A continuous detonation robot imitating a fire-brigade worm and a control method thereof

Abstract

This invention discloses a pyrosome-inspired continuous blast robot and its control method. The robot includes a shell with a camera mounted on it. A jet nozzle is located at the rear end of the shell, and a one-way water filtration device is located at the front end. This one-way water filtration device allows water to enter the cavity of the shell but prevents water from leaving the cavity. An elastic diaphragm and a chemical energy release reaction device are mounted on the side wall of the cavity. The chemical energy release reaction device causes the elastic diaphragm to expand through a chemical energy release reaction, thereby expelling water from the cavity through the jet nozzle, thus enabling the continuous blast robot to move. This invention utilizes a chemical energy release reaction to drive the rapid expansion of the elastic diaphragm, achieving continuous water jet propulsion. This results in a significant increase in thrust within a short time, significantly improving the robot's start-up speed and motion response capability.

Description

Technical Field

[0001] This invention relates to the field of robotics, specifically to a continuous blast robot inspired by a pyrosome and its control method. Background Technology

[0002] With the continuous development of technology, the demand for robots in underwater surveying, monitoring, and rescue is increasing. The complex underwater environment places extremely high demands on the structure, materials, circuitry, and sensors of robots. Traditional hardware robots typically use rigid materials and electric motors as their main components, which makes them subject to significant resistance and energy loss during underwater movement, and also susceptible to collisions and corrosion. Furthermore, traditional hardware robots require batteries or connecting cables for power supply, which limits their underwater working time and range. Therefore, developing a small, lightweight, flexible, adaptable, continuously driven, and highly maneuverable underwater robot is particularly important.

[0003] Chemical reaction-driven technology refers to the release of energy through chemical reactions, converting it into mechanical energy or other forms of energy. In underwater environments, robots often find it difficult to use traditional energy sources such as batteries or fuel cells due to their large size and weight. Chemical reaction energy, on the other hand, has the advantages of high energy density, small size, and light weight, making it better suited to the needs of underwater environments.

[0004] Bionic soft robots are robots that mimic the structure and movement of living organisms. Using flexible materials as their main components, they possess characteristics such as flexibility, deformability, adaptability, and light weight. This allows them to operate with low drag and energy loss underwater and to adapt to complex and varied underwater environments. Bionic soft robots can also achieve various morphological changes and functional switching to meet the needs of different tasks. Therefore, bionic soft robots have broad application prospects in underwater surveying, monitoring, and rescue.

[0005] Unlike traditional robot motor drives, the actuation of soft robots primarily depends on the deformation and elasticity of their materials. Soft robot materials possess deformable, compressible, and bendable properties; common materials include dielectric elastomers (DE), ionomer-metal composites (IPMC), shape memory alloys (SMA), and shape memory polymers (SMP). Soft robot actuation methods can be categorized into external and internal actuation. External actuation refers to using external forces or fields to manipulate the soft robot's movement, such as magnetic fields, electric fields, and light fields. Internal actuation utilizes internal energy sources within the soft robot to generate motion, such as air pressure or hydraulic pressure. However, traditional actuation methods cannot generate significant thrust in a short time, affecting the startup speed and movement of underwater soft robots.

[0006] Traditional underwater image recognition often adopts a "single image processing" mode, which is extremely inefficient for massive amounts of data (especially since underwater robots have limited computing power); underwater image batch processing algorithms often lack a collaborative mechanism of "targeted clustering + representative recognition", which easily leads to repeated calculation of similar images; at the same time, they do not take into account the characteristics of the underwater environment (such as image degradation) to design the linkage logic between clustering and recognition. Summary of the Invention

[0007] The purpose of this invention is to provide a continuous blast robot that mimics a pyrosome and its control method, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A continuous blasting robot mimicking a pyrosome includes a shell with a camera mounted on it. A jet nozzle is located at the rear end of the shell, and a one-way water filtration device is located at the front end. This one-way water filtration device allows water to enter the cavity of the shell but prevents water from leaving the cavity. An elastic diaphragm and a chemical energy release reaction device are mounted on the side wall of the cavity. The chemical energy release reaction device causes the elastic diaphragm to expand through a chemical energy release reaction, thereby expelling water from the cavity through the jet nozzle, thus enabling the continuous blasting robot to move.

[0010] Furthermore, the one-way water filtration device includes a filter plate with a plurality of water holes and a one-way valve on each water hole.

[0011] Furthermore, the outer shell and its cavity have a structure that gradually thickens from front to back.

