Motion mode control method and system for robot with water-air dual-purpose tail fin

Abstract

The invention provides a motion modal control method and system for a robot with a water-air dual-purpose tail fin, and belongs to the technical field of water-air amphibious cross-medium robotics.Data are collected through a sensor system and a feedback system, modal selection and control are carried out through a modal selector based on reinforcement learning, and the motion modal of the robot with the water-air dual-purpose tail fin is controlled. Stable switching of an underwater mode, a water-to-air mode, an air mode and an air-to-water mode is achieved, the problem that multi-mode motion compatibility is insufficient is solved, meanwhile, by designing the water-to-air dual-purpose tail fin system and adopting V-shaped adjustment of a tail fin upper control surface and motion of a tail fin lower control surface, efficient and high-maneuverability posture adjustment is achieved, and the stability of the water-to-air dual-purpose tail fin system is improved. The problem that the overall performance of the robot system is unbalanced is solved, and the working efficiency and adaptability of the robot system in a complex environment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of water-air amphibious cross-medium robots, and in particular to a robot motion mode control method and system with a water-air dual-purpose tail fin. Background Art

[0002] The water-to-air cross-medium robot is a new type of integrated robot system that has the ability to navigate underwater, fly in the air, and freely transition between media. It breaks the inherent pattern of traditional robot systems and greatly improves the operating efficiency and adaptability of unmanned systems in complex water environments. It has broad application prospects in the fields of maritime reconnaissance, penetration, marine ecological environment monitoring, and marine resource exploration.

[0003] The water-air cross-media robot system in the prior art mainly includes propeller-propelled rotor cross-media UAV, fixed-wing cross-media UAV, bionic cross-media robot, flying fish-like cross-media robot and squid-like cross-media robot. The propeller-propelled rotor-spanning UAV uses multiple propeller differential drives to achieve propulsion and maneuvering movements. It has no structural compatibility design, and its underwater propulsion speed and efficiency are seriously insufficient. The fixed-wing cross-medium UAV uses propellers as underwater and aerial propulsion devices, and some wings adopt a variable-sweep and foldable design, which can be fully unfolded in the air to enhance lift. It has the advantages of high flight efficiency and high speed, but the underwater folding ratio is small and cannot be completely folded into the body, and it usually does not have a dedicated underwater maneuvering mechanism. Bionic cross-medium robots usually have a specific movement advantage, and it is difficult to take into account multiple movement modes such as underwater navigation and aerial flight, and the overall performance is insufficient. The flying fish-like cross-medium robot adopts underwater fish-like tail swinging propulsion and folding pectoral fin design. It has good underwater navigation capabilities but insufficient aerial flight capabilities. The squid-like cross-medium robot adopts jet propulsion to achieve rapid medium transition, etc., but lacks underwater and aerial navigation capabilities. In summary, the water-air cross-medium robot systems in the existing technologies have certain technical advantages in their respective fields, but they generally have problems such as lack of intelligence in multimodal motion control, insufficient system compatibility, and unbalanced overall performance. Summary of the invention

[0004] In order to solve the problems in the above-mentioned prior art, the present invention provides a method and system for controlling the motion mode of a robot with a dual-purpose tail fin for water and air. The invention collects data through a sensor system and a feedback system, and uses a mode selector based on reinforcement learning to select and control the mode, thereby realizing the smooth switching of underwater mode, water-to-air mode, air mode, and air-to-water mode, and improving the problems of lack of intelligence in multi-modal motion control and insufficient system compatibility. At the same time, by designing a dual-purpose tail fin system for water and air, the V-shaped adjustment of the rudder surface on the tail fin and the movement of the rudder surface under the tail fin are adopted to realize efficient and highly maneuverable posture adjustment, improve the problem of overall performance imbalance of the robot system, and improve the operating efficiency and adaptability of the robot system in complex environments. To achieve the above-mentioned purpose, the technical scheme is as follows:

[0005] In one aspect, the present invention provides a method for controlling the motion mode of a robot having a dual-purpose tail fin for water and air, the method comprising:

[0006] S1. Collect data through the sensor system to obtain the current state data set;

[0007] S2. According to the data set of the current state, a modality selector based on reinforcement learning is used to select a modality to obtain the current environment modality;

[0008] S3, according to the current environment mode, controlling the robot through the controller system to obtain the current form of the robot;

[0009] S4, obtaining the actual state of the robot through the robot's feedback system according to the current state of the robot;

[0010] S5. According to the actual state of the robot, the controller system is adjusted to obtain a stable form of the robot.

