CN112124537B - Intelligent control method for underwater robot for autonomous absorption and fishing of benthos - Google Patents
Intelligent control method for underwater robot for autonomous absorption and fishing of benthos Download PDFInfo
- Publication number
- CN112124537B CN112124537B CN202011009529.1A CN202011009529A CN112124537B CN 112124537 B CN112124537 B CN 112124537B CN 202011009529 A CN202011009529 A CN 202011009529A CN 112124537 B CN112124537 B CN 112124537B
- Authority
- CN
- China
- Prior art keywords
- fuzzy
- motion
- fishing
- horizontal
- type
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000010521 absorption reaction Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000002787 reinforcement Effects 0.000 claims abstract description 23
- 230000004438 eyesight Effects 0.000 claims abstract description 14
- 230000006870 function Effects 0.000 claims description 28
- 230000004913 activation Effects 0.000 claims description 18
- 238000013459 approach Methods 0.000 claims description 10
- 230000008447 perception Effects 0.000 claims description 10
- 230000000007 visual effect Effects 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 6
- 238000001514 detection method Methods 0.000 abstract description 5
- 230000008901 benefit Effects 0.000 abstract description 4
- 238000013461 design Methods 0.000 abstract description 4
- 238000011160 research Methods 0.000 abstract description 4
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 230000009123 feedback regulation Effects 0.000 abstract description 3
- 230000009471 action Effects 0.000 description 12
- 239000002245 particle Substances 0.000 description 10
- 238000005457 optimization Methods 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 7
- 238000005086 pumping Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 230000003044 adaptive effect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000006399 behavior Effects 0.000 description 4
- 230000006872 improvement Effects 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 2
- 238000002955 isolation Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000016776 visual perception Effects 0.000 description 2
- 241000257465 Echinoidea Species 0.000 description 1
- 241000251511 Holothuroidea Species 0.000 description 1
- 241000237503 Pectinidae Species 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 238000009360 aquaculture Methods 0.000 description 1
- 244000144974 aquaculture Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 239000001064 degrader Substances 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000005911 diet Nutrition 0.000 description 1
- 230000000378 dietary effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000011664 nicotinic acid Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 235000020637 scallop Nutrition 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B63—SHIPS OR OTHER WATERBORNE VESSELS; RELATED EQUIPMENT
- B63C—LAUNCHING, HAULING-OUT, OR DRY-DOCKING OF VESSELS; LIFE-SAVING IN WATER; EQUIPMENT FOR DWELLING OR WORKING UNDER WATER; MEANS FOR SALVAGING OR SEARCHING FOR UNDERWATER OBJECTS
- B63C11/00—Equipment for dwelling or working underwater; Means for searching for underwater objects
- B63C11/48—Means for searching for underwater objects
- B63C11/49—Floating structures with underwater viewing devices, e.g. with windows ; Arrangements on floating structures of underwater viewing devices, e.g. on boats
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B63—SHIPS OR OTHER WATERBORNE VESSELS; RELATED EQUIPMENT
- B63C—LAUNCHING, HAULING-OUT, OR DRY-DOCKING OF VESSELS; LIFE-SAVING IN WATER; EQUIPMENT FOR DWELLING OR WORKING UNDER WATER; MEANS FOR SALVAGING OR SEARCHING FOR UNDERWATER OBJECTS
- B63C11/00—Equipment for dwelling or working underwater; Means for searching for underwater objects
- B63C11/52—Tools specially adapted for working underwater, not otherwise provided for
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Ocean & Marine Engineering (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Feedback Control In General (AREA)
Abstract
The invention belongs to the technical field of intelligent control of underwater robots, and particularly relates to an intelligent control method of an underwater robot for autonomous absorption and fishing of benthos. The invention is mainly used for completing the detection and identification of target organisms in a complex underwater environment, guiding the operation of a robot and realizing the accurate absorption of a specified target. During operation, the suction robot firstly identifies and tracks an operation target through an underwater vision and reinforcement learning algorithm, and then guides and completes autonomous suction and fishing operation of benthos through self pose feedback regulation and an intelligent control system of platform motion of the robot to deduce and optimize fuzzy rules. The invention can realize continuous and stable tracking and autonomous suction of targets based on advanced achievements in artificial intelligence research, has the advantages of accurate identification, high intelligence degree, high fishing efficiency, low operation cost and the like, is practically applied to underwater robot system design, and has important significance for efficient autonomous suction and fishing of marine organisms.
Description
Technical Field
The invention belongs to the technical field of intelligent control of underwater robots, and particularly relates to an intelligent control method of an underwater robot for autonomous absorption and fishing of benthos.