[0012] Furthermore, the elastic diaphragm and the inner wall of the outer shell form a driving gas chamber. The chemical energy release reaction device includes a gas charging module, a control module, an air inlet valve, and a spark plug. The air inlet valve and the spark plug are disposed on the driving gas chamber. The gas charging module is connected to the air inlet valve to supply chemical energy release reaction gas to the driving gas chamber. The control module is connected to the spark plug to control the spark plug discharge.

[0013] When the inflation module supplies air to the drive air chamber and the control module controls the spark plug to discharge, the chemical energy release reaction gas in the drive air chamber undergoes a chemical energy release reaction, causing the elastic diaphragm to bulge. The elastic diaphragm pushes the water in the cavity to be ejected outward from the injection port, thereby generating the driving force that propels the continuous detonation robot forward.

[0014] The present invention also provides a control method for the continuous detonation robot mimicking the pyrosome as described above, comprising:

[0015] Step 1: Collect images and videos using a camera to achieve unified input of underwater images;

[0016] Step 2: Preprocess the uniformly input underwater images using the pyrosome swarm image processing algorithm;

[0017] Step 3: Use the YOLOX algorithm to recognize objects in front of the preprocessed image;

[0018] Step 4: Control the robot's movement based on the recognition results; when an obstacle is detected ahead, control the expansion of the elastic diaphragm of the continuous detonation robot to achieve rapid obstacle avoidance; when a living organism that needs to be followed is detected ahead, increase the working frequency of the chemical energy release reaction device to achieve rapid following.

[0019] Furthermore, the fire swarm image processing algorithm in step 2 includes the following steps:

[0020] S100 optimizes the underwater environment in underwater images through an underwater image preprocessing module, including noise reduction and contrast enhancement;

[0021] The S200 uses the lightweight YOLOX-tiny network for initial rapid identification.

[0022] S300 extracts key features from the initial recognition results through the feature extraction module;

[0023] S400, the clustering and classification module, based on feature vector similarity, groups similar images into one class, simulating the "information aggregation" characteristics of pyrosome group collaboration; it performs underwater image feature similarity calculation, assuming the feature vector of image i is... , Let n represent the component, n be the feature dimension, and the feature vector of image j be... Similarity for:

[0024] ,

[0025] in, Let D(i,j) be the vector dot product, and D(i,j) be the underwater image degradation distance. As an environmental adaptation factor, and The magnitude of the eigenvector;

[0026] S500 represents the image selection module using entropy scoring to select 1-3 images with the "most informational content" from four categories: marine life, seabed environment, seabed resources, and underwater structures; and represents the image scoring formula. as follows:

[0027] ,

[0028] in, Image entropy, Representing an image Medium pixel value The probability, For clarity, The image is in coordinates gradient magnitude at that point It is the image size. For the target percentage, For weights.

[0029] Furthermore, step 3 uses the YOLOX algorithm to identify the preprocessed image and predict the bounding box coordinates. The calculation formula is:

[0030] ,

[0031] ,

[0032] ,

[0033] ,

[0034] Among them, the coordinates of the bounding box Including center point coordinates Width w, Height h The x and y coordinates of the current grid center are given. For the current width, At the current altitude, coordinates of the center point Predict the offset. Predict the offset for width w. Predict the offset for height h. For the sigmoid function, and is the scaling factor for width and height, and e is the natural constant;

[0035] loss function for:

[0036] ,

[0037] Where IoU is the intersection-union ratio. Center of the prediction box Center of the real frame The square of the Euclidean distance, The length of the diagonal of the smallest rectangle enclosing the two frames. , It is a balancing factor.

[0038] Compared with the prior art, the beneficial effects of the present invention are:

[0039] (1) High-efficiency drive and continuous motion capability: The elastic diaphragm is rapidly expanded by chemical energy release reaction (such as the detonation of a mixture of propane and oxygen) to achieve continuous jet propulsion of water flow. It can generate a large thrust in a short time, which significantly improves the robot's start-up speed and motion response capability. It overcomes the problem of insufficient thrust in traditional soft robot drive methods and the problem of limited drive times in the original underwater robot driven by chemical energy release reaction.

[0040] (2) Bionic structural design: The robot shell imitates the streamlined structure of the pyrosome with a large diameter water inlet at the front and a small diameter water outlet at the rear. Combined with a one-way water filtration device, it effectively reduces water flow resistance and improves movement efficiency. The use of flexible materials (such as elastic diaphragms) enhances the robot's adaptability and collision resistance, and reduces the risk of corrosion.