[0011] Optionally, the sensor system comprises:

[0012] Visual sensors, used to collect images and environmental features;

[0013] Inertial measurement sensor, used to collect acceleration, angular velocity and attitude data of the robot;

[0014] The depth sensor is used to collect the vertical position information of the robot, and the vertical position information includes the depth in water and the flying height in the air.

[0015] Optionally, in S1, data is collected through a sensor system to obtain a data set of the current state, including:

[0016] S11, according to the visual sensor, using a convolutional neural network to extract features to obtain visual features of the environment;

[0017] S12, obtaining the acceleration, angular velocity and posture data of the robot through collection according to the inertial measurement sensor;

[0018] S13, obtaining the vertical position information of the robot through collection according to the depth sensor;

[0019] S14. Perform matrix processing according to the visual features of the environment, the acceleration of the robot, the angular velocity of the robot, the posture data of the robot and the vertical position information of the robot to obtain a data set of the current state.

[0020] Optionally, the training method of the reinforcement learning-based modality selector includes:

[0021] S21, by setting the initialization reinforcement learning environment, an initialized Q network is obtained;

[0022] S22, inputting the training data set into the Q network, adopting a greedy strategy, and obtaining the robot mode under the training data set state;

[0023] S23, according to the robot mode in the state of the training data set, controlling the robot through the controller system to obtain the current training form and action of the robot;

[0024] S24, obtaining the actual training state of the robot through the feedback system of the robot according to the current training state of the robot;

[0025] S25. According to the actual training state of the robot and the current training state of the robot, the controller system is adjusted and compared and analyzed, and the correctness reward function of the modality and the adjusted robot training state are obtained through formula (1).

[0026] r t =R correct -R incorrect -R switch_cost +R efficiency (1)

[0027] In the formula, r t is the correctness reward function of the modality, R correct is the reward for choosing the correct mode, R incorrect is the penalty for choosing the wrong mode, R switch_cost is the penalty for frequent mode switching, R efficiency incentives for energy efficiency;

[0028] S26, repeating steps S22 to S25, storing the current training form of the robot, the action of the robot, the correctness reward function of the mode, and the adjusted training state of the robot into the experience replay pool to obtain a training experience replay pool;

[0029] S27, input the training experience playback pool into the Q network, update the parameters of the Q network through formula (2), and obtain the final parameters of the Q network.

[0030]

[0031] In the formula, Q(s t ,a t ) is the action quality function of the current training form, s t is the current training state of the robot, a t is the action of the robot, a′ is the action of any robot, r t is the correctness reward function of the modality, s t+1 is the adjusted robot training state, γ is the discount factor, and α is the learning rate;

[0032] S28. Input the parameters of the final Q network into the Q network to obtain a mode selector based on reinforcement learning.

[0033] Optionally, the greedy strategy includes:

[0034] Strategy 1: Select a random robot action instruction with probability ∈;

[0035] Strategy 2: Select the robot's action instruction with the maximum action quality function of the current training form with probability 1-∈.

[0036] Optionally, in S3, according to the current environmental modality, the robot is controlled by a controller system to obtain the current form of the robot, including:

[0037] S31, according to the current environmental mode, controlling the fixed-wing propulsion system through the controller system to obtain the form of the fixed-wing propulsion system;

[0038] S32, according to the current environmental mode, controlling the variable structure membrane wing system through the controller system to obtain the shape of the variable structure membrane wing system;

[0039] S33, according to the current environmental mode, controlling the pectoral fin system through the controller system to obtain the shape of the pectoral fin system;

[0040] S34, according to the current environment mode, controlling the body cabin system through the controller system to obtain the form of the body cabin system;

[0041] S35, according to the current environmental mode, controlling the multi-joint tail cabin system through the controller system to obtain the shape of the multi-joint tail cabin system;

[0042] S36, according to the current environmental mode, controlling the dual-purpose tail fin system through the controller system to obtain the shape of the dual-purpose tail fin system;

[0043] S37. Obtain the current form of the robot based on the form of the fixed-wing propulsion system, the form of the variable structure membrane wing system, the form of the pectoral fin system, the form of the body cabin system, the form of the multi-joint tail cabin system and the form of the dual-purpose water-air tail fin system.