Background
With the development of marine economy and the improvement of the living standard of Chinese people, the dietary demand of people on high-protein marine organisms is higher and higher, and the current fishing of the marine organisms such as sea cucumbers, scallops, sea urchins and the like in a marine ranch is mainly completed by a diver. The manual fishing not only easily causes life and injury threats to divers, but also has less divers in the aspect recently, so that the autonomous fishing robot for marine organisms improves the fishing economy and greatly reduces the industrial crisis of marine organism fishing.
In recent years, researchers have conducted some research on marine life fishing robots, and for example, patent document "precious marine life fishing robot (cn201410686861. x)" mainly walks through the sea floor, and a manually-controlled robot observes and sucks marine life underwater to complete fishing of marine life. Patent documents "autonomous navigation and man-machine cooperative fishing operation system of benthos fishing robot (CN 201210553378.5)" and "a bionic benthos fishing robot (CN 201210553365.8)" have a certain autonomy in navigation and walking of underwater operation, but still rely on certain intervention of an operator on path planning, target searching, fishing instructions, and the like, and the autonomy is limited. The above invention is difficult to realize the intelligent control of the marine life and the autonomous absorption and capture, is difficult to use and popularize by common fishermen, and needs to invest certain manpower support and use training. In view of these problems, there is a high necessity for a control method of a marine organism fishing robot capable of realizing autonomous capturing in the field of aquaculture.
Disclosure of Invention
The invention aims to provide an intelligent control method of an underwater robot for the autonomous suction and fishing of benthos, which can realize the continuous tracking and the autonomous suction of targets.
The purpose of the invention is realized by the following technical scheme: the underwater robot for autonomous suction and catching of the benthos comprises a motion module, a suction and catching module, an underwater vision perception module and a II-type fuzzy approximation control module based on reinforcement learning; the motion module comprises a horizontal propeller and a vertical propeller; the underwater vision perception module comprises an absorption tracking observation camera arranged on the absorption fishing module; the method comprises the following steps:
step 1: setting a desired region in the shooting range of the absorption tracking observation camera, and selecting n desired points in the desired region; obtaining a suction fishing target relative to a desired areaPosition state error set St ═ St for each desired point in the domainx,Sty}; wherein StxSet of horizontal position state errors, Stx={pex1,pex2,...,pexn};StySet of errors for vertical position states, Sty={pey1,pey2,...,peyn};
Step 2: inputting the position state error set St of the suction-catching target relative to the reference point into a II-type fuzzy approximation control module based on reinforcement learning, wherein the II-type fuzzy approximation control module based on reinforcement learning comprises nxFuzzy rule of horizontal motion and nyFuzzy rules of vertical motion; set horizontal position state errors StxInputting the motion information into all horizontal motion fuzzy rules to obtain the output result of each horizontal motion fuzzy ruleix=1,2,...,nxAnd calculating the standard deviation sigma of the output results of all horizontal motion blur rulesx(ii) a Set of vertical position state errors StyInputting the motion information into all the vertical motion fuzzy rules to obtain the output result of each vertical motion fuzzy ruleiy=1,2,...,nyAnd calculating the standard deviation sigma of the output results of all horizontal motion blur rulesy;
And step 3: calculating an upper bound for activation strength for all horizontal motion blur rulesAnd lower boundConstruction of a type II fuzzy set of horizontal motionCalculating an upper bound for activation strength for all vertical motion blur rulesAnd lower boundConstructing type II fuzzy sets of vertical motion
Step 3.1: calculating a set of horizontal position state errors Stx={pex1,pex2,...,pexnState error pe of each horizontal position in thexjCorresponding to each horizontal motion blur rule ixUpper bound of the gaussian primary membership function ofAnd lower bound
step 3.2: calculating a set of vertical position state errors Sty={pey1,pey2,...,peynAt each vertical positionAttitude error peyjCorresponding to each vertical motion blur rule iyUpper bound of the gaussian primary membership function ofAnd lower bound
step 3.3: calculating an upper bound for activation strength for all horizontal motion blur rulesAnd lower boundConstruction of a type II fuzzy set of horizontal motion
Step 3.4: calculating an upper bound for activation strength for all vertical motion blur rulesAnd lower boundConstructing type II fuzzy sets of vertical motion
And 4, step 4: assembling horizontally moving II-type fuzzy setsAnd vertically moving type II fuzzy setsConverting the model reduction link into a linear fuzzy set, and calculating the voltage control value u of the horizontal thrusterxAnd a voltage control value u of the vertical thrustery;
Wherein, LcxAnd RcxSets of type II fuzzy moving horizontally respectivelyThe left and right intersection points of the Gaussian main membership function; lcyAnd RcySet of II types of blur, each in vertical motionThe left and right intersection points of the Gaussian main membership function;
and 5: II-type fuzzy approximation control module based on reinforcement learning is used for controlling voltage u of horizontal thrusterxAnd a voltage control value u of the vertical thrusteryThe target is transmitted to a motion module to ensure that the target can approach and always keep in a safe area of the visual field; judging whether the underwater robot which is automatically sucked and caught by the benthos finishes sucking the target or not; and if not, returning to the step 1 to perform intelligent control at the next moment.