[0041] (3) Intelligent image recognition and control: Based on the pyrosome swarm image processing algorithm and the YOLOX target detection algorithm, efficient processing and target recognition of underwater images are realized. The algorithm simulates the "information aggregation" characteristics of pyrosome swarm cooperation, and reduces computational redundancy and improves recognition efficiency through clustering and representative image filtering mechanisms, which is especially suitable for underwater robot platforms with limited computing power.

[0042] (4) Environmental adaptability: The image processing algorithm incorporates underwater image degradation features (such as noise and color decay), and dynamically adjusts the clustering threshold through environmental adaptation factors, thereby improving the accuracy and robustness of image recognition and overcoming the problem of high misclassification rate of traditional underwater image processing algorithms in complex environments. Attached Figure Description

[0043] Figure 1 A schematic diagram of a continuous detonation underwater robot that mimics the structure of a pyrosome.

[0044] Figure 2 A flowchart illustrating the continuous detonation drive module and jet drive method of a continuous detonation underwater robot that mimics a pyrosome.

[0045] Figure 3 Flowcharts for underwater image recognition algorithms and robot motion control algorithms;

[0046] Figure 4 Diagram of underwater image recognition and control algorithm for pyrosome swarms

[0047] In the diagram: 1-outer shell; 2-one-way water filtration device; 3-one-way valve; 4-camera; 5-control module; 6-inlet valve; 7-spark plug; 8-exhaust valve; 9-elastic diaphragm. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Please see Figures 1-3 A continuous blasting robot, modeled after a pyrosome, includes a shell 1 with a miniature camera 4 mounted on it. A jet nozzle is located at the rear end of the shell 1, and a one-way water filtration device 2 is located at the front end. The one-way water filtration device 2 allows water to enter the cavity of the shell 1 but prevents water from leaving the cavity. Elastic diaphragms 9 and chemical energy release reaction devices are mounted on the left and right sidewalls of the cavity of the shell 1. The chemical energy release reaction device causes the elastic diaphragms 9 to expand through a chemical energy release reaction, thereby expelling water from the cavity through the jet nozzle, thus enabling the continuous blasting robot to move.

[0050] Continue reading Figure 1 The one-way water filtration device 2 includes a water filter plate with several water holes and a one-way valve 3 installed on each water hole.

[0051] Continue reading Figure 1 The outer shell 1 and its cavity have a structure that gradually thickens from front to back.

[0052] Continue reading Figure 1 The elastic diaphragm 9 and the inner wall of the outer shell 1 form a driving gas chamber. The chemical energy release reaction device includes a gas filling module, a control module 5, an inlet valve 6, a spark plug 7, and an exhaust valve 8. The inlet valve 6, spark plug 7, and exhaust valve 8 are located on one side of the driving gas chamber. The inlet valve 6 and exhaust valve 8 control the inlet and outlet of the driving gas chamber, respectively, and the spark plug 7 discharges the driving gas chamber. The gas filling module is connected to the inlet valve 6 to supply the chemical energy release reaction gas to the driving gas chamber. The selection of the chemical energy release reaction gas is a well-known technology and will not be described in detail. The control module 5 is connected to the spark plug 7 to control the discharge of the spark plug 7. When the gas filling module supplies gas to the driving gas chamber and the control module 5 controls the spark plug 7 to discharge, the chemical energy release reaction gas in the driving gas chamber undergoes a chemical energy release reaction, causing the elastic diaphragm 9 to bulge. The elastic diaphragm 9 pushes the water in the cavity to be ejected outward from the spray port, thereby generating a driving force that propels the continuous detonation robot forward.

[0053] The present invention also provides a control method for the continuous detonation robot mimicking the pyrosome as described above, comprising:

[0054] Step 1: Collect images and videos using a camera to achieve unified input of underwater images.

[0055] Step 2 involves preprocessing the uniformly input underwater images using a pyrosome swarm image processing algorithm. This algorithm, analogous to the biological characteristics of pyrosomes—"individual specialization and group aggregation"—maps the process from "massive images → clustering and aggregation → representative identification" to swarm intelligence behavior, forming a novel "coarse screening - fine screening" collaborative framework. This includes the following steps:

[0056] The S100 optimizes the underwater environment in underwater images through an underwater image preprocessing module, including noise reduction (Gaussian filtering / non-local mean filtering) and contrast enhancement (adaptive histogram equalization).

[0057] The S200 uses the lightweight YOLOX-tiny network for initial rapid identification.

[0058] The S300 extracts key features (such as edges, textures, and target contours) from the initial recognition results through a feature extraction module, combined with the color attenuation features unique to underwater targets.