[0044] Optionally, the dual-purpose tail fin system for water and air includes:

[0045] The rudder under the tail fin is used to provide a control surface below to adjust the direction of the flow in water or air;

[0046] The rudder surface on the tail fin is used to provide an upper control surface to adjust the direction of the flow in water or air;

[0047] The rudder surface under the tail fin drives the servo, which is used to adjust the angle change of the rudder surface under the tail fin;

[0048] Rudder arm, used to transmit driving force;

[0049] A bracket, used for fixing the lower rudder surface of the tail fin and the upper rudder surface of the tail fin;

[0050] A planetary gear is used to adjust the angle of the rudder surface on the tail fin;

[0051] A planetary gear fixing frame, used for fixing the planetary gear and connecting the multi-joint tail cabin system;

[0052] The lower tail fin rudder surface drives the servo motor and is fixedly connected to the rudder arm, and the rudder arm is fixedly connected to the lower tail fin rudder surface. The lower tail fin rudder surface drives the servo motor to rotate, so as to adjust the angle between the lower tail fin rudder surface and the upper tail fin rudder surface.

[0053] The planetary gear is fixedly connected to the upper control surface of the tail fin, and the rotation of the planetary gear realizes the V-shaped adjustment movement of the upper control surface of the tail fin.

[0054] Optionally, the planetary gear comprises a driving gear, a transmission gear 1, a transmission gear 2 and a transmission gear 3. The driving gear rotates under the drive of the transmission shaft, and the transmission gear 1, the transmission gear 2 and the transmission gear 3 are respectively meshed with the driving gear.

[0055] Optionally, the stable forms of the robot include underwater mode, water-into-air mode, aerial mode and air-into-water mode.

[0056] On the other hand, the present invention provides a robot motion mode control system with a dual-purpose tail fin for water and air, the system is applied to a robot motion mode control method with a dual-purpose tail fin for water and air, the system comprising:

[0057] The data acquisition module is used to collect data through the sensor system to obtain a data set of the current state;

[0058] A mode selection module is used to select a mode based on the current state data set and obtain the current environment mode by using a mode selector based on reinforcement learning;

[0059] A first acquisition module is used to control the robot through a controller system according to the current environmental modality to obtain the current form of the robot;

[0060] A second acquisition module is used to obtain the actual state of the robot through the feedback system of the robot according to the current state of the robot;

[0061] The third acquisition module is used to obtain a stable form of the robot through adjustment of the controller system according to the actual state of the robot.

[0062] Compared with the prior art, the technical solution of the present invention has at least the following beneficial effects:

[0063] On the one hand, the above scheme collects data through the sensor system and feedback system, and uses the mode selector based on reinforcement learning to select and control the mode, thereby realizing the smooth and intelligent switching of underwater mode, water-to-air mode, aerial mode and air-to-water mode, and improving the problems of lack of intelligence in multi-modal motion control and insufficient system compatibility. On the other hand, by designing a water-air dual-purpose tail fin system, the V-shaped adjustment of the rudder surface on the tail fin and the movement of the rudder surface under the tail fin are adopted to realize efficient and highly maneuverable attitude adjustment, improve the problem of overall performance imbalance of the robot system, and improve the operation efficiency and adaptability of the robot system in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0065] Figure 1 is a flow chart of an embodiment of a method for controlling motion modes of a robot with a dual-purpose tail fin for water and air of the present invention;

[0066] Figure 2 It is a flow chart of obtaining a data set of a current state in an embodiment of a method for controlling a robot motion mode with a dual-purpose tail fin for water and air of the present invention;

[0067] Figure 3It is a flow chart of a training method of a mode selector based on reinforcement learning in an embodiment of a method for controlling a robot motion mode with a dual-purpose tail fin for water and air of the present invention;

[0068] Figure 4 It is a flow chart of obtaining the current state of the robot in an embodiment of the method for controlling the motion mode of the robot with a dual-purpose tail fin for water and air of the present invention;