The invention has the beneficial effects that:
the invention is mainly used for completing the detection and identification of target organisms in a complex underwater environment, guiding the operation of a robot and realizing the accurate absorption of a specified target. During operation, the suction robot firstly identifies and tracks an operation target through an underwater vision and reinforcement learning algorithm, and then guides and completes autonomous suction and fishing operation of benthos through self pose feedback regulation and an intelligent control system of platform motion of the robot to deduce and optimize fuzzy rules. The invention can realize continuous and stable tracking and autonomous suction of targets based on advanced achievements in artificial intelligence research, has the advantages of accurate identification, high intelligence degree, high fishing efficiency, low operation cost and the like, is practically applied to underwater robot system design, and has important significance for efficient autonomous suction and fishing of marine organisms.
Drawings
Fig. 1 is a main body diagram of an underwater robot for autonomous sucking and catching of benthos.
Fig. 2 is a general block diagram of a control system of an underwater robot for autonomous suction and fishing of benthos.
Fig. 3 is a schematic diagram of a type II fuzzy learning controller in accordance with the present invention.
Fig. 4 is a schematic diagram of a discrete grid of 2048 × 1536CCD image planes.
FIG. 5 is a schematic of the S-plane function.
Fig. 6 is a schematic diagram of a video frame in scene II of an autonomous capture process.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The invention provides an intelligent control method of an underwater robot for autonomous absorption and fishing of benthos, which is mainly used for completing detection and identification of target creatures in a complex underwater environment, guiding the robot to operate and realizing accurate absorption of a specified target. The invention comprises the following steps: the underwater robot carrier for the absorption type marine organism fishing, the underwater robot intelligent motion control system, the underwater vision perception system based on deep learning and the like. During operation, the suction robot firstly identifies and tracks an operation target through an underwater vision and reinforcement learning algorithm, and then guides and completes autonomous suction and fishing operation of benthos through self pose feedback regulation and an intelligent control system of platform motion of the robot to deduce and optimize fuzzy rules. The invention can realize continuous and stable tracking and autonomous suction of targets based on advanced achievements in artificial intelligence research, has the advantages of accurate identification, high intelligence degree, high fishing efficiency, low operation cost and the like, is practically applied to underwater robot system design, and has important significance for efficient autonomous suction and fishing of marine organisms.
An intelligent control method for an underwater robot for autonomous absorption and fishing of benthos, wherein the underwater robot for autonomous absorption and fishing of benthos comprises a motion module, an absorption and fishing module, an underwater vision perception module and a II-type fuzzy approximation control module based on reinforcement learning; the motion module comprises a horizontal propeller and a vertical propeller; the underwater vision perception module comprises an absorption tracking observation camera arranged on the absorption fishing module; the specific control method comprises the following steps:
step 1: setting a desired region in the shooting range of the absorption tracking observation camera, and selecting n desired points in the desired region; acquiring a position state error set St ═ St { St } of the suction fishing target relative to each expected point in the expected areax,Sty}; wherein StxSet of horizontal position state errors, Stx={pex1,pex2,...,pexn};StySet of errors for vertical position states, Sty={pey1,pey2,...,peyn};
Step 2: inputting the position state error set St of the suction-catching target relative to the reference point into a II-type fuzzy approximation control module based on reinforcement learning, wherein the II-type fuzzy approximation control module based on reinforcement learning comprises nxFuzzy rule of horizontal motion and nyFuzzy rules of vertical motion; set horizontal position state errors StxInputting the motion information into all horizontal motion fuzzy rules to obtain the output result of each horizontal motion fuzzy ruleix=1,2,...,nxAnd calculating the standard deviation sigma of the output results of all horizontal motion blur rulesx(ii) a Set of vertical position state errors StyInputting the motion information into all the vertical motion fuzzy rules to obtain the output result of each vertical motion fuzzy ruleiy=1,2,...,nyAnd calculating the standard deviation sigma of the output results of all horizontal motion blur rulesy;
And step 3: calculating an upper bound for activation strength for all horizontal motion blur rulesAnd lower boundConstruction of a type II fuzzy set of horizontal motionCalculating an upper bound for activation strength for all vertical motion blur rulesAnd lower boundConstructing type II fuzzy sets of vertical motion
Step 3.1: calculating a set of horizontal position state errors Stx={pex1,pex2,...,pexnState error pe of each horizontal position in thexjCorresponding to each horizontal motion blur rule ixUpper bound of the gaussian primary membership function ofAnd lower bound
step 3.2: calculating a set of vertical position state errors Sty={pey1,pey2,...,peynThe state error pe of each vertical position in theyjCorresponding to each vertical motion blur rule iyUpper bound of the gaussian primary membership function ofAnd lower bound
step 3.3: calculating an upper bound for activation strength for all horizontal motion blur rulesAnd lower boundConstruction of a type II fuzzy set of horizontal motion
Step 3.4: calculating an upper bound for activation strength for all vertical motion blur rulesAnd lower boundConstructing type II fuzzy sets of vertical motion
And 4, step 4: assembling horizontally moving II-type fuzzy setsAnd vertically moving type II fuzzy setsConverting the model reduction link into a linear fuzzy set, and calculating the voltage control value u of the horizontal thrusterxAnd a voltage control value u of the vertical thrustery;
Wherein, LcxAnd RcxSets of type II fuzzy moving horizontally respectivelyThe left and right intersection points of the Gaussian main membership function; lcyAnd RcySet of II types of blur, each in vertical motionThe left and right intersection points of the Gaussian main membership function;
and 5: II-type fuzzy approximation control module based on reinforcement learning is used for controlling voltage u of horizontal thrusterxAnd a voltage control value u of the vertical thrusteryThe target is transmitted to a motion module to ensure that the target can approach and always keep in a safe area of the visual field; judging whether the underwater robot which is automatically sucked and caught by the benthos finishes sucking the target or not; and if not, returning to the step 1 to perform intelligent control at the next moment.
