Ground-air amphibious dual-rotor unmanned aerial vehicle and ground-air unified planning method under airfoil state constraint thereof

Abstract

The invention belongs to the technical field of unmanned aerial vehicles, discloses a ground-air amphibious dual-rotor unmanned aerial vehicle and a ground-air unified planning method under airfoil state constraint thereof, and solves the problems that the airfoil state limitation of the unmanned aerial vehicle at a take-off point of the ground-air amphibious dual-rotor unmanned aerial vehicle is difficult to guarantee, a large amount of time is wasted by movement mode conversion before take-off, and the unmanned aerial vehicle cannot take off easily in the prior art. And the control continuity and smoothness during mode conversion are influenced. The method mainly comprises the following steps: step 1, fusing binocular vision and an inertial sensor to carry out positioning and mapping of an unmanned aerial vehicle; 2, generating an initial ground-air trajectory based on front-end motion dynamics path search and rear-end B-spline trajectory optimization; and step 3, generating a final ground track based on the ground track rolling optimization of the uniform expansion of the safety space. According to the invention, the continuous trajectory planning and motion control of the air-ground amphibious dual-rotor unmanned aerial vehicle in the obstacle environment are realized, the time consumed for switching the motion modes before taking off is shortened, and the control continuity and smoothness during mode switching are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of unmanned aerial vehicles, and in particular relates to a ground-to-air amphibious dual-rotor unmanned aerial vehicle and a ground-to-air unified planning method under wing surface state constraints thereof. Background Art

[0002] Design a rotor UAV with multi-mode motion capability, so that it has both air flight capability and ground motion capability. By using "wing flywheel rolling" and "cross-domain perching" in unstructured scenarios, the application scenarios of UAVs can be expanded to the ground, narrow gaps, pipelines, etc., so as to complete complex tasks such as integrated inspection and strike, cross-domain search and rescue, and contact detection. In addition, enhancing the autonomous planning and decision-making capabilities of multi-mode UAVs, so that they can correctly judge and plan a desired motion path based on environmental information in various complex environments, and actively explore the target area along the desired path, and autonomously switch motion modes to climb over various obstacles with the minimum energy cost, is of great significance and value for improving the robot's ability to adapt to complex environments, giving full play to the advantages of multi-mode motion, and enhancing the robot's autonomous control and intelligent research.

[0003] The ground-to-air amphibious flight platform based on the dual-rotor UAV has the characteristics of compact structure, can realize ground-to-air amphibious movement without adding additional actuators, and has the advantages of energy saving and noise reduction of ground movement. However, the dual-rotor UAV must ensure that the wing surface is vertically facing upward before taking off on the horizontal ground. The ground-to-air amphibious dual-rotor UAV of the present invention has the structural characteristics of "wings rotate with wheels". Before taking off, it is subject to the coupling constraints of the take-off point and the wing surface state. The existing ground-to-air planning algorithm is difficult to ensure the wing surface state restriction of the UAV at the take-off point. The motion mode conversion before take-off wastes a lot of time, delaying the execution of the overall mission trajectory. Developing a ground-to-air unified planning method under the wing surface state constraint that meets the structural conditions of the above-mentioned UAV has important practical significance for improving the safety of autonomous exploration of UAVs and the smoothness of state switching. Summary of the invention

[0004] The purpose of the present invention is to provide a ground-to-air amphibious dual-rotor UAV and a ground-to-air unified planning method under the constraints of its wing surface state, so as to solve the above-mentioned technical problems.