[0059] S400, the clustering and classification module, based on feature vector similarity, groups similar images into one class, simulating the "information aggregation" characteristics of pyrosome group collaboration; it performs underwater image feature similarity calculation, assuming the feature vector of image i is... , Let n represent the component, n be the feature dimension, and the feature vector of image j be... Similarity for:

[0060]

[0061] in, D(i,j) is the vector inner product (which measures the directional similarity between two vectors), and D(i,j) is the underwater image degradation distance (calculated based on noise variance). Environmental adaptation factors ( (Increases when underwater turbidity). and denoted as the modulus of the eigenvector.

[0062] S500 represents the image selection module using entropy scoring to select 1-3 images with the "most informational content" from four categories: marine life, seabed environment, seabed resources, and underwater structures; and represents the image scoring formula. as follows:

[0063]

[0064] in, Image entropy, Representing an image Medium pixel value The probability, For clarity, The image is in coordinates gradient magnitude at that point It is the image size. For the target percentage, Weights (dynamically adjusted based on the characteristics of the underwater target).

[0065] Step 3: Use the YOLOX algorithm to identify the coordinates of the bounding box in the preprocessed image. The calculation formula is:

[0066]

[0067]

[0068]

[0069]

[0070] Among them, the coordinates of the bounding box Including center point coordinates Width w, Height h The x and y coordinates of the current grid center are given. For the current width, At the current altitude, coordinates of the center point Predict the offset. Predict the offset for width w. Predict the offset for height h. For the sigmoid function, and is a scaling factor for width and height, allowing the box size to be dynamically adjusted; e is a natural constant.

[0071] The coordinates of the true bounding box ( For pre-defined fixed values, the coordinates of the predicted bounding box should be infinitely close to the coordinates of the actual bounding box. ( ) represents the coordinates of the center point. For width, For height.

[0072] loss function for:

[0073]

[0074] Where IoU is the intersection-union ratio. Center of the prediction box Center of the real frame The square of the Euclidean distance, The length of the diagonal of the smallest rectangle enclosing the two frames. , It is a balancing factor.

[0075] Step 4: Control the robot's movement based on the recognition results; when an obstacle is detected ahead, control the expansion of the elastic diaphragm of the continuous detonation robot to achieve rapid obstacle avoidance; when a living organism that needs to be followed is detected ahead, increase the working frequency of the chemical energy release reaction device to achieve rapid following.

[0076] Continue reading Figure 4 This is a schematic diagram of the image processing algorithm for fire-body insect swarms, including an underwater image preprocessing module, a lightweight network-based initial fast recognition module, a feature extraction module, a clustering and classification module, and a representative image selection module.

[0077] by Figure 1 , Figure 2 The working process of the continuous detonation underwater robot, modeled after a pyrosome, is explained using an example. The outer shell 1 mimics the shape of a pyrosome, with a large opening at the front for water entry and a small opening at the rear for water exit. A one-way filtration device 2 and a one-way valve 3 are designed to ensure that water ejected from the cavity can immediately pass through the one-way filtration device 2 back into the chamber, while the one-way valve 3 prevents water from flowing back out. When underwater images and recognition data captured by the camera 4 mounted on the robot's head are transmitted to the control module 5, they are analyzed. Based on the analysis, propane and oxygen are premixed and injected into the drive gas chamber, igniting the spark plug 7. The elastic diaphragm rapidly expands, causing water in the conical opening to be ejected quickly, driving the underwater robot to move rapidly through the reaction force. After the reaction is complete, exhaust gas is discharged from the exhaust valve 8, ready for the next drive.

[0078] by Figure 3 , Figure 4 The following example illustrates the underwater image recognition and control algorithm for pyrosome swarms. The overall process includes unified input of underwater images, pyrosome swarm image processing, recognition of the preprocessed images based on the YOLOX algorithm, and control of robot movement based on the recognition results. When an obstacle is detected ahead, the expansion degree of the elastic diaphragms 9 on both sides of the robot is controlled to achieve rapid obstacle avoidance. When an underwater creature that needs to be followed is detected ahead, the working frequency of the spark plug 7 and the air intake valve 6 is increased to achieve rapid following. The pyrosome swarm image processing algorithm includes an underwater image preprocessing module, a lightweight network initial rapid recognition module, a feature extraction module, a clustering and classification module, and a representative image selection module.

[0079] The pyrosome swarm image processing algorithm has the following characteristics:

[0080] 1. Inspired by the collaborative mechanism of pyrosomes, and drawing analogy to the biological characteristics of pyrosomes' "individual division of labor - group aggregation", the process of "massive images → clustering and aggregation → representative recognition" is mapped to swarm intelligence behavior, forming a brand-new "coarse screening - fine screening" collaborative framework;

[0081] 2. Environmentally Adaptive Clustering: The clustering algorithm incorporates underwater image degradation features (such as color channel attenuation models) to solve the problem of "high misclassification rate" in traditional clustering for underwater images;

[0082] 3. Dynamic threshold control: The clustering threshold and the number of representative images are adjusted in real time by the control coordination module to achieve a dynamic balance between computing power and accuracy.