[0069] Figure 5 1 is a schematic structural diagram of a robot with a dual-purpose tail fin for water and air in an embodiment of a method for controlling a robot motion mode for water and air of the present invention;

[0070] Figure 6 1 is a schematic structural diagram of a dual-purpose water-air tail fin system in an embodiment of a method for controlling a robot motion mode having a dual-purpose water-air tail fin according to the present invention;

[0071] Figure 7 1 is a schematic structural diagram of a planetary gear of a water-air dual-purpose tail fin system in an embodiment of a method for controlling a motion mode of a robot with a water-air dual-purpose tail fin of the present invention;

[0072] Figure 8 It is a system block diagram of an embodiment of a robot motion modal control system with a water-air dual-purpose tail fin of the present invention.

[0073] Explanation of the numbers in the figure: fixed-wing propulsion system 1, variable structure membrane wing system 2, pectoral fin system 3, body cabin system 4, multi-joint tail cabin system 5, water-air dual-purpose tail fin system 6, tail fin lower rudder surface 61, tail fin upper rudder surface 62, tail fin lower rudder surface driving servo 63, rudder arm 64, bracket 65, planetary gear 66, planetary gear fixing frame 67, driving gear 661, transmission gear 1 662, transmission gear 2 663, transmission gear 3 664. DETAILED DESCRIPTION

[0074] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0075] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0076] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0077] like Figure 1The flowchart of the embodiment of the motion mode control method of the robot with a dual-purpose tail fin of the present invention is shown in FIG. The present invention provides a motion mode control method of the robot with a dual-purpose tail fin of the present invention, which is implemented by a motion mode control system of the robot with a dual-purpose tail fin of the present invention, and the method comprises:

[0078] S1. Collect data through the sensor system to obtain the current state data set;

[0079] Specifically, the sensor system comprises:

[0080] Visual sensors, used to collect images and environmental features;

[0081] Inertial measurement sensor, used to collect acceleration, angular velocity and attitude data of the robot;

[0082] The depth sensor is used to collect the vertical position information of the robot, and the vertical position information includes the depth in water and the flying height in the air.

[0083] like Figure 2 The flowchart of obtaining a data set of the current state in the embodiment of the method for controlling the motion mode of a robot with a dual-purpose tail fin of water and air of the present invention is shown, wherein in S1, data is collected by a sensor system to obtain a data set of the current state, including:

[0084] S11, according to the visual sensor, using a convolutional neural network to extract features to obtain visual features of the environment;

[0085] S12, obtaining the acceleration, angular velocity and posture data of the robot through collection according to the inertial measurement sensor;

[0086] S13, obtaining the vertical position information of the robot through collection according to the depth sensor;

[0087] S14. Perform matrix processing according to the visual features of the environment, the acceleration of the robot, the angular velocity of the robot, the posture data of the robot and the vertical position information of the robot to obtain a data set of the current state.

[0088] S2. According to the data set of the current state, a modality selector based on reinforcement learning is used to select a modality to obtain the current environment modality;

[0089] Specifically, Figure 3 The flowchart of the training method of the mode selector based on reinforcement learning in the embodiment of the motion mode control method of the robot with a dual-purpose tail fin of water and air of the present invention is shown, and the training method of the mode selector based on reinforcement learning includes:

[0090] S21, by setting the initialization reinforcement learning environment, an initialized Q network is obtained;

[0091] S22, inputting the training data set into the Q network, adopting a greedy strategy, and obtaining the robot mode under the training data set state;

[0092] S23, according to the robot mode in the state of the training data set, controlling the robot through the controller system to obtain the current training form and action of the robot;

[0093] S24, obtaining the actual training state of the robot through the feedback system of the robot according to the current training state of the robot;

[0094] S25. According to the actual training state of the robot and the current training state of the robot, the controller system is adjusted and compared and analyzed, and the correctness reward function of the modality and the adjusted robot training state are obtained through formula (1).

