A flying robot and its flight control method

Abstract

The present invention provides a flying robot and its flight control method; the flying robot is characterized by comprising an airbag main body (100), a rotor power system (200) mounted on the airbag main body (100) and used for driving the airbag main body (100), a control system (300) mounted on the airbag main body (100) and electrically connected to the rotor power system (200) and used for controlling the operation of the rotor power system (200), and a ground system (400) communicatively connected to the control system (300) and used for sending control instructions to the control system (300). The flying robot of the present invention greatly increases the endurance time by combining the rotor power system with the airbag main body. The flight control method of the present invention can eliminate the coupling of the mobility and attitude adjustment of the flexible flying robot.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicles, and particularly to a flying robot and a flight control method thereof. Background Art

[0002] In the past decade or more, unmanned aerial vehicles based on the horizontal coaxial multi-rotor technology have developed rapidly, breaking through the boundaries of the fixed-wing and helicopter technology systems and creating a field of unmanned aerial vehicles with hover self-stabilization and 6-degree-of-freedom maneuverability. The current key development direction of unmanned aerial vehicle technology is the electric vertical takeoff and landing unmanned aerial vehicle eVTOL combined with fixed wings, and the design goal is to develop towards traditional aircraft with long range and large payload. However, another frontier development direction is to continue the advantages of the multi-rotor technology system, and the design goal is to develop towards a flying robot with super maneuverability, safe and easy to operate, and long endurance. Its application scenarios are in ultra-low altitude complex ground environments indoors and outdoors to complete aerial operations such as aerial photography monitoring, remote sensing monitoring, voice interaction, and visual obstacle avoidance, and even multi-machine complex collaborations involving physical contact and mechanical interaction such as charging and battery swapping, logistics relay, and industrial cleaning.

[0003] Traditionally, flying robots have horizontal coaxial four-rotors that control the attitude based on the rotational speed difference, and later vertical coaxial dual-rotors based on ducted fans. However, they all have the problems of short endurance due to large self-weight and potential safety hazards of crashing and colliding caused by rigid structures. Summary of the Invention

[0004] The present invention aims at the above technical problems and provides a flying robot and a flight control method thereof.

[0005] The technical solution proposed by the present invention is as follows:

[0006] The present invention proposes a flying robot, which includes an airbag body, a rotor power system installed on the airbag body for driving the airbag body, a control system installed on the airbag body and electrically connected to the rotor power system for controlling the operation of the rotor power system, and a ground system communicatively connected to the control system for sending control commands to the control system.

[0007] In the above flying robot of the present invention, the rotor power system further includes four sets of vector propulsion rotors, and the four sets of vector propulsion rotors are evenly distributed in the circumferential direction of the airbag body;

[0008] The vector propulsion rotor includes a first motor installed on the airbag body and electrically connected to the control system; the first motor has a horizontally oriented propeller.

[0009] In the above flying robot of the present invention, the rotor power system includes four sets of horizontal coaxial rotors, and the four sets of horizontal coaxial rotors are evenly distributed in the circumferential direction of the airbag body;

[0010] The horizontal coaxial rotors include a servo mounted on the airbag body and a second motor mounted on the servo and electrically connected to the control system; the second motor has a vertically oriented propeller.

[0011] In the above-mentioned flying robot of the present invention, the four sets of horizontal coaxial rotors and the four sets of vector propulsion rotors are arranged at intervals and alternately.

[0012] The present invention proposes a flight control method for the above-mentioned flying robot, including the following steps:

[0013] Step 1: Construct an incremental reinforcement learning controller, using the flight state and attitude information of the flying robot and the flight stability measurement value as the input layer nodes of the incremental reinforcement learning controller, and using the flight control parameters corresponding to the flight state and attitude information of the flying robot and the flight stability measurement value as the output layer nodes of the incremental reinforcement learning controller;

[0014] At the same time, construct a reward function RF according to the weighting coefficient given by different scenarios and the flight stability measurement value of the flying robot;

[0015] Step 2: Collect the historical flight state and attitude information of the flying robot, the corresponding historical flight control parameters and the historical flight stability measurement value, and input them into the incremental reinforcement learning controller for training together to obtain a preliminarily trained reinforcement learning controller;