Example 1:
the invention relates to an underwater robot for automatically sucking and catching benthos, which mainly comprises the following parts: the underwater robot carrier for the absorption type marine organism fishing, the underwater robot intelligent motion control system, the underwater vision perception system based on deep learning and the like. An underwater robot for catching marine organisms is taken as a carrier, and a deep learning method for underwater visual perception is combined with an underwater robot intelligent motion control method for control and detection. The carrier platform consists of a propeller, a pumping structure and a storage box; the underwater visual perception method based on deep learning jointly completes biological density detection and establishment of a kinematic relationship between an image and an absorbed target on the basis of two sets of fisheye cameras; the underwater robot intelligent motion control method takes a cabin PC104 hardware circuit as a core, and intelligent suction capture of marine organism targets is completed through an intelligent suction motion control method based on reinforcement learning optimization. The method establishes fuzzy rules by determining a reasoning system of II-type fuzzy, designs an S-surface adaptive function by the rules derived by subdivision, and finally ensures that the robot finishes accurate absorption of biological targets appearing in a formulated area by a reinforcement learning method.
Constructing behavior rules according to the current and expected states of the depth and direction of the marine organism pumping suction pipe, and determining fuzzy rules, degradation types and deblurring output by the following methods: determining a fuzzy rule:
in the formula, mijIs an uncertain mean valueσijIs a fixed standard deviation;representing a degree of membership, which is a bounded set,andrespectively representUpper and lower bounds off i;Upper and lower bounds for activation intensity, respectively; pe (t) represents the target position state error within the camera range; n is the number of states input.
Type reduction:
wherein mr is a rule number; lc and Rc are the left and right intersections, respectively; a isiIs the corresponding behavior of the fuzzy rule; n number of states input.
And (3) resolving fuzzy output to a thruster voltage:
enabling the target to appear in a proper suction area in a distance camera coordinate system according to a fuzzy rule, and then selecting the optimal values of the particles and the particle swarm to be updated in a non-stage mode, wherein the method is realized by the following fitness function:
in the formula, keAnd kuIs an adjustable parameter; k is a radical ofpIs a proportionality coefficient; k is a radical ofdIs a differential coefficient; e.g. of the typeiAnd aiThe error in a certain direction and the corresponding behavior of the fuzzy rule.
The underwater robot for automatically sucking and fishing the benthos mainly comprises: the underwater robot carrier for the absorption type marine organism fishing, the underwater robot intelligent motion control system, the underwater vision perception system based on deep learning and the like. The absorption type marine organism fishing underwater robot carrier comprises a carrier frame, four horizontal propellers, two vertical propellers, a marine organism suction pump device, an absorption storage tank and the like, wherein the four horizontal propellers are arranged in a vector mode. The underwater vision perception system based on the deep learning comprises a fisheye camera arranged in front of the robot and a suction tracking observation camera arranged on a suction pipe; the fish-eye camera detects the existence of marine life and determines the density of the marine life through deep learning and distortion processing, and the suction tracking observation camera on the suction pipe establishes the kinematic relationship between the camera image and the suction target based on the kinematic modeling of the image. The underwater robot intelligent motion control system mainly takes a PC104 control computer in a pressure-resistant cabin as a core, and controls the motion and the suction of the underwater robot by an intelligent suction motion control method based on reinforcement learning according to the feedback of a magnetic compass, a depth meter and an underwater camera, thereby realizing the intelligent suction and capture of marine organisms.