[0095] r t =R correct -R incorrect -R switch_cost +R efficiency (1)

[0096] In the formula, r t is the correctness reward function of the modality, R correct is the reward for choosing the correct mode, R incorrect is the penalty for choosing the wrong mode, R switch_cost is the penalty for frequent mode switching, R efficiency incentives for energy efficiency;

[0097] S26, repeating steps S22 to S25, storing the current training form of the robot, the action of the robot, the correctness reward function of the mode, and the adjusted training state of the robot into the experience replay pool to obtain a training experience replay pool;

[0098] S27, input the training experience playback pool into the Q network, update the parameters of the Q network through formula (2), and obtain the final parameters of the Q network.

[0099]

[0100] In the formula, Q(s t ,a t ) is the action quality function of the current training form, s t is the current training state of the robot, a t is the action of the robot, a′ is the action of any robot, r t is the correctness reward function of the modality, st+1 is the adjusted robot training state, γ is the discount factor, and α is the learning rate;

[0101] S28. Input the parameters of the final Q network into the Q network to obtain a mode selector based on reinforcement learning.

[0102] The greedy strategy includes:

[0103] Strategy 1: Select a random robot action instruction with probability ∈;

[0104] Strategy 2: Select the action instruction of the robot with the maximum action quality function of the current training form with probability 1-ε.

[0105] S3, according to the current environment mode, controlling the robot through the controller system to obtain the current form of the robot;

[0106] Specifically, Figure 4 The flowchart and the example of the method for controlling the motion mode of a robot with a dual-purpose tail fin of the present invention are shown in FIG. Figure 5 The schematic diagram of the structure of the robot with a dual-purpose tail fin for water and air in the embodiment of the method for controlling the motion mode of the robot with a dual-purpose tail fin for water and air of the present invention is shown, and in S3, the robot is controlled by the controller system according to the current environmental mode to obtain the current form of the robot, including:

[0107] S31, according to the current environmental mode, controlling the fixed-wing propulsion system 1 through the controller system to obtain the form of the fixed-wing propulsion system;

[0108] S32, according to the current environmental mode, controlling the variable structure membrane wing system 2 through the controller system to obtain the shape of the variable structure membrane wing system;

[0109] S33, according to the current environmental mode, controlling the pectoral fin system 3 through the controller system to obtain the shape of the pectoral fin system;

[0110] S34, according to the current environment mode, the body cabin system 4 is controlled by the controller system to obtain the form of the body cabin system;

[0111] S35, according to the current environmental mode, the multi-joint tail cabin system 5 is controlled by the controller system to obtain the shape of the multi-joint tail cabin system;

[0112] S36, according to the current environmental mode, controlling the water-air dual-purpose tail fin system 6 through the controller system to obtain the shape of the water-air dual-purpose tail fin system;

[0113] S37. Obtain the current form of the robot based on the form of the fixed-wing propulsion system, the form of the variable structure membrane wing system, the form of the pectoral fin system, the form of the body cabin system, the form of the multi-joint tail cabin system and the form of the dual-purpose water-air tail fin system.

[0114] Furthermore, the fixed-wing propulsion system 1 is used as a power source during aerial flight, the variable structure membrane wing system 2 is used for flight stability and generation of flight lift during aerial flight, the pectoral fin system 3 is used for pitch, yaw and other controls during underwater swimming, the body cabin system is used for control, perception, communication and contraction of the variable structure membrane wing system 2 in the underwater swimming state, the multi-joint tail cabin 5 is used for a multi-joint fish-like tail swinging propulsion method, and the water-air dual-purpose tail fin system 6 is used as the main mechanism for generating propulsion force for the fish-like tail fin in the underwater mode and as the V-shaped tail fin structure of the fixed-wing robot in the aerial mode to achieve adjustment of pitch, yaw and other postures.

[0115] like Figure 6 The structure diagram of the dual-purpose tail fin system of the robot motion mode control method with dual-purpose tail fin of the present invention is shown, and the dual-purpose tail fin system 6 includes:

[0116] The lower rudder surface 61 of the tail fin is used to provide a control surface below to adjust the direction of the fluid in the water or in the air;

[0117] The rudder surface 62 on the tail fin is used to provide an upper control surface to adjust the direction of the fluid in the water or in the air;

[0118] The lower tail fin control surface drives the steering gear 63, which is used to adjust the angle change of the lower tail fin control surface 61;

[0119] A rudder arm 64, used to transmit driving force;

[0120] A bracket 65, used for fixing the lower rudder surface 61 of the tail fin and the upper rudder surface 62 of the tail fin;