[0016] Step 3: Use the preliminarily trained reinforcement learning controller to control the flight of the flying robot, and during the flight control, calculate the reward function RF in real time; when the reward function RF calculated in real time exceeds the preset threshold, use the corresponding flight state and attitude information and the flight stability measurement value to perform iterative learning training on the preliminarily trained reinforcement learning controller to generate corresponding flight control parameters;

[0017] Among them, when performing iterative learning training, calculate the change rate of the flight state and attitude state of the flying robot according to the flight state and attitude information of the flying robot, and then use the change rate of the flight state and attitude state of the flying robot and use the adaptive incremental algorithm to enhance and strengthen the preliminarily trained reinforcement learning controller, so that the flight control parameters output by the enhanced and strengthened reinforcement learning controller converge; then use the enhanced and strengthened reinforcement learning controller to control the flight of the flying robot.

[0018] In the above-mentioned flight control method of the present invention, the flight state and attitude information of the flying robot includes GPS data, geomagnetic data, barometer data, gyroscope data, rotor angle data, rotor speed data and flight speed data.

[0019] In the above flight control method of the present invention, the flight stability measurement values of the flying robot include, but are not limited to, the X-axis offset value, Y-axis offset value, and Z-axis offset value of the flying robot.

[0020] In the above flight control method of the present invention, the flight control parameters of the flying robot include the action control parameters when the flying robot performs pitch, roll, and yaw attitude actions.

[0021] The flying robot of the present invention combines a rotor power system with an airbag main body. By filling the airbag main body with helium or hydrogen and then carrying out weight trimming, it obtains the hovering ability with a static lift of 0, greatly increasing the endurance time. At the same time, the airbag main body provides soft protection for the rigid structure, not only increasing the safety and robustness of the airframe, but also being friendly to the operating environment. The flight control method of the present invention can integrate the flight control of a vector-pushing quadrotor and a horizontal coaxial quadrotor. When two rotors in the horizontal coaxial quadrotor rotate in a specific direction to generate an upward moment, the other two corresponding rotors rotate in the opposite direction to generate a downward moment. In this way, the two cancel each other out and no longer interfere with the vector-pushing rotors, but only generate attitude adjustments for the airframe, eliminating the coupling between the mobility and attitude adjustment of the flexible flying robot. Description of the Drawings

[0022] Figure 1 The structural schematic diagram of the flying robot according to the preferred embodiment of the present invention is shown. Detailed Embodiments

[0023] In order to make the technical solutions, technical objectives, and technical effects of the present invention clearer so that those skilled in the art can understand and implement the present invention, the present invention will be further described in detail below in conjunction with the drawings and specific embodiments.

[0024] As Figure 1 shown, Figure 1 The structural schematic diagram of the flying robot according to the preferred embodiment of the present invention is shown. The flying robot includes an airbag main body 100, a rotor power system 200 installed on the airbag main body 100 for driving the airbag main body 100, a control system 300 installed on the airbag main body 100 and electrically connected to the rotor power system 200 for controlling the operation of the rotor power system 200, and a ground system 400 communicatively connected to the control system 300 for sending control commands to the control system 300.

[0025] The above technical solution is the basic solution. By combining the rotor power system 200 with the airbag main body 100, filling the airbag main body 100 with helium or hydrogen, and then carrying a counterweight for trimming, a hovering ability with a static lift of 0 is obtained, greatly increasing the endurance time. At the same time, the airbag main body 100 provides soft protection for the rigid structure, not only increasing the safety and robustness of the airframe, but also being friendly to the operating environment.

[0026] In the above technical solution, the functions of the rotor power system 200 are as follows: 1) Generate lift to balance the gravity of the flying robot and the component forces in the vertical direction; 2) Generate a forward horizontal component force to overcome air resistance and make the flying robot move forward; 3) When the flying robot hovers, generate a lateral or backward horizontal component force to make the flying robot fly sideways or backward; 4) Generate component forces and torques to control or maneuver the flying robot, similar to various control surfaces on an airplane. Different from aerostats and airships, the flying robot of the present invention does not obtain positive and negative buoyancy to move up and down by changing the volume and density of the airbag main body 100. Instead, based on the vector propulsion technology of the rotor power system 200, it controls an aircraft with a weight of 0 to achieve super maneuverability in six degrees of freedom of up, down, front, back, left, and right, and realizes attitude control of pitch, roll, and yaw.