[0083] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A control method for a continuous detonation robot mimicking a pyrosome, characterized in that, The continuous blasting robot, which mimics the fire-body insect, includes an outer shell with a camera mounted on it. A jet nozzle is located at the rear end of the shell, and a one-way water filtration device is located at the front end. This one-way water filtration device allows water to enter the cavity of the shell but prevents water from leaving the cavity. An elastic diaphragm and a chemical energy release reaction device are mounted on the side wall of the cavity. The chemical energy release reaction device causes the elastic diaphragm to expand through a chemical energy release reaction, thereby expelling water from the cavity through the jet nozzle, thus enabling the continuous blasting robot to move. The control method includes: Step 1: Collect images and videos using a camera to achieve unified input of underwater images; Step 2: Preprocess the uniformly input underwater images using the pyrosome swarm image processing algorithm; the pyrosome swarm image processing algorithm includes the following steps: S100 optimizes the underwater environment in underwater images through an underwater image preprocessing module, including noise reduction and contrast enhancement; The S200 uses the lightweight YOLOX-tiny network for initial rapid identification. S300 extracts key features from the initial recognition results through the feature extraction module; S400, the clustering and classification module, based on feature vector similarity, groups similar images into one class, simulating the "information aggregation" characteristics of pyrosome group collaboration; it performs underwater image feature similarity calculation, assuming the feature vector of image i is... , Let n represent the component, n be the feature dimension, and the feature vector of image j be... Similarity for: ， in, Let D(i,j) be the vector dot product, and D(i,j) be the underwater image degradation distance. As an environmental adaptation factor, and The magnitude of the eigenvector; S500 represents the image filtering module's scoring based on entropy values, and represents the image scoring formula. as follows: ， in, Image entropy, Representing an image Medium pixel value The probability, For clarity, The image is in coordinates gradient magnitude at that point It refers to the image size. For the target percentage, As weight; Step 3: Use the YOLOX algorithm to recognize objects in front of the preprocessed image; Step 4: Control the robot's movement based on the recognition results; when an obstacle is detected ahead, control the expansion of the elastic diaphragm of the continuous detonation robot to achieve rapid obstacle avoidance; when a living organism that needs to be followed is detected ahead, increase the working frequency of the chemical energy release reaction device to achieve rapid following.

2. The control method for a continuous detonation robot mimicking a pyrosome according to claim 1, characterized in that, The one-way water filtration device includes a filter plate with several water holes and a one-way valve on each water hole.

3. The control method for a continuous detonation robot mimicking a pyrosome according to claim 1, characterized in that, The outer shell and its cavity have a structure that gradually thickens from front to back.

4. The control method for a continuous detonation robot mimicking a pyrosome according to claim 1, characterized in that, The elastic diaphragm and the inner wall of the shell form a driving gas chamber. The chemical energy release reaction device includes a gas charging module, a control module, an air inlet valve and a spark plug. The air inlet valve and the spark plug are disposed on the driving gas chamber. The gas charging module is connected to the air inlet valve to supply chemical energy release reaction gas to the driving gas chamber. The control module is connected to the spark plug to control the spark plug discharge. When the inflation module supplies air to the drive air chamber and the control module controls the spark plug to discharge, the chemical energy release reaction gas in the drive air chamber undergoes a chemical energy release reaction, causing the elastic diaphragm to bulge. The elastic diaphragm pushes the water in the cavity to be ejected outward from the injection port, thereby generating the driving force that propels the continuous detonation robot forward.

5. The control method for a continuous detonation robot mimicking a pyrosome according to claim 1, characterized in that, Step 3 uses the YOLOX algorithm to identify the coordinates of the preprocessed image and the bounding box prediction. The calculation formula is: ， ， ， ， Among them, the coordinates of the bounding box Including center point coordinates Width w, Height h The x and y coordinates of the current grid center are given. For the current width, At the current altitude, coordinates of the center point Predict the offset. Predict the offset for width w. Predict the offset for height h. For the sigmoid function, and is the scaling factor for width and height, and e is the natural constant; loss function for: ， Where IoU is the intersection-union ratio. Center of the prediction box Center of the real frame The square of the Euclidean distance, The length of the diagonal of the smallest rectangle enclosing the two frames. , As a balance factor, For width, For height.

Images