[0121] The planetary gear 66 is used to adjust the angle of the control surface 62 on the tail fin;

[0122] A planetary gear fixing frame 67, used for fixing the planetary gear 66 and connecting the multi-joint tail cabin system 5;

[0123] The tail fin lower rudder surface driving steering engine 63 is fixedly connected to the rudder arm 64, and the rudder arm 64 is fixedly connected to the tail fin lower rudder surface 61. The tail fin lower rudder surface driving steering engine 63 performs rotational motion to adjust the angle between the tail fin lower rudder surface 61 and the tail fin upper rudder surface 62;

[0124] When the output shaft of the tail fin lower rudder driving servo 63 rotates, the angle between the tail fin lower rudder 61 and the tail fin upper rudder 62 can be adjusted. When the tail fin lower rudder 61 is in a parallel position, the output shaft of the tail fin lower rudder driving servo 63 keeps the same position, which is an underwater mode. That is, in the underwater mode, the tail fin lower rudder 61 and the tail fin upper rudder 62 remain stationary, only serving as underwater propulsion sources, and underwater propulsion is achieved through the swing of the multi-joint tail cabin system 5; in the air mode, the tail fin upper rudder 62 forms a V-shaped tail angle, that is, the planetary gear 66 makes the two tail fin upper rudders 62 rotate in opposite directions.

[0125] The lower tail fin rudder surface 61 drives the rudder arm 64 to rotate through the lower tail fin rudder surface, thereby pulling the bracket 65 to achieve deflection, and achieve pitch and yaw control when flying in the air.

[0126] The planetary gear 66 is fixedly connected to the upper control surface 62 of the tail fin, and the rotation of the planetary gear 66 realizes the V-shaped adjustment movement of the upper control surface 62 of the tail fin.

[0127] Furthermore, if Figure 7 The schematic diagram of the structure of the planetary gear of the dual-purpose tail fin system in the embodiment of the motion mode control method of the robot with dual-purpose tail fin of the present invention is shown, the planetary gear 66 includes a driving gear 661, a transmission gear 1 662, a transmission gear 2 663 and a transmission gear 3 664, the driving gear 661 rotates under the drive of the transmission shaft, and the transmission gear 1 662, the transmission gear 2 663 and the transmission gear 3 664 are respectively meshed with the driving gear 661.

[0128] S4, obtaining the actual state of the robot through the robot's feedback system according to the current state of the robot;

[0129] S5. According to the actual state of the robot, the controller system is adjusted to obtain a stable form of the robot.

[0130] Specifically, the stable forms of the robot include underwater mode, water-into-air mode, aerial mode and air-into-water mode.

[0131] Furthermore, in the aerial mode, the V-shaped tail angle of the robot is usually fixed, and the aerial attitude adjustment is achieved only by adjusting the angle between the lower rudder surface 61 of the tail fin and the upper rudder surface 62 of the tail fin, that is, adjusting the lower rudder surface 61 of the tail fin to change the received aerodynamic force; in the underwater mode, the V-shaped tail angle of the robot is usually 180 degrees, and the angle between the lower rudder surface 61 of the tail fin and the upper rudder surface 62 of the tail fin is also 180 degrees, that is, the water-air dual-purpose tail fin system 6 constitutes an integral crescent-shaped tail fin, similar to the tail fin of a tuna.

[0132] like Figure 8The system block diagram of the embodiment of the robot motion mode control system with a dual-purpose tail fin of the present invention is shown in the figure. The present invention provides a robot motion mode control system with a dual-purpose tail fin of the present invention. The system is applied to a robot motion mode control method with a dual-purpose tail fin of the present invention. The system includes a data acquisition module, a mode selection module, a first acquisition module, a second acquisition module and a third acquisition module. Specifically,

[0133] The data acquisition module is used to collect data through the sensor system to obtain a data set of the current state;

[0134] A mode selection module is used to select a mode based on the current state data set and obtain the current environment mode by using a mode selector based on reinforcement learning;

[0135] A first acquisition module is used to control the robot through a controller system according to the current environmental modality to obtain the current form of the robot;

[0136] A second acquisition module is used to obtain the actual state of the robot through the feedback system of the robot according to the current state of the robot;

[0137] The third acquisition module is used to obtain a stable form of the robot through adjustment of the controller system according to the actual state of the robot.