[0027] Furthermore, the airbag main body 100 can be a special-shaped airbag with different shapes, filled with low-density gases such as helium or hydrogen in the middle to generate buoyancy, thereby reducing the load driven by the battery and greatly extending the endurance time. The size and shape of the airbag main body 100 can be calculated through buoyancy, power, and load by comprehensively considering actual load, endurance, and other requirements. Its aerodynamic characteristics can be obtained through wind tunnel tests and software calculation simulations to obtain attitude adjustment parameters. The present invention takes a spherical airbag as an example, but is not limited to spherical, disc-shaped, or other centrally symmetric variable bodies, etc.

[0028] Furthermore, the rotor power system 200 further includes four sets of vector propulsion rotors 220, and the four sets of vector propulsion rotors 220 are evenly distributed in the circumferential direction of the airbag main body 100;

[0029] The vector propulsion rotor 220 includes a first motor 221 installed on the airbag main body 100 and electrically connected to the control system 300; the first motor 221 has a horizontally oriented propeller.

[0030] Here, the four sets of vector propulsion rotors 220 are mounted at symmetric positions on both sides of the equator on the surface of the airbag main body 100, and on the spherical airbag main body 100, they are at positions separated by 90 degrees every other in the middle of the spherical surface. The orientation of the vector propulsion rotor 220 is the horizontal direction, generating a pulling force or a thrust in the horizontal direction. By controlling and managing the different rotational speeds of the multiple vector propulsion rotors 220, corresponding vector synthesis thrusts can be generated to push the airbag main body 100 to move in six degrees of freedom of up, down, forward, left, and right.

[0031] Furthermore, the rotor power system 200 includes four sets of horizontally coaxial rotors 210, and the four sets of horizontally coaxial rotors 210 are evenly distributed in the circumferential direction of the airbag main body 100;

[0032] The horizontally coaxial rotor 210 includes a servo 211 mounted on the airbag main body 100 and a second motor 212 mounted on the servo 211 and electrically connected to the control system 300; the second motor 212 has a vertically oriented propeller.

[0033] Here, the four sets of horizontally coaxial rotors 210 are mounted at symmetric positions on both sides of the equator on the surface of the airbag main body 100, and on the spherical airbag main body 100, they are at positions separated by 90 degrees from each other in the middle of the spherical surface. The four sets of horizontally coaxial rotors 210 and the four sets of vector propulsion rotors 220 are arranged alternately at intervals of 45 degrees. The directions of the four sets of horizontally coaxial rotors 210 are in the vertical direction, generating pulling force or thrust in the vertical direction. When two rotors among the four sets of horizontally coaxial rotors 210 generate vertical upward moment components, the corresponding other two rotors generate vertical downward moment components, which cancel each other out, so that only pitch, roll, and yaw attitude movements are generated.

[0034] Furthermore, the control system 300 includes a main control module, a power battery, a communication module, etc., and can be dispersed or integrated in a pod and mounted at the bottom of the airbag main body 100, so that the overall center of gravity of the airbag main body 100 is relatively low, thereby maintaining the vertical orientation of the airbag main body 100. The main control module outputs signals through a control algorithm to control the motors of the vector propulsion rotor structure, thereby controlling the movement of the flying robot. The main control module also controls the attitude of the flying robot through the horizontally coaxial rotors 210. The size and capacity of the power battery are configured according to the requirements of the entire set of horizontally coaxial rotors 210. The communication module is used for communication between the main control module and the ground system. The control system 300 can add other modules such as a positioning module as needed to achieve the ability of autonomous flight.