An intelligent control method of an underwater robot for autonomous absorption and fishing of benthos comprises the following steps: firstly, establishing a II-type fuzzy inference system based on the depth and the angle of a target relative to an image, and initially establishing a fuzzy rule; secondly, designing an improved particle optimization method to subdivide and optimize the derivative fuzzy rule, particularly designing an S-surface-based adaptive function for the adaptive judgment of the derivative fuzzy, and determining and judging the fuzzy rule; and finally, further training fuzzy rule output by combining the current suction state and the external environment state of the robot by using a reinforcement learning method, thereby ensuring that the target can approach and always keep in a safe area of the visual field, and controlling a suction pipe of the underwater robot to gradually approach the target to finish suction of marine life.
And the intelligent approach controller of the suction robot enables the pumping pipeline to be continuously close to the target to finish suction according to the visual characteristics of the image. The advantage of this controller is that the control performance is improved by the context-specific automatic selection strategy. In addition, for low cost robots without horizontal velocity or position sensors, the present invention can stabilize the target within a safe suction area, with the pumping tubing constantly approaching the target. Type II fuzzy systems help to minimize the uncertain effects of unknown environments compared to type I fuzzy systems. The invention realizes the stable and accurate motion control of the ROV through the intelligent motion control system, and ensures that the target is in a safe range and is continuously close to the target.
The type II fuzzy inference system aims at constructing action rules and determining action set output ranges according to current and expected states of the depth and the direction of a pipeline. Wherein the action set indicates that a group of action robots of the robot will perform actions such as heading, sideways and heave through state evaluation to stabilize their pose, hold the target in the camera, control the ROV to approach the target and be ready for absorption. Referring to fig. 4 of the drawings, a 2048 × 1536 pixel CCD image plane may be divided into 32 × 24 discrete grids each containing 64 × 64 pixels, with the different shaded areas in the figure representing the desired reach, approach and danger areas, respectively. The definition of the type II fuzzy set in the invention is designed as follows:
in the formula,pe (t) represents the target position state error within the camera range; μ (t) represents a fuzzy membership function; st represents the state set of pe (t); js denotes the members in pe (t).
Its primary gaussian membership function can be expressed as:
in the formula, mijIs an uncertain mean valueσijIs a fixed standard deviation;representing the degree of membership, which is a bounded set,andrespectively representUpper and lower bounds of (1):
the operation of fuzzy t norm in the inference machine is realized by algebraic product, the rule activation strength Fi is interval type I fuzzy set, and the activation strength of the suction fishing target corresponding to the ith rule at present can be obtained by continuous multiplication of fuzzy operation:
wherein,f i;the activation intensity is respectively the upper bound and the lower bound, because the output of the system inference engine is a type II fuzzy set, the fuzzy set is converted into a type I fuzzy set through a model reduction link, and then the fuzzy solution operation can be carried out, the invention uses the following method to carry out the model reduction operation:
wherein mr is a rule number; lc and Rc are the left and right intersections, respectively;f i;upper and lower bounds for activation intensity, respectively; a isiIs the result deduced by the ith fuzzy rule; n is the number of states input. The voltage value u finally output to the propeller0Comprises the following steps:
a learning-based Particle Swarm Optimization (PSO) fuzzy rule optimizer is developed for uncertain environment and dynamic changes and factors in the capture process. The purpose of the fuzzy rule is to select an ROV action to approach the aspiration target based on the current and expected state of the depth and orientation of the ROV tube. The subdivision and derivation optimization method is as follows: the adaptive function of the rule enables the target to quickly reach a desired region when the target is far away from a suction region in a camera coordinate system through a fuzzy rule, and enables a suction pipe of the underwater robot to be stabilized in the suction region of the camera coordinate system through the fuzzy rule subdivision control quantity when the target is near the suction region in the camera coordinate system, and suction is completed. Considering that different field environments exist in the capturing process, the fixed fuzzy rule is lack of universality, and the fuzzy rule is trained and optimized through a particle swarm optimization algorithm. The fuzzy rule optimization algorithm process is described as follows:
(1) initialization: fuzzy rules for actions and input states have been randomly generated before PSO optimization.
(2) Fitness function determination: for each fuzzy rule test, the fitness function is important to determine the optimal action of the controller. Because the designed learning-based II-type fuzzy approximation controller is accurate to realize online, the selection of an effective and intelligent fitness function is of great significance for realizing the quick iteration and optimization of the fuzzy rule in each step length released by the control command. To achieve rapid target access and absorption maintenance, it is desirable to control the motion loosely when the deviation is large and tightly when the deviation is small. The S-plane function shown in FIG. 5 of the specification is one of these functions, and the adaptive function is designed as follows:
in the formula, keAnd kuIs an adjustable parameter; k is a radical ofpIs a proportionality coefficient; k is a radical ofdIs a differential coefficient; e.g. of the typeiAnd aiThe error in a certain direction and the corresponding behavior of the fuzzy rule.