[0005] In order to solve the above technical problems, the specific technical solutions of the ground-to-air amphibious dual-rotor UAV and the ground-to-air unified planning method under the wing state constraint of the present invention are as follows:

[0006] A ground-to-air amphibious dual-rotor unmanned aerial vehicle comprises a flight control, a binocular camera, an airborne computer, a 360° steering gear, a flight motor and wheels. The two wheels are coaxially mounted on a transmission shaft between the steering gear and the flight motor in a fastening manner. The flight motor is mounted on the outer side of the wheel and at the end of the transmission shaft. The rotor is mounted on the flight motor and driven to rotate by the flight motor. The rotor rotates with the wheel. The flight control, binocular camera, airborne computer and 360° steering gear are mounted on the transmission shaft. The 360° steering gear replaces the tilt steering gear on the traditional dual-rotor, and can not only control the tilt of the motor in the air mode to realize the motion flight of the dual-rotor, but also drive the differential rotation of the wheels in the ground mode to realize the control movement of the two-wheel vehicle. The airborne computer is used to run the ground-to-air unified planning algorithm, and convert the finally generated trajectory into a service message and send it to the flight control for execution after dynamic solution. The binocular camera is used to identify the environment and locate.

[0007] The present invention also discloses a ground-to-air unified planning method for a ground-to-air amphibious dual-rotor UAV under the wing surface state constraint, comprising the following steps:

[0008] Step 1: Fuse binocular vision and inertial sensors to locate and map the drone;

[0009] Step 2: Initial ground-to-air trajectory generation based on front-end motion dynamics path search and back-end B-spline trajectory optimization;

[0010] Step 3: Generate the final ground track based on the rolling optimization of the ground track with uniform expansion of the safety space. Furthermore, the step 1 uses the binocular camera 2 to obtain the image data within the field of view of the ground-to-air amphibious dual-rotor UAV in real time and integrates the IMU data information built into the flight control 1 to construct a three-dimensional grid map and perform environmental positioning.

[0011] Furthermore, the environmental positioning extracts the above-mentioned binocular image features through the Shi-Tomasi corner detection method, and uses the LK optical flow algorithm to track the image features, pre-integrates the IMU data between the two frames of images, and then incorporates the image observation results and the IMU pre-integration results into the back-end sliding window optimization algorithm to calculate the drone position and attitude values.

[0012] Furthermore, in the construction of the three-dimensional grid map, each cubic space is taken as a voxel, and whether it is occupied by an obstacle is determined based on the Bayesian probability law. By calculating the distance from the unoccupied voxel to the nearest obstacle surface, a positive or negative sign is assigned according to whether the voxel is inside or outside the obstacle. The generated distance field is smoothed and optimized to obtain a complete continuous distance field ESDF for the subsequent trajectory planning algorithm; each voxel has a corresponding precise signed distance value for the subsequent trajectory planning algorithm.

[0013] Furthermore, the step 2 uses Kinodynamic A* to preferentially search for a dynamic path, uses motion primitives instead of straight lines as graph edges, and adds additional energy costs to motion primitives in the air to ensure that the path search tends to plan a ground path;

[0014] When expanding the motion primitive outward, the discrete degree of the input quantity is controlled based on the current yaw angle of the drone, thereby limiting the search angle range;

[0015] The change of z-axis height information in the path information is used to determine the switching of motion modes, and thus divide the ground track and the air track.

[0016] Furthermore, the dynamic path is converted into a mathematical trajectory expression using a B-spline curve, and cost functions for smoothness terms, dynamic feasibility terms and collision terms are constructed to construct an optimization problem, thereby performing gradient-based B-spline trajectory optimization.

[0017] Furthermore, in the trajectory optimization, the generated path is reparameterized and represented by a uniform B-spline through a series of control points Q = {Q0, Q1, ..., Q N}, where Q a , Q t To distinguish between air and ground control points, in particular, ground control points are described as Each control point has two-dimensional coordinates Composition, i∈[0,M], the following cost function is introduced for optimization:

[0018] f1=λ s f s +λ c f c +λ f (f v +f a )

[0019] where f s is a smoothing cost term designed as an elastic band cost function, f c is the collision cost based on ESDF gradient information, f v and f a is the dynamic feasibility cost of limiting speed and acceleration, λ s ,λ c ,λ f is the weight of each cost term. Due to the convex hull property of B-spline, the above cost terms only constrain the control point Q to ensure safety and dynamic feasibility;

[0020] Q tAn additional penalty is imposed to limit the curvature of the ground track to avoid the track being too curved and causing large turning tracking errors:

[0021]

[0022] Among them, F n (Q ti ) is a differentiable cost function, C max The maximum curvature threshold is specified as:

[0023]

[0024] Therefore, the overall objective cost function is expressed as follows:

[0025] f total =λ s f s +λ c f c +λ f (f v +f a )+λ n f n .

