjust the corridor margin. With linkage threshold When comparing, Not greater than At that time, the controller determines that the ejection trigger signal and the motor cut-off signal are output at the same trigger moment, and sets the preset interval time. Setting it to zero, the controller defines the same trigger time as... and ,in The moment when the motor cuts off the signal output. When the ejection trigger signal is output, Greater than At that time, the controller determines that the motor cut-off signal is output before the ejection trigger signal, and uses a preset interval time. The controller defines the moment when the motor cut-off signal is output as follows: (This is the initial statement of the controller's output timing for determining the output timing of the two signals.) The ejection trigger signal output time is defined as In one implementation method, The propeller stop time parameter and actuator delay parameter in the structural parameters are jointly determined: The controller first reads the propeller stop time parameter, which is the constant time required for the motor to shut down until the speed drops to the safe speed threshold. The controller then reads the motor cut-off delay and ejection trigger delay in the actuator delay parameters. The controller adds the above time constants together according to a preset safety margin to obtain the final value. and will Limiting the interval to between the preset minimum and maximum intervals, so that This is a time value that can be directly used for scheduling;

[0108] The controller outputs the motor cut-off signal at the appropriate time. Generate motor cut-off signal The This includes enable / disable commands for all motors; the controller reads the number of motors from the motor configuration parameters. And generate motor enable / disable command vectors. ,in For length is The vector, with each component being a closed value, will be controlled by the controller. With timestamp Packaged as and will Write to the timed output queue of the motor control interface so that the motor control interface can... Always output enable / disable commands to all motors;

[0109] The controller outputs based on the ejection trigger signal timing. Generate ejection trigger signal The For the catapult trigger control input, the controller reads the catapult trigger channel identifier and trigger pulse width parameter, and generates a signal containing the trigger valid level, pulse width, and timestamp. The controller encapsulates the trigger frame into a trigger frame. And write it into the timed output queue of the catapult control interface, so that the catapult control interface... Output trigger pulses at all times;

[0110] The controller will trigger the ejection signal. Disconnect signal from motor Encapsulated as a parachute trigger command The At least including the instruction serial number, , , With the catapult trigger frame, the controller will Output to the execution layer so that the execution layer can... and Complete the coordinated execution of motor cut-off and ejection trigger, and maintain... and The corresponding timing rules do not change within a single parachute drop procedure.

[0111] In this embodiment, step S6 specifically includes:

[0112] The controller is triggered by the parachute drop command. After the order is issued, the parachute drop execution monitoring process begins, in which... To encapsulate the parachute trigger command, which combines the ejection trigger signal and the motor cut-off signal, the controller samples the data at the specified time. Obtain the body acceleration output by the inertial measurement unit ,in The controller obtains the sampling period for the three-axis acceleration numerical vector of the machine system. ,in The controller sets the time interval between adjacent sampling times as the numerical value. Performing differential calculations yields changes in the body's acceleration. The difference operation is performed by subtracting the currently sampled body acceleration vector from the previously sampled body acceleration vector, and then... Perform normalization so that Characterizing the abrupt change in acceleration per unit time, the controller... The amplitude of acceleration change is obtained by performing amplitude calculation. The amplitude calculation uses the square root of the summation of the squares of the three-axis components to obtain a one-dimensional value. The controller is in a length of... Extract within the sliding window The peak value of the window is used as an indicator of acceleration impact. ,in To determine the duration of the impact observation window, the number of sampling points included in the window is adjusted from... and The conversion yields, It is a one-dimensional numerical index;

[0113] The controller acquires the paracord force signal output by the paracord force sensor at the same sampling time. ,in The controller is for one-dimensional force values. The validity determination is performed to obtain the valid paracord force signal. The validity determination satisfies the following constraint: reading the sensor health indicator. And require For a valid state, where As a health indicator for the paracord force sensor, Execute range determination and requirements Falling into the lower limit of the force range With the upper limit of the force range Between, among This is the lower limit value of the force range. The upper limit value of the force range is calculated by dividing the force difference between adjacent samples and then... Normalization yields the rate of change of force, and the absolute value of the rate of change of force must not exceed a threshold value. ,in The threshold for the rate of change of force is set; when the above constraints are simultaneously satisfied, the controller will... Set as When any constraint is not satisfied, the controller will Set to the previous valid sample value and set the invalid counter. ,in For consecutive invalid sample count values, the controller in After exceeding the preset invalid limit, Set to zero to avoid holding indefinitely;