[0138] The present invention provides a method and system for controlling the motion modes of a robot with a dual-purpose tail fin for water and air. The invention collects data through a sensor system and a feedback system, and uses a mode selector based on reinforcement learning to select and control the mode, thereby realizing the smooth switching of underwater mode, water-to-air mode, air mode and air-to-water mode, and improving the problems of lack of intelligence in multi-modal motion control and insufficient system compatibility. At the same time, by designing a dual-purpose tail fin system for water and air, the V-shaped adjustment of the rudder surface on the tail fin and the movement of the rudder surface under the tail fin are adopted to realize efficient and highly maneuverable posture adjustment, thereby improving the problem of overall performance imbalance of the robot system and improving the operating efficiency and adaptability of the robot system in complex environments.

[0139] It is to be understood that the present invention is described by the above embodiments and should not be construed as limiting the embodiments of the present invention and the scope of the present invention. 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 method for controlling the motion mode of a robot with a dual-purpose tail fin for water and air, characterized in that: The method comprises: S1. Collect data through the sensor system to obtain the current state data set; S2. According to the data set of the current state, a modality selector based on reinforcement learning is used to select a modality to obtain a current environment modality; S3, according to the current environment mode, controlling the robot through the controller system to obtain the current form of the robot; S4. Obtaining the actual state of the robot through the feedback system of the robot according to the current state of the robot; S5. According to the actual state of the robot, the controller system is adjusted to obtain a stable form of the robot.

2. The motion mode control method of a robot with a dual-purpose tail fin for water and air according to claim 1, characterized in that: The sensor system comprises: Visual sensors, used to collect images and environmental features; An inertial measurement sensor, used to collect acceleration, angular velocity and posture data of the robot; The depth sensor is used to collect the vertical position information of the robot, and the vertical position information includes the depth in water and the flying height in the air.

3. The motion mode control method of a robot with a dual-purpose tail fin for water and air according to claim 2, characterized in that: In S1, data is collected through the sensor system to obtain a data set of the current state, including: S11, extracting features using a convolutional neural network according to the visual sensor to obtain visual features of the environment; S12, obtaining the acceleration, angular velocity and posture data of the robot by collecting data according to the inertial measurement sensor; S13, obtaining the vertical position information of the robot through collection according to the depth sensor; S14. Perform matrix processing according to the visual features of the environment, the acceleration of the robot, the angular velocity of the robot, the posture data of the robot and the position information of the robot in the vertical direction to obtain a data set of the current state.

4. The motion mode control method of a robot with a dual-purpose tail fin for water and air according to claim 1, characterized in that: The training method of the mode selector based on reinforcement learning includes: S21, by setting the initialization reinforcement learning environment, an initialized Q network is obtained; S22, inputting the training data set into the Q network, adopting a greedy strategy, and obtaining the robot mode under the training data set state; S23, controlling the robot through the controller system according to the robot mode in the state of the training data set to obtain the current training form and action of the robot; S24, obtaining the actual training state of the robot through the feedback system of the robot according to the current training state of the robot; S25, according to the actual training state of the robot and the current training state of the robot, the controller system is adjusted and compared and analyzed, and the correctness reward function of the modality and the adjusted robot training state are obtained through formula (1), r t =R correct -R incorrect -R switch_cost +R efficiency (1) In the formula, r t is the correctness reward function of the modality, R correct is the reward for choosing the correct mode, R incorrect is the penalty for choosing the wrong mode, R switch_cost is the penalty for frequent mode switching, R efficiency incentives for energy efficiency; S26, repeating steps S22 to S25, storing the current training form of the robot, the action of the robot, the correctness reward function of the modality, and the adjusted robot training state into the experience replay pool to obtain a training experience replay pool; S27, input the training experience playback pool into the Q network, and update the parameters of the Q network by formula (2) to obtain the final parameters of the Q network. In the formula, Q(s t ,a t ) is the action quality function of the current training form, s t is the current training state of the robot, a t is the action of the robot, a′ is the action of any robot, r t is the correctness reward function of the modality, s t+1 is the adjusted robot training state, γ is the discount factor, and α is the learning rate; S28. Input the parameters of the final Q network into the Q network to obtain a mode selector based on reinforcement learning.