(3) Particle memory and selection: each rule particle will be evaluated by remembering its own fitness value and selecting the largest one as the fitness.
Each particle is further modified based on the above method. The above steps are repeatedly performed until a significant improvement is achieved. The particle having the best adaptability is a globally optimal particle, and thus, an optimal fuzzy rule adaptability value is obtained to cope with environmental disturbance and load variation during the work.
With reference to fig. 1, the designed marine organism autonomous sucking robot based on the intelligent control method comprises a carrier platform, a control cabin, a pumping suction pipe, a propeller of a propeller, a depth meter and a doppler velocimeter; the control cabin comprises hardware such as a PC104 core module, an I/O board, a CAN data acquisition board, an isolation serial port board, a data acquisition board, a direct current servo motor control board, a propeller motor driver, a magnetic compass and a marine organism capturing and storing cabin.
With reference to fig. 2, the core of the unmanned underwater organism suction robot control system is a PC104 core module, and the PC104 module acquires data of a doppler velocimeter and a magnetic compass through an isolation serial port plate and performs dead reckoning to obtain the current pose of the unmanned underwater vehicle; and the PC104 module sends a control instruction to the motor driver through the direct-current servo motor control panel according to the result of the coordinated motion controller, and controls the suction robot propeller to move so as to realize the stable and accurate motion of the carrier platform in the suction fishing operation process.
With reference to fig. 3, the unmanned underwater organism suction robot utilizes a II-type fuzzy approximation controller based on reinforcement learning to retain the visual characteristics of the marine organisms to be caught in the image, and can simultaneously approximate the pumping pipeline to the suction target. Defining a set of target position state error sets St ═ { pe) within a camera region1 pe2 ... penAnd inputting a clear accurate value by the fuzzy controller, and inputting a deviation amount between the position of the robot and an expected arrival area by an input state layer. Fuzzification operation fuzzifies an input accurate value into a fuzzy set, each node corresponding to input in the layer defines an interval type II Gaussian membership function, the rule of the type II fuzzy system adopts an If-Then form, and the rule is as follows: if pe1(t)is St1 and,…,pen(t)is Stn,then u1(t)is a1(t)and,…,um(t)is am(t). A fuzzy rule may determine a type ii output fuzzy set, from which the inference engine infers to determine the action to be taken next. Fuzzy inference carries out fuzzy operation through continuous multiplication to obtain activation strength, a degrader maps a type II fuzzy set to a type I fuzzy set, and a defuzzifier converts fuzzy output defuzzification into a clear set.
Referring to fig. 4, a CCD image plane of 2048 × 1536 pixels in the unmanned underwater bioabsorption robot underwater vision system may be divided into 32 × 24 discrete grids each including 64 × 64 pixels. Fig. 4 can show the current state and the expected state of the depth and the orientation of the pumping pipeline, the visual field of the camera comprises a desired area, an approaching area, a dangerous area, a safe area and the like which are distinguished by different colors according to the graph, different areas correspond to different sucking success rates, and the control system aims to control the submarine organism sucking robot suction pipe to be stabilized in the desired area of a red square.
With reference to fig. 5, the determination of the fitness function in the fuzzy rule of the reinforcement learning of the unmanned underwater organism suction robot plays a role in determining the optimal action of the controller. The II-type fuzzy approximation controller based on reinforcement learning aims to ensure accurate online realization and complete stable absorption capture. Therefore, the invention selects an effective and intelligent fitness function to realize the quick iteration and optimization of the fuzzy rule in each step of the release of the control command. In order to achieve rapid target approach and absorption maintenance, it is desirable to have a relaxed control action when the deviation is large and a strict control action when the deviation is small, so the S-plane function shown in fig. 5 is selected to achieve the above control purpose.