[0114] The controller correlates the changes in the body acceleration with the effective parachute line force signals in a time sequence to construct the parachute descent execution state variables. The controller is and Configure unified timestamp And write to the circular buffer, the controller with The rising edge is taken as the moment of the force event. ,in For the moment of a force event, the rising edge determination uses the force trigger threshold. With minimum duration ,in The force trigger threshold, The minimum number of sampling steps required to continuously satisfy the threshold is determined by the following criteria: continuous Steps not less than The controller is in Centered on, with a length of Retrieve acceleration impact index within the associated window The maximum value is used as the associated impact amount ,in For time-series associated window duration, For a one-dimensional impact quantity aligned with the force event, the controller uses Continuous not less than the force holding threshold The cumulative duration of the stress is used to obtain the stress duration index. ,in To maintain the threshold under stress, This represents the duration of the force application, accumulated in [the following format]. Not less than Press at time Gradually accumulate and Less than The controller will be reset to zero. and Parachute execution state variables are composed in a fixed order. ,in It is a two-dimensional state quantity and includes acceleration impact index and force duration index;

[0115] The controller will execute the parachute state quantity. With lock threshold The locking threshold is compared. At least including the impact lock-in threshold With persistent lock threshold ,in This is the threshold value for impact locking. To continuously lock the threshold value, when Not less than and Not less than At that time, the controller generates a parachute mode lock command. In the locked state, where This is a parachute mode lock command; the controller remains locked if any condition is not met. In the unlocked state, the controller is When in a locked state, the parachute mode state machine is set to locked and interlocked with the command channel: the motor cut-off signal cancellation is prohibited, and the parachute trigger command is prohibited. Regenerated, while maintaining monitoring of parachute line force signals and changes in aircraft acceleration for status reporting and fault recording. The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for anti-winding parachute landing of multi-rotor unmanned aerial vehicle based on phase plane safety corridor and attitude synchronous ejection, characterized in that, The method comprises the following steps: S1, obtaining the attitude angle, angular rate, angular rate change rate of the inertial measurement unit, obtaining the height and motor state, fusing the attitude angle and angular rate to obtain the attitude solution and angular velocity solution; S2, constructing the phase plane state quantity containing the attitude angle and angular rate according to the attitude solution and angular velocity solution, determining the residual control torque according to the motor state, determining the disturbance intensity according to the angular rate change rate, determining the dynamic shrinkage safety corridor boundary varying with the height under the constraint of the residual control torque and the disturbance intensity, calculating the corridor margin of the phase plane state quantity relative to the dynamic shrinkage safety corridor boundary, and determining the parachute demand sign based on the comparison result of the corridor margin and the corridor threshold; S3, determining the propeller rotation plane according to the parachute demand sign and the structure parameter, generating the propeller disk exclusion model, and generating the ejection trajectory envelope according to the ejection device installation direction and the ejection initial speed; S4, predicting the rotation vector according to the attitude solution and angular velocity solution to obtain the roll phase prediction result, calculating the centrifugal throw-off index according to the angular velocity solution and the structure parameter, mapping the ejection trajectory envelope to the propeller disk exclusion model under the constraint of the roll phase prediction result, calculating the minimum interval and screening the clearance window satisfying the clearance threshold and the centrifugal throw-off index satisfying the throw-off threshold, backstepping the clearance window according to the actuator time delay parameter, and determining the triggering time in combination with the corridor margin; S5, generating the parachute triggering instruction according to the parachute demand sign, the triggering time and the corridor margin, wherein the parachute triggering instruction comprises the ejection triggering signal and the motor cut-off signal; S6, obtaining the parachute execution state quantity formed by the body acceleration change and the parachute rope stress signal, and generating the parachute mode locking instruction based on the parachute execution state quantity.

2. The multi-copter parachute deployment method based on phase plane safety corridor and attitude synchronized catapult of claim 1, wherein, S1 specifically comprises: reading the attitude angle and angular rate output by the inertial measurement unit at the same sampling time, and performing zero offset compensation on the angular rate to form a compensated angular rate; performing differential operation on the compensated angular rate to obtain the angular rate change rate, and expressing the attitude angle, compensated angular rate and angular rate change rate in the body coordinate system; obtaining the height and motor state, wherein the motor state comprises the motor speed and the motor enable state; taking the attitude angle and compensated angular rate as the fusion input, and adopting state filtering fusion to output the attitude solution and angular velocity solution.

3. The multi-copter parachute deployment method based on phase plane safety corridor and attitude synchronization catapulting of claim 1, wherein, S2 specifically comprises: pairing the attitude solution and angular velocity solution according to the same attitude axis to construct the phase plane state quantity containing the attitude angle component and angular rate component, and taking the phase plane state quantity as the state input for subsequent stability judgment; extracting the motor enable state and motor speed from the motor state, determining the residual control torque by mapping the motor state to the torque availability, and taking the residual control torque as the control constraint input for recoverable domain calculation; performing amplitude calculation on the angular rate change rate and forming the disturbance intensity by using the amplitude limiting rule, and taking the disturbance intensity as the external disturbance constraint input for recoverable domain calculation; generating the shrinkage coefficient varying monotonously with the height according to the height, and using the shrinkage coefficient to adjust the safety corridor boundary to form the dynamic shrinkage safety corridor boundary varying with the height; The stable domain of the phase plane state quantity is determined by using a Lyapunov stability criterion under the constraints of the remaining control torque and the disturbance intensity, and boundary parameters of a dynamically contracted safe corridor boundary are output to limit the allowable range of the phase plane state quantity; The distance between the phase plane state quantity and the boundary parameters of the dynamically contracted safe corridor boundary is calculated to obtain a corridor margin representing the relative position of the phase plane state quantity to the dynamically contracted safe corridor boundary, and the corridor margin is kept as a continuous quantity for subsequent determination of a trigger time; The corridor margin is compared with a corridor threshold, and when the corridor margin is less than or equal to the corridor threshold, a parachute deployment demand flag is generated, and when the corridor margin is greater than the corridor threshold, the parachute deployment demand flag is kept in a non-trigger state, and the corridor margin and the parachute deployment demand flag are output to a subsequent clear window screening and trigger time determination process.

4. The multi-copter parachute deployment method based on phase plane safety corridor and attitude synchronization catapulting of claim 1, wherein, In the process of determining the stable domain of the phase plane state quantity by using the Lyapunov stability criterion under the constraints of the remaining control torque and the disturbance intensity and outputting the boundary parameters of the dynamically contracted safe corridor boundary, the boundary budget corresponding to the dynamically contracted safe corridor boundary is determined by a boundary budget formula: ； wherein, is a Lyapunov boundary budget, is a reference boundary budget constant, is a moment availability coefficient, is a disturbance intensity, is a disturbance intensity reference value, is a contraction coefficient.

5. The multi-copter parachute deployment method based on phase plane safety corridor and attitude synchronization catapulting of claim 1, wherein, S3 specifically is: The structural parameters are called according to the parachute deployment demand flag, and three-dimensional coordinates of the center of the propeller rotation plane, three-dimensional coordinates of the normal of the propeller rotation plane and one-dimensional coordinates of the propeller disc radius are extracted to determine the propeller rotation plane; A propeller disc exclusion model is generated by using a conditional exclusion distance network, and the three-dimensional coordinates of the center of the propeller rotation plane, the three-dimensional coordinates of the normal of the propeller rotation plane, the one-dimensional coordinates of the propeller disc radius, the one-dimensional coordinates of the motor state and the three-dimensional coordinates of the spatial point to be determined in the machine system are used as eleven-dimensional inputs, the spatial point to be determined is transformed into a propeller disc normal distance, a propeller disc radial distance and a ratio of the radial distance to the propeller disc radius in the feature generation layer, and the motor state is mapped into an exclusion expansion to output a six-dimensional feature vector, and a one-dimensional exclusion distance is output by a three-layer fully connected network and using a linear rectifier unit; A machine system ejection direction vector is generated according to the installation direction of the catapult, and a time-evolving ejection trajectory point set is generated in combination with the initial speed of the ejection to form an ejection trajectory envelope; The propeller disc exclusion model and the ejection trajectory envelope are output.