[0026] Furthermore, the ground track rolling optimization in step 3 is triggered when a take-off point appears in the planning field of view, and the integer multiple relationship between the remaining ground track length before take-off and the wheel circumference is compared to determine whether the soft constraint of the wing-up state is satisfied when the two-wheeled vehicle mode runs to the take-off point: θ′ L ,θ′ R ∈[-θ°, θ°], where θ′ L ,θ′ R are the angles between the left and right rotors and the z-axis when theoretically reaching the take-off point, and θ is the tolerance value of the rotor angle. By uniformly expanding the current trajectory length in the safety space and continuously approaching the soft constraint of the wing-up state, a final ground trajectory is optimized;

[0027] The finally optimized ground trajectory and air trajectory are sent to the flight control 1 via the Mavlink communication protocol via the onboard computer 3, and the position information in the trajectory is resolved into the rotation speed of the flight motor 5 and the wheel speed of the wheel 6 of the drone for tracking.

[0028] Furthermore, before takeoff, the flight control 1 servo controls the 360° servo 4 to return to the 0 point to satisfy θ′ L =θ′ R =θ=0° is a hard constraint upward on the wing surface.

[0029] The present invention provides a ground-to-air amphibious dual-rotor UAV and a ground-to-air unified planning method under wing surface state constraints thereof, which has the following advantages: the present invention develops a ground-to-air unified planning method for a ground-to-air amphibious dual-rotor UAV that satisfies its wing surface state constraints, realizes continuous trajectory planning and motion control of the ground-to-air amphibious dual-rotor UAV in an obstacle environment, shortens the time consumption of motion mode switching before takeoff, and improves control continuity and smoothness during mode conversion. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of the ground-to-air amphibious dual-rotor UAV of the present invention.

[0031] Figure 2 This is a framework diagram of the ground-to-air amphibious dual-rotor UAV and the ground-to-air unified planning method under the constraints of its wing surface state.

[0032] Figure 3 This is a flow chart of ground track rolling optimization judgment of the present invention.

[0033] Explanation of the reference numerals: 1. flight control; 2. binocular camera; 3. onboard computer; 4. 360° servo; 5. flight motor; 6. wheel. DETAILED DESCRIPTION

[0034] In order to better understand the purpose, structure and function of the present invention, the following is a further detailed description of a ground-to-air amphibious dual-rotor UAV and a ground-to-air unified planning method under wing surface state constraints of the present invention in conjunction with the accompanying drawings.

[0035] like Figure 1 As shown, a ground-to-air amphibious dual-rotor unmanned aerial vehicle of the present invention comprises a flight control 1, a binocular camera 2, an onboard computer 3, a 360° steering gear 4, a flight motor 5 and a wheel 6. The two wheels 6 are concentrically mounted on the transmission shaft between the steering gear 4 and the flight motor 5 in a fastening manner, and a spoke space is reserved to avoid blocking the propeller rotation interference of the flight motor 5. The flight motor 5 is mounted on the outer side of the wheel 6 and the end of the transmission shaft. The rotor is mounted on the flight motor 5 and driven to rotate by the flight motor 5. The rotor rotates with the wheel 6. The flight control 1, the binocular camera 2, the onboard computer 3, and the 360° steering gear 4 are mounted on the transmission shaft. The 360° steering gear 4 replaces the tilt steering gear on the traditional dual-rotor, which can not only control the motor tilt in the air mode to realize the motion flight of the dual-rotor, but also drive the wheel 6 to rotate differentially in the ground mode to realize the control movement of the two-wheel vehicle. The onboard computer 3 is used to run the ground-to-air unified planning algorithm, and convert the finally generated trajectory into a service message and send it to the flight control 1 for execution after dynamic solution. The binocular camera 2 is used to identify the environment and locate.

[0036] like Figure 2 , Figure 3As shown, the ground-to-air unified planning method under the wing state constraint for the functional characteristics of the ground-to-air amphibious dual-rotor UAV of the present invention mainly includes the following steps:

[0037] Step 1: Fuse binocular vision and inertial sensors to locate and map the drone;

[0038] The binocular camera 2 is used to obtain the image data within the field of view of the ground-to-air amphibious dual-rotor UAV in real time and integrate the IMU data information built into the flight control 1 to construct a three-dimensional grid map and perform environmental positioning;

[0039] The environmental positioning extracts the binocular image features by Shi-Tomasi corner detection method, and uses LK optical flow algorithm to track image features, pre-integrate IMU data between two frames of images, and then incorporates image observation results and IMU pre-integration results into the back-end sliding window optimization algorithm to calculate the position and attitude value of the drone;

[0040] In the construction of the three-dimensional grid map, each cubic space is regarded as a voxel, and whether it is occupied by an obstacle is determined based on the Bayesian probability law. By calculating the distance from the unoccupied voxel to the nearest obstacle surface, a positive or negative sign is assigned according to whether the voxel is inside or outside the obstacle. The generated distance field is smoothed and optimized to obtain a complete continuous distance field ESDF, which is used in the subsequent trajectory planning algorithm;

[0041] Each voxel has a corresponding accurate signed distance value for use in the subsequent trajectory planning algorithm;

[0042] Step 2: Initial ground-to-air trajectory generation based on front-end motion dynamics path search and back-end B-spline trajectory optimization;

[0043] Use Kinodynamic A* to prioritize a dynamic path, use motion primitives instead of straight lines as graph edges, and add extra energy costs to motion primitives in the air to ensure that the path search tends to plan a ground path;

[0044] In order to ensure that the initial trajectory satisfies the kinematic constraints of the ground-to-air amphibious dual-rotor UAV as much as possible, especially the turning radius limit of the ground two-wheel vehicle mode, when expanding the motion primitive outward, the discrete degree of the input quantity is controlled based on the current yaw angle of the UAV, thereby limiting the search angle range;

[0045] The change of z-axis height information in the path information is used to determine the switching of motion modes, and thus to divide the ground track and the air track;

[0046] The above dynamic path is converted into a mathematical trajectory expression using a B-spline curve, and the cost functions for the smoothness term, the dynamic feasibility term and the collision term are constructed to construct the optimization problem, and then the gradient-based B-spline trajectory optimization is performed;

[0047] In trajectory optimization, we reparameterize the generated path using a uniform B-spline representation, which is characterized by a series of control points Q = {Q0, Q1, ..., Q N} to define. Among them, Q a , Q t To distinguish between air and ground control points, in particular, ground control points are described as Each control point has two-dimensional coordinates Composition, i∈[0,M]. In order to improve the quality of the trajectory, we introduce the following cost function for optimization:

[0048] f1=λ s f s +λ c f c +λ f (f v +f a )

[0049] where f s is a smoothing cost term designed as an elastic band cost function. c is the collision cost based on the ESDF gradient information. v and f a is the dynamic feasibility cost of limiting speed and acceleration. s ,λ c ,λ f is the weight of each cost term. Due to the convex hull property of B-spline, the above cost terms only constrain the control point Q to ensure safety and dynamic feasibility.

[0050] Q t An additional penalty is imposed to limit the curvature of the ground track to avoid the track being too curved and causing large turning tracking errors:

[0051]

[0052] Among them, F n (Q ti ) is a differentiable cost function, C max The maximum curvature threshold is specified as:

[0053]

[0054] Therefore, the overall objective cost function is expressed as follows:

[0055] f total =λ s f s +λ c f c +λ f (f v +f a )+λ n f n

[0056] Step 3: The final ground track generation based on the ground track rolling optimization with uniform expansion of the safety space.

[0057] The ground track rolling optimization is triggered when the take-off point appears in the planning field of view. The relationship between the remaining ground track length before take-off and the integer multiple of the wheel circumference is compared to determine whether the soft constraint of the wing-up state is met when running to the take-off point in the two-wheeled vehicle mode: θ L ′,θ′ R ∈[﹣θ°,θ°], where θ L ′、θ′ R are the angles between the left and right rotors and the z-axis when theoretically reaching the take-off point, and θ is the tolerance value of the rotor angle. By uniformly expanding the current trajectory length in the safety space and continuously approaching the soft constraint of the wing-up state, a final ground trajectory is optimized;

[0058] In order to avoid the difficulty in solving due to the strict restriction of the wing upward constraint, the rotor angle tolerance value θ can be adjusted according to the specific situation;

[0059] The finally optimized ground track and air track are sent to the flight control 1 by the onboard computer 3 through the Mavlink communication protocol, and the position information in the track is solved into the rotation speed of the flight motor 5 and the wheel speed of the wheel 6 of the UAV for tracking;

[0060] Furthermore, in order to ensure that the wing surface of the ground-to-air amphibious dual-rotor UAV of the present invention faces upward at the take-off point, the flight control 1 servo controls the 360° steering gear 4 to return to the 0 point before take-off to satisfy θ L ′=θ′ R =θ=0° is a hard constraint upward on the wing surface.

[0061] It is to be understood that the present invention is described by some embodiments, and it is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the scope of protection of the present invention.

Claims

1. A ground-to-air amphibious dual-rotor drone, comprising a flight control (1), a binocular camera (2), an onboard computer (3), a 360° steering gear (4), a flight motor (5) and wheels (6), characterized in that: The two wheels (6) are coaxially mounted on a transmission shaft between a steering gear (4) and a flight motor (5) in a fastening manner. The flight motor (5) is mounted on the outer side of the wheel (6) and at the end of the transmission shaft. The rotor is mounted on the flight motor (5) and driven to rotate by the flight motor (5). The rotor rotates along with the wheel (6). The flight control (1), the binocular camera (2), the airborne computer (3), and the 360° steering gear (4) are mounted on the transmission shaft. The 360° steering gear (4) replaces the tilt steering gear on the traditional twin rotors and can not only control the tilt of the motor in the air mode to realize the motion flight of the twin rotors, but also drive the wheels (6) to rotate at a differential speed in the ground mode to realize the control movement of the two-wheeled vehicle. The airborne computer (3) is used to run a ground-to-air unified planning algorithm and convert the finally generated trajectory into a service message and send it to the flight control (1) for execution after dynamics solution. The binocular camera (2) is used to identify the environment and locate.

2. The ground-to-air unified planning method of the ground-to-air amphibious dual-rotor UAV under the wing surface state constraint according to claim 1 is characterized in that: The steps include: Step 1: Fuse binocular vision and inertial sensors to locate and map the drone; Step 2: Initial ground-to-air trajectory generation based on front-end motion dynamics path search and back-end B-spline trajectory optimization; Step 3: The final ground track generation based on the ground track rolling optimization with uniform expansion of the safety space.

3. The ground-air unified planning method according to claim 2, characterized in that: The step 1 uses a binocular camera (2) to obtain image data within the field of view of the ground-to-air amphibious dual-rotor UAV in real time and fuses the IMU data information built into the flight control (1) to construct a three-dimensional grid map and perform environmental positioning.

4. The ground-air unified planning method according to claim 3 is characterized in that: The environmental positioning extracts the above binocular image features through the Shi-Tomasi corner detection method, and uses the LK optical flow algorithm to track the image features, pre-integrates the IMU data between the two frames of images, and then incorporates the image observation results and the IMU pre-integration results into the back-end sliding window optimization algorithm to calculate the drone position and attitude values.

5. The ground-air unified planning method according to claim 3, characterized in that: In the three-dimensional grid map, each cubic space is taken as a voxel, and whether it is occupied by an obstacle is determined based on the Bayesian probability law. By calculating the distance from the unoccupied voxel to the nearest obstacle surface, a positive or negative sign is assigned according to whether the voxel is inside or outside the obstacle. The generated distance field is smoothed and optimized to obtain a complete continuous distance field ESDF for the subsequent trajectory planning algorithm; each voxel has a corresponding accurate signed distance value for the subsequent trajectory planning algorithm.

6. The ground-air unified planning method according to claim 2, characterized in that: The step 2 uses KinodynamicA* to preferentially search for a dynamic path, uses motion primitives instead of straight lines as graph edges, and adds additional energy costs to motion primitives in the air to ensure that the path search tends to plan a ground path; When expanding the motion primitive outward, the discrete degree of the input quantity is controlled based on the current yaw angle of the drone, thereby limiting the search angle range; The change of z-axis height information in the path information is used to determine the switching of motion modes, and thus divide the ground track and the air track.

7. The ground-air unified planning method according to claim 6, characterized in that: The dynamic path is converted into a mathematical trajectory expression using a B-spline curve, and cost functions for smoothness terms, dynamic feasibility terms and collision terms are constructed to construct an optimization problem, thereby performing gradient-based B-spline trajectory optimization.

8. The ground-air unified planning method according to claim 7, characterized in that: In the trajectory optimization, the generated path is reparameterized and represented by uniform B-spline through a series of control points Q = {Q0, Q1, ..., Q N }, where Q a , X t To distinguish between air and ground control points, in particular, ground control points are described as Each control point has two-dimensional coordinates Composition, i∈[0,M], the following cost function is introduced for optimization: f1=λ s f s +λ c f c +λ f (f v +f a ) where f s is a smoothing cost term designed as an elastic band cost function, f c is the collision cost based on ESDF gradient information, f v and f a is the dynamic feasibility cost of limiting speed and acceleration, λ s ,λ c ,λ f is the weight of each cost term. Due to the convex hull property of B-spline, the above cost terms only constrain the control point Q to ensure safety and dynamic feasibility; Q t An additional penalty is imposed to limit the curvature of the ground track to avoid the track being too curved and causing large turning tracking errors: Among them, F n (Q ti ) is a differentiable cost function, c max The maximum curvature threshold is specified as: Therefore, the overall objective cost function is expressed as follows: f total =λ s f s +λ c f c +λ f (f v +f a )+λ n f n 。 9. The ground-air unified planning method according to claim 2, characterized in that: The ground track rolling optimization in step 3 is triggered when a take-off point appears in the planning field of view. The relationship between the remaining ground track length before take-off and the integer multiple of the wheel circumference is compared to determine whether the soft constraint of the wing-up state is satisfied when the two-wheeled vehicle mode runs to the take-off point: θ′ L ,θ′ R ∈[﹣θ°,θ°], where θ′ L ,θ′ R are the angles between the left and right rotors and the z-axis when theoretically reaching the take-off point, and θ is the tolerance value of the rotor angle. By uniformly expanding the current trajectory length in the safety space and continuously approaching the soft constraint of the wing-up state, a final ground trajectory is optimized; The finally optimized ground trajectory and air trajectory are sent to the flight control (1) via the onboard computer (3) through the Mavlink communication protocol, and the position information in the trajectory is resolved into the rotation speed of the UAV's flight motor (5) and the wheel speed (6) for tracking.

10. The ground-air unified planning method according to claim 9, characterized in that: Before takeoff, the flight control (1) servo controls the 360° servo (4) to return to the zero point to satisfy θ′ L =θ′ R =θ=0° is a hard constraint upward on the wing surface.

Images