5. The motion mode control method of a robot with a dual-purpose tail fin for water and air according to claim 4, characterized in that: The greedy strategy includes: Strategy 1: Select a random robot action instruction with probability ε; Strategy 2: Select the action instruction of the robot with the maximum action quality function of the current training form with probability 1-ε.

6. The method for controlling the motion mode of a robot with a dual-purpose tail fin for water and air according to claim 1, characterized in that: In S3, according to the current environment modality, the robot is controlled by the controller system to obtain the current form of the robot, including: S31, controlling the fixed-wing propulsion system through a controller system according to the current environmental mode to obtain a form of the fixed-wing propulsion system; S32, according to the current environmental mode, controlling the variable structure membrane wing system through the controller system to obtain the shape of the variable structure membrane wing system; S33, controlling the pectoral fin system through the controller system according to the current environmental modality to obtain the shape of the pectoral fin system; S34, controlling the body cabin system through the controller system according to the current environmental modality to obtain the form of the body cabin system; S35, according to the current environmental modality, controlling the multi-joint tail cabin system through the controller system to obtain the shape of the multi-joint tail cabin system; S36, according to the current environmental mode, controlling the dual-purpose tail fin system through the controller system to obtain the shape of the dual-purpose tail fin system; S37. Obtain the current form of the robot based on the form of the fixed-wing propulsion system, the form of the variable structure membrane wing system, the form of the pectoral fin system, the form of the body cabin system, the form of the multi-joint tail cabin system and the form of the dual-purpose water-air tail fin system.

7. The motion mode control method of a robot with a dual-purpose tail fin for water and air according to claim 6, characterized in that: The dual-purpose tail fin system for water and air includes: The rudder under the tail fin is used to provide a control surface below to adjust the direction of the flow in water or air; The rudder surface on the tail fin is used to provide an upper control surface to adjust the direction of the flow in water or air; The rudder surface under the tail fin drives the servo, which is used to adjust the angle change of the rudder surface under the tail fin; Rudder arm, used to transmit driving force; A bracket, used for fixing the lower rudder surface of the tail fin and the upper rudder surface of the tail fin; A planetary gear, used to adjust the angle of the control surface on the tail fin; A planetary gear fixing frame, used for fixing the planetary gear and connecting the multi-joint tail cabin system; The lower tail fin rudder surface driving steering gear is fixedly connected to the rudder arm, and the rudder arm is fixedly connected to the lower tail fin rudder surface, and the lower tail fin rudder surface driving steering gear performs rotational motion to adjust the angle between the lower tail fin rudder surface and the upper tail fin rudder surface; The planetary gear is fixedly connected to the upper control surface of the tail fin, and the rotation of the planetary gear realizes the V-shaped adjustment movement of the upper control surface of the tail fin.

8. The method for controlling the motion mode of a robot with a dual-purpose tail fin for water and air according to claim 7, characterized in that: The planetary gear comprises a driving gear, a transmission gear 1, a transmission gear 2 and a transmission gear 3. The driving gear rotates under the drive of the transmission shaft, and the transmission gear 1, the transmission gear 2 and the transmission gear 3 are respectively meshed with the driving gear.

9. The method for controlling the motion mode of a robot with a dual-purpose tail fin for water and air according to claim 1, characterized in that: The stable forms of the robot include underwater mode, water-into-air mode, air mode and air-into-water mode.

10. A robot motion mode control system with a dual-purpose tail fin for water and air, used to implement the robot motion mode control method with a dual-purpose tail fin for water and air as claimed in any one of claims 1 to 9, characterized in that: The system comprises: The data acquisition module is used to collect data through the sensor system to obtain a data set of the current state; A mode selection module, configured to select a mode based on the data set of the current state by using a mode selector based on reinforcement learning to obtain a current environment mode; A first acquisition module is used to control the robot through a controller system according to the current environmental modality to obtain the current form of the robot; A second acquisition module is used to obtain the actual state of the robot through the feedback system of the robot according to the current state of the robot; The third acquisition module is used to obtain a stable form of the robot through adjustment by the controller system according to the actual state of the robot.

Images