6. The multi-copter parachute deployment method based on phase plane safety corridor and attitude synchronization catapulting of claim 1, wherein, S4 specifically is: The prediction time domain is determined by taking the attitude solution and the angular velocity solution as inputs, and the angular velocity solution is integrated to predict a rotation vector sequence; The rotation vector sequence is converted into a roll phase prediction result, and the roll phase prediction result includes the attitude change of the machine body at each time in the prediction time domain; A centrifugal throw-off index sequence is generated according to the angular velocity solution and the structural parameters, and the structural parameters at least include a throw-off radius parameter and a relative machine body coordinate parameter of the parachute rope connection point; The ejection trajectory envelope is subjected to attitude transformation under the constraint of the roll phase prediction result to obtain three-dimensional coordinates of the spatial point to be determined in the machine system at each time in the prediction time domain; The three-dimensional coordinates of the spatial point to be determined in the machine system are input into the propeller disc exclusion model to obtain the corresponding exclusion distance, and the exclusion distance at each prediction time is subjected to a minimum operation to obtain a minimum interval sequence; The minimum interval sequence is compared with the clearance threshold to obtain a clearance window satisfying the clearance threshold, and the centrifugal throwing-off indicator sequence is compared with a throwing-off threshold to re-screen the clearance window, so as to obtain a clearance window satisfying both the clearance threshold and the throwing-off threshold; The actuator time delay parameter is used to time back-propagate the clearance window to form a candidate trigger time set, and the candidate trigger time set is associated with the corridor margin to determine the trigger time.

7. The multi-copter parachute deployment method based on phase plane safety corridor and attitude synchronization catapulting of claim 1, wherein, In the process of time back-propagating the clearance window according to the actuator time delay parameter to form a candidate trigger time set, and associating the candidate trigger time set with the corridor margin to determine the trigger time, a cost function is used to associate and screen the candidate trigger time in the candidate trigger time set, and the candidate trigger time with the minimum cost is determined as the trigger time, and the cost function is: ； wherein, is a trigger time, is an operation that takes the candidate time that minimizes the bracketed value, is a candidate trigger time, is a set of candidate trigger times, is a corridor margin cost weight, is a corridor threshold, is a corridor margin, is a clearance cost weight, is a clearance threshold, is a function that takes is a function that takes is the minimum interval value read after mapping to the sequence of prediction times, is a throw-off cost weight, is a throw-off threshold, is a function that takes is a function that takes is the centrifugal throw-off indicator value read after mapping to the sequence of prediction times, is a back-pull delay, is a time priority weight, is a current sampling time, is a prediction horizon length, is an operation that takes the greater of zero and the value in the brackets.

8. The multi-copter parachute deployment method based on phase plane safety corridor and attitude synchronization catapulting of claim 1, wherein, S5 specifically is: When the parachute drop demand sign is in the trigger state, the trigger time and the corridor margin are read as instruction generation inputs; The corridor margin is compared with the linkage threshold, and when the corridor margin is not greater than the linkage threshold, it is determined that the ejection trigger signal and the motor cut-off signal are output at the same trigger time, and when the corridor margin is greater than the linkage threshold, it is determined that the motor cut-off signal is output before the ejection trigger signal and a preset interval time is determined; Based on the trigger time, a motor cut-off signal is generated, and the motor cut-off signal includes an enable closing instruction for all motors; Based on the trigger time and the preset interval time, an ejection trigger signal is generated, and the ejection trigger signal and the motor cut-off signal are packaged as a parachute drop trigger instruction.

9. The multi-copter parachute deployment method based on phase plane safety corridor and attitude synchronization catapulting of claim 1, wherein, Step S6 specifically is: The body acceleration output by the inertial measurement unit is obtained, and the body acceleration is differentiated to obtain the body acceleration change; The parachute rope force sensor output is obtained, and the parachute rope force signal is effectively determined to obtain the effective parachute rope force signal; The body acceleration change and the effective parachute rope force signal are time-sequentially associated to construct a parachute drop execution state quantity containing an acceleration impact indicator and a force duration indicator; The parachute drop execution state quantity is compared with the locking threshold, and when the parachute drop execution state quantity satisfies the locking threshold, a parachute drop mode locking instruction is generated, and the parachute drop mode locking instruction is used to maintain the parachute drop mode and prohibit the motor cut-off signal from being revoked and the parachute drop trigger instruction from being generated again.

Images