With reference to fig. 6, six frame images show the process of switching the autonomous recognition target from the environmental disturbance to the biological target block during the operation of the unmanned underwater bioabsorption robot. The invention obtains the relation between the pixel distance and the actual distance through underwater height-setting calibration measurement, and meanwhile, the II-type fuzzy controller based on reinforcement learning can optimize the tracking rule and continuously adapt to the environment, thereby realizing the continuous tracking of the target in the complex underwater environment.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (1)
1. An intelligent control method for an underwater robot for autonomous absorption and fishing of benthos, wherein the underwater robot for autonomous absorption and fishing of benthos comprises a motion module, an absorption and fishing module, an underwater vision perception module and a II-type fuzzy approximation control module based on reinforcement learning; the motion module comprises a horizontal propeller and a vertical propeller; the underwater vision perception module comprises an absorption tracking observation camera arranged on the absorption fishing module; the method is characterized by comprising the following steps:
step 1: setting a desired region in the shooting range of the absorption tracking observation camera, and selecting n desired points in the desired region; obtaining the position of the suction fishing target relative to each expected point in the expected areaSet of state errors St ═ Stx,Sty}; wherein StxSet of horizontal position state errors, Stx={pex1,pex2,...,pexn};StySet of errors for vertical position states, Sty={pey1,pey2,...,peyn};
Step 2: inputting the position state error set St of the suction fishing target relative to each expected point in the expected area into a II type fuzzy approximation control module based on reinforcement learning, wherein the II type fuzzy approximation control module based on reinforcement learning comprises nxFuzzy rule of horizontal motion and nyFuzzy rules of vertical motion; set horizontal position state errors StxInputting the motion information into all horizontal motion fuzzy rules to obtain the output result a of each horizontal motion fuzzy ruleix,ix=1,2,...,nxAnd calculating the standard deviation sigma of the output results of all horizontal motion blur rulesx(ii) a Set of vertical position state errors StyInputting the motion information into all the vertical motion fuzzy rules to obtain the output result a of each vertical motion fuzzy ruleiy,iy=1,2,...,nyAnd calculating the standard deviation sigma of the output results of all horizontal motion blur rulesy;
And step 3: calculating an upper bound for activation strength for all horizontal motion blur rulesAnd lower boundConstruction of a type II fuzzy set of horizontal motionCalculating an upper bound for activation strength for all vertical motion blur rulesAnd lower boundConstructing type II fuzzy sets of vertical motion
Step 3.1: calculating a set of horizontal position state errors Stx={pex1,pex2,...,pexnState error pe of each horizontal position in thexjCorresponding to each horizontal motion blur rule ixUpper bound of the gaussian primary membership function ofAnd lower bound
step 3.2: calculating a set of vertical position state errors Sty={pey1,pey2,...,peynThe state error pe of each vertical position in theyjCorresponding to each vertical motion blur rule iyDegree of gaussian principal membershipUpper bound of functionAnd lower bound
step 3.3: calculating an upper bound for activation strength for all horizontal motion blur rulesAnd lower boundConstruction of a type II fuzzy set of horizontal motion
Step 3.4: calculating an upper bound for activation strength for all vertical motion blur rulesAnd lower boundConstructing type II fuzzy sets of vertical motion
And 4, step 4: assembling horizontally moving II-type fuzzy setsAnd vertically moving type II fuzzy setsConverting the model reduction link into a linear fuzzy set, and calculating the voltage control value u of the horizontal thrusterxAnd a voltage control value u of the vertical thrustery;
Wherein, LcxAnd RcxSets of type II fuzzy moving horizontally respectivelyThe left and right intersection points of the Gaussian main membership function; lcyAnd RcySet of II types of blur, each in vertical motionThe left and right intersection points of the Gaussian main membership function;
and 5: II-type fuzzy approximation control module based on reinforcement learning is used for controlling voltage u of horizontal thrusterxAnd a voltage control value u of the vertical thrusteryThe target is transmitted to a motion module to ensure that the target can approach and always keep in a safe area of the visual field; judging whether the underwater robot which is automatically sucked and caught by the benthos finishes sucking the target or not; and if not, returning to the step 1 to perform intelligent control at the next moment.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011009529.1A CN112124537B (en) | 2020-09-23 | 2020-09-23 | Intelligent control method for underwater robot for autonomous absorption and fishing of benthos |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011009529.1A CN112124537B (en) | 2020-09-23 | 2020-09-23 | Intelligent control method for underwater robot for autonomous absorption and fishing of benthos |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112124537A CN112124537A (en) | 2020-12-25 |
CN112124537B true CN112124537B (en) | 2021-07-13 |
Family
ID=73842905