[0035] Furthermore, the main control module includes a basic flight control module based on PID and a PID parameter adjustment and optimization module based on reinforcement learning. In the present invention, the theoretical parameters of the basic flight control module are first set as a baseline, and the PID parameters are calibrated through a large number of tests and tuning based on various methods such as critical ratio and response curve. Because the flying robot introduces horizontally coaxial rotors for attitude control, when the control accuracy jitters greatly and an overflowing component moment is generated in the horizontal direction, it will interfere with the maneuverability of the flying robot. Therefore, when exceeding the PID control boundary, the tuning module will be activated to further test and adjust the parameters. The present invention achieves the purpose of quickly converging and calibrating the PID parameters through the reinforcement learning algorithm.

[0036] The present invention proposes a flight control method based on the above-mentioned flying robot, including the following steps:

[0037] Step 1: Construct an Incremental Reinforcement Learning (IRL) controller. Use the flight state and attitude information of the flying robot and the flight stability measurement value as the input layer nodes of the incremental reinforcement learning controller, and use the flight control parameters corresponding to the flight state and attitude information of the flying robot and the flight stability measurement value as the output layer nodes of the incremental reinforcement learning controller;

[0038] Meanwhile, construct a reward function RF according to the weighting coefficients given in different scenarios and the flight stability measurement value of the flying robot;

[0039] In this step, the flight state and attitude information of the flying robot are collected by various sensors on the flying robot, including but not limited to GPS data, geomagnetic data, barometer data, gyroscope data, rotor angle data, rotor speed data, and flight speed data;

[0040] The flight stability measurement value of the flying robot includes but not limited to the X-axis offset value, Y-axis offset value, and Z-axis offset value of the flying robot;

[0041] The flight control parameters of the flying robot include but not limited to the action control parameters when the flying robot performs pitch, roll, and yaw attitude actions.

[0042] In this step, the constructed reward function RF is used to evaluate the stability of the incremental reinforcement learning controller.

[0043] Step 2: Collect the historical flight state and attitude information of the flying robot, the corresponding historical flight control parameters, and the historical flight stability measurement value, and input them into the incremental reinforcement learning controller for training to obtain a preliminarily trained reinforcement learning controller;

[0044] Step 3: Use the preliminarily trained reinforcement learning controller to control the flight of the flying robot, and calculate the reward function RF in real time during the flight control; when the reward function RF calculated in real time exceeds the preset threshold, use the corresponding flight state and attitude information and the flight stability measurement value to perform iterative learning training on the preliminarily trained reinforcement learning controller to generate corresponding flight control parameters;

[0045] Among them, during iterative learning training, calculate the change rate of the flight state and attitude of the flying robot according to the flight state and attitude information of the flying robot, and then use the change rate of the flight state and attitude of the flying robot and use the adaptive incremental algorithm to enhance and strengthen the preliminarily trained reinforcement learning controller to make the flight control parameters output by the enhanced and strengthened reinforcement learning controller converge; then use the enhanced and strengthened reinforcement learning controller to control the flight of the flying robot.

[0046] In this step, enhancing and strengthening the initially trained reinforcement learning controller according to the flight state and the change rate of the attitude state of the flying robot and using an adaptive increment algorithm includes:

[0047] Increasing or decreasing the increment value according to the flight state and the change rate of the attitude state of the flying robot, and using the adaptive increment algorithm to enhance and strengthen the convergence of the initially trained reinforcement learning controller.

[0048] Through repeated iteration, the flight control parameters are converged to meet the flight state and attitude control conditions and the flight stability conditions, so as to confirm the flight control parameters and transfer the parameters to the PID main control module.

[0049] The ground system 400 generally consists of a remote controller and / or a control computer, manually remotely controls the attitude and movement of the flexible robot, and sends commands such as autonomous flight based on various attitude and movement information of the flexible robot, including movement in six degrees of freedom and attitude adjustments such as pitch, roll, and yaw.

[0050] The present invention describes a flying robot integrating a vector-pushing rotor and a horizontal coaxial rotor and a specific control method. Since the traditional horizontal coaxial multi-rotor technology provides a torque in the horizontal direction through the speed difference between different rotors, and each rotor will form a vertically upward torque, simply combining a horizontal coaxial quadcopter and a vector-pushing quadcopter will cause redundant torques in each degree of freedom, and the two will be coupled and interfere with each other. The present invention uses an optimized PID control method to integrate the flight control of the vector-pushing quadcopter and the horizontal coaxial quadcopter. When two rotors in the horizontal coaxial quadcopter rotate in a specific direction to generate an upward torque, the other two corresponding rotors rotate in the opposite direction to generate a downward torque. In this way, the two cancel each other out and no longer interfere with the vector-pushing rotor, but only generate an attitude adjustment for the fuselage, eliminating the coupling between the mobility and attitude adjustment of the flexible flying robot.

[0051] The embodiments of the present invention have been described above with reference to the drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the claims, and all of these fall within the protection scope of the present invention.

Claims

1. A flight control method for a flying robot, characterized in that The flying robot includes an airbag main body (100), a rotor power system (200) installed on the airbag main body (100) for driving the airbag main body (100), a control system (300) installed on the airbag main body (100) and electrically connected to the rotor power system (200) for controlling the operation of the rotor power system (200), and a ground system (400) communicatively connected to the control system (300) for sending control instructions to the control system (300); the flight control method includes the following steps: Step 1, construct an incremental reinforcement learning controller, using the flight state and attitude information of the flying robot and the flight stability measurement value as the input layer nodes of the incremental reinforcement learning controller, and using the flight control parameters corresponding to the flight state and attitude information of the flying robot and the flight stability measurement value as the output layer nodes of the incremental reinforcement learning controller; Meanwhile, construct a reward function RF according to the weighting coefficient given by different scenarios and the flight stability measurement value of the flying robot; Step 2, collect the historical flight state and attitude information of the flying robot, the corresponding historical flight control parameters and the historical flight stability measurement value, and input them into the incremental reinforcement learning controller for training together to obtain a preliminarily trained reinforcement learning controller; Step 3, perform flight control on the flying robot through the preliminarily trained reinforcement learning controller, and during the flight control, calculate the reward function RF in real time; when the reward function RF calculated in real time exceeds the preset threshold, use the corresponding flight state and attitude information and the flight stability measurement value to perform iterative learning training on the preliminarily trained reinforcement learning controller to generate corresponding flight control parameters; Among them, during iterative learning training, calculate the change rate of the flight state and attitude of the flying robot according to the flight state and attitude information of the flying robot, and then use the adaptive incremental algorithm to enhance and strengthen the preliminarily trained reinforcement learning controller according to the change rate of the flight state and attitude of the flying robot, so that the flight control parameters output by the enhanced and strengthened reinforcement learning controller converge; then use the enhanced and strengthened reinforcement learning controller to perform flight control on the flying robot.

2. The flight control method according to claim 1, wherein The flight state and attitude information of the flying robot includes GPS data, geomagnetic data, barometer data, gyroscope data, rotor angle data, rotor speed data and flight speed data.

3. The flight control method according to claim 1, wherein The flight stability measurement value of the flying robot includes but is not limited to the X-axis offset value, Y-axis offset value and Z-axis offset value of the flying robot.

4. The flight control method according to claim 1, wherein The flight control parameters of the flying robot include the action control parameters when the flying robot performs pitch, roll and yaw attitude actions.

5. The flight control method according to claim 1, characterized in that, The rotor power system (200) further includes four sets of vector propulsion rotors (220), and the four sets of vector propulsion rotors (220) are evenly distributed in the circumferential direction of the airbag main body (100); The vector propulsion rotor (220) includes a first motor (221) installed on the airbag main body (100) and electrically connected to the control system (300); the first motor (221) has a horizontally oriented propeller.

6. The flight control method according to claim 5, characterized in that, The rotor power system (200) includes four sets of horizontally coaxial rotors (210), and the four sets of horizontally coaxial rotors (210) are evenly distributed in the circumferential direction of the airbag body (100); The horizontally coaxial rotor (210) includes a servo (211) mounted on the airbag body (100) and a second motor (212) mounted on the servo (211) and electrically connected to the control system (300); the second motor (212) has a vertically oriented propeller.

7. The flight control method according to claim 6, characterized in that, The four sets of horizontally coaxial rotors (210) and the four sets of vector propulsion rotors (220) are arranged at intervals and alternately.

Images