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011009529.1A Active CN112124537B (en) | 2020-09-23 | 2020-09-23 | Intelligent control method for underwater robot for autonomous absorption and fishing of benthos |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112124537B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112987741B (en) * | 2021-02-24 | 2024-06-18 | 武汉理工大学 | Ship course intelligent control method oriented to uncertain disturbance |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103057678A (en) * | 2012-12-18 | 2013-04-24 | 浙江工业大学 | Autonomous navigation and man-machine coordination catching operating system of benthic organism catching robot |
JP2016150681A (en) * | 2015-02-18 | 2016-08-22 | 学校法人金沢工業大学 | Robot for amphibian survey |
CN107499476A (en) * | 2017-08-21 | 2017-12-22 | 江苏科技大学 | Underwater robot control system and motion control method |
CN108319286A (en) * | 2018-03-12 | 2018-07-24 | 西北工业大学 | A kind of unmanned plane Air Combat Maneuvering Decision Method based on intensified learning |
CN108423141A (en) * | 2018-02-05 | 2018-08-21 | 浙江大学 | A kind of underwater operation robot and its control method |
CN109240091A (en) * | 2018-11-13 | 2019-01-18 | 燕山大学 | A kind of underwater robot control method based on intensified learning and its control method tracked |
KR20190129379A (en) * | 2018-05-10 | 2019-11-20 | 한국해양과학기술원 | Underwater Remote Unmanned System |
CN111062990A (en) * | 2019-12-13 | 2020-04-24 | 哈尔滨工程大学 | Binocular vision positioning method for underwater robot target grabbing |
CN111079936A (en) * | 2019-11-06 | 2020-04-28 | 中国科学院自动化研究所 | Wave fin propulsion underwater operation robot tracking control method based on reinforcement learning |
-
2020
- 2020-09-23 CN CN202011009529.1A patent/CN112124537B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103057678A (en) * | 2012-12-18 | 2013-04-24 | 浙江工业大学 | Autonomous navigation and man-machine coordination catching operating system of benthic organism catching robot |
JP2016150681A (en) * | 2015-02-18 | 2016-08-22 | 学校法人金沢工業大学 | Robot for amphibian survey |
CN107499476A (en) * | 2017-08-21 | 2017-12-22 | 江苏科技大学 | Underwater robot control system and motion control method |
CN108423141A (en) * | 2018-02-05 | 2018-08-21 | 浙江大学 | A kind of underwater operation robot and its control method |
CN108319286A (en) * | 2018-03-12 | 2018-07-24 | 西北工业大学 | A kind of unmanned plane Air Combat Maneuvering Decision Method based on intensified learning |
KR20190129379A (en) * | 2018-05-10 | 2019-11-20 | 한국해양과학기술원 | Underwater Remote Unmanned System |
CN109240091A (en) * | 2018-11-13 | 2019-01-18 | 燕山大学 | A kind of underwater robot control method based on intensified learning and its control method tracked |
CN111079936A (en) * | 2019-11-06 | 2020-04-28 | 中国科学院自动化研究所 | Wave fin propulsion underwater operation robot tracking control method based on reinforcement learning |
CN111062990A (en) * | 2019-12-13 | 2020-04-24 | 哈尔滨工程大学 | Binocular vision positioning method for underwater robot target grabbing |
Non-Patent Citations (1)
Title |
---|
《基于视觉感知的海生物吸纳式水下机器人目标捕获控制》;周浩等;《机器人》;20190331;第41卷(第2期);第242-249页 * |
Also Published As
Publication number | Publication date |
---|---|
CN112124537A (en) | 2020-12-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11363929B2 (en) | Apparatus and methods for programming and training of robotic household appliances | |
Zhu et al. | Complete coverage path planning of autonomous underwater vehicle based on GBNN algorithm | |
Cai | ROBOTICS: From Manipulator to Mobilebot | |
Liu et al. | Biologically inspired behaviour design for autonomous robotic fish | |
CN103057678B (en) | The autonomous navigation of benthic organism hauls robot and man-machine coordination fishing operation system | |
Myint et al. | Dual-eyes vision-based docking system for autonomous underwater vehicle: an approach and experiments | |
Ji-Yong et al. | Design and vision based autonomous capture of sea organism with absorptive type remotely operated vehicle | |
CN111746728B (en) | Novel overwater cleaning robot based on reinforcement learning and control method | |
CN114005021B (en) | Laser vision fusion based unmanned inspection system and method for aquaculture workshop | |
CN109784201A (en) | AUV dynamic obstacle avoidance method based on four-dimensional risk assessment | |
CN112124537B (en) | Intelligent control method for underwater robot for autonomous absorption and fishing of benthos | |
Farooq et al. | Microcontroller based neural network controlled low cost autonomous vehicle | |
Yan et al. | Real-world learning control for autonomous exploration of a biomimetic robotic shark | |
Fu et al. | Towards computational models of insect motion detectors for robot vision | |
Ishii et al. | A self-organizing map based navigation system for an underwater robot | |
CN115805595B (en) | Robot navigation method and device and sundry cleaning robot | |
CN115373383B (en) | Autonomous obstacle avoidance method and device for garbage recycling unmanned ship and related equipment | |
Tang et al. | Modeling and control of hybrid-driven gliding motion for an underwater gliding snake-like robot | |
Chou et al. | An AI AUV enabling vision-based diver-following and obstacle avoidance with 3D-modeling dataset | |
Qi et al. | Autonomous underwater rescue technology | |
Liu et al. | A Multi-AUVs Cooperative Hunting Algorithm for Environment with Ocean Current | |
Nakamura et al. | Development of Dual-eyes Docking System for AUV with Lighting 3D Marker | |
Aamer et al. | A Novel Algorithm for Autonomous Robot Navigation System Using Neural Network | |
Igbinedion et al. | Learning When to Ask for Help: Efficient Interactive Navigation via Implicit Uncertainty Estimation | |
Derbas et al. | SLAM Algorithm for Omni-Directional Robots based on ANN and EKF |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |