Underwater robot system and method for self-adaptive perception calculation scheduling

Abstract

The invention discloses an underwater robot system and method for self-adaptive perception calculation scheduling, and relates to the technical field of big data analysis. According to the method, node data is formed by collecting data such as the state and task progress of an underwater robot; constructing a network and electing a temporary leader node to form a distributed data set; constructing an energy change curve based on the node data; sending a transmission request when the energy is lower than a threshold value; and the response node calculates the transmission efficiency and generates a data packet. The system comprises a data sensing module, a dynamic networking module, an energy prediction module, a cooperation request module and an energy transmission module, and the modules respectively execute the functions of data acquisition and encapsulation, network construction, energy prediction, request sending, energy transmission and the like. According to the invention, through mechanisms such as dynamic networking, energy prediction and adaptive transmission, the energy management efficiency and task execution capability of the underwater robot are improved.

Description

Technical Field

[0001] This invention relates to the field of big data analytics, specifically to an adaptive perception-computation scheduling underwater robot system and method. Background Technology

[0002] The ocean covers most of the Earth's surface, but its depths still hold many unknowns. Underwater robots can carry various scientific instruments, such as acoustic sensors, optical sensors, and water quality monitoring equipment, to conduct detailed observations and data collection on the physical, chemical, and biological aspects of the ocean. They can reach deep-sea areas inaccessible to humans, providing invaluable data for marine scientific research.

[0003] However, existing underwater robots suffer from insufficient power endurance. During long-term operations or long-distance exploration missions, they often have to interrupt their missions due to energy depletion, which limits their application scope and work efficiency. Summary of the Invention

[0004] The purpose of this invention is to provide an adaptive perception-computation scheduling underwater robot system and method to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an adaptive perception-computation scheduling method for underwater robots, the underwater robot method comprising the following steps: Step S1: Collect the underwater robot's own status data and task progress to form a data set, generate the corresponding timestamp, and package the data set and timestamp into a node, while embedding three-dimensional position coordinates to form node data; Step S2: Construct an underwater robot network, collect node data of all underwater robots, elect a temporary leader node, and form a distributed data set; Step S3: Based on the underwater robot's own state data and mission progress in the node data, construct an energy change curve; Step S4: Collect the location information of all underwater robots. When the energy of any underwater robot is lower than the preset threshold, send an energy transfer request to the temporary leader node. Step S5: The underwater robot that receives the energy transfer request calculates the required energy transfer efficiency based on its own location and generates a data packet for energy transfer.

[0006] Furthermore, in step S1, the self-state data includes energy, speed, and mission load, constructing a vector S=[E, V, P]. sen E represents energy, V represents speed, and P represents speed. sen The task load; the task progress data includes the progress of completed tasks and the remaining task duration, constructing a vector T=[A doneT left A done For the progress of the completed task, T left The remaining task duration is denoted as (x, y, z) using an inertial navigation system and underwater acoustic positioning technology. The three-dimensional coordinates are obtained using an inertial navigation system and underwater acoustic positioning technology, and are denoted as (x, y, z) with a timestamp t added. Here, x is the horizontal coordinate, y is the vertical coordinate, z is the water depth where the underwater robot is located, and t is the time when the three-dimensional coordinates were obtained.

[0007] Furthermore, the specific steps of step S2 are as follows: Step S2-1: Each underwater robot calculates its own comprehensive index in real time. The calculation formula is as follows: ; Where C represents the underwater robot's overall performance indicators, and E represents the underwater robot's current energy. Where α is the task load rate, D is the communication distance of the node, and α, β, and γ are weighting coefficients. A temporary leader node is elected through a consensus mechanism. The temporary leader node is the elected temporary network manager. Non-temporary leader nodes are connected with adjacent underwater robot nodes to form subnets. The subnets are local networks automatically formed by non-leader nodes according to their communication radius. When the energy of the temporary leader node falls below a threshold, a new leader node is elected. Step S2-2: Node data within each subnet is synchronized to the entire underwater robot network through a temporary leader node, forming a distributed data set covering all underwater robots.

[0008] Furthermore, the specific steps of step S3 are as follows: Step S3-1: Parse the sailing speed sequence {V} from the node data. t Energy change ΔE t =E t -E t−1 The task consumption intensity λ, where the sailing speed sequence is a set of sailing speeds in a discrete time series; a sliding window technique is introduced to calculate the average speed μ within the window. V Energy standard deviation σ E Task completion rate ρ, construct feature vector F t =[μ V ,σ E ,ρ,λ]; Where the energy standard deviation σ E The calculation formula is: ; in the formula Let be the energy value at time t within the window. Let n be the average energy within the window, and n be the number of points within the window. The mean velocity μ within the windowV The calculation formula is: ; in the formula Let n be the speed at time t within the window, and n be the number of points within the window. The formula for calculating the task consumption intensity λ is: ; in the formula The task is complete, and t represents time. The formula for calculating the task completion rate ρ is: ; in the formula The task has been completed. For the overall task; Step S3-2: Use the ARIMA model to capture the time-series periodicity of energy changes, and use an LSTM network to learn the nonlinear effects of airspeed and mission load on energy; establish a joint prediction model: ; in To predict the remaining energy, ARIMA is an autoregressive integral moving average model. for LSTM output characteristics , ,..., Let be the node energy at times t-1, t-2, ..., tn; Step S3-3: Construct a time window, Arrange them in chronological order and plot the energy change curves.

[0009] Furthermore, the specific steps of step S4 are as follows: Step S4-1: Construct a 3D adjacency graph including the depth dimension, mark the neighboring nodes of a given node as neighboring nodes, and obtain all neighboring nodes to form a neighborhood set: ; in Let be the Euclidean distance between nodes i and j, h be the depth tolerance, and z be the depth coordinates of the underwater robot. For communication radius; Step S4-2: Sort the neighboring nodes in the neighborhood set in descending order according to the comprehensive value and store them in the temporary leader node. A node obtains the neighboring node order and sends requests in descending order. The node that sends the request is marked as the requesting node, and the neighboring nodes that receive the request are marked as the responding nodes. The comprehensive value is calculated. The formula is: ; Where w is the weighting coefficient. The current energy of the neighboring nodes. Let be the Euclidean distance between node i and node j. This is a function to indicate task conflicts. Step S4-3: Obtain the remaining energy of the response node itself. Remaining task duration The transferable energy of neighboring nodes is calculated using the following formula: ; in In response to the current energy of the neighboring nodes, In response to the remaining task time of neighboring nodes, Total task duration for the response node; like If >0, then accept the request; if If the value is less than 0, then calculate the transferable energy of the next node.

[0010] Furthermore, the shortest path from the response node to the request node is searched in the three-dimensional neighborhood graph. If a node exists on the shortest path, a multi-level transmission mode is adopted. This multi-level transmission mode involves energy forwarding through nodes on the shortest path. The mapping relationship between transmission rate and water depth is calculated, and the transmission rate is adjusted according to the current water depth. The calculation formula is as follows: ; in The energy transfer rate at the current water depth, denoted as , where k is the surface energy transfer rate, k is the seawater attenuation coefficient, and z is the current water depth. After the transmission is complete, the requesting node sends a data packet containing the amount of energy received and the transmission time to the responding node.

[0011] Furthermore, an adaptive perception computing scheduling underwater robot system is provided, the underwater robot system comprising a data perception module, a dynamic networking module, an energy prediction module, a cooperation request module, and an energy transmission module; The data sensing module is used to collect the underwater robot's own status data and task progress data, generate timestamps, and embed three-dimensional position coordinates to form node data; the dynamic networking module is used to construct an underwater robot network, collecting all node data to form a distributed dataset; the energy prediction module is used to construct an energy consumption prediction curve based on the underwater robot's operating characteristics and task progress in the node data; the cooperation request module is used to collect the position information of all underwater robots, and send an energy transfer request to adjacent underwater robots when the energy of any underwater robot is lower than a preset threshold; the energy transfer module is used to receive the energy transfer request from the underwater robot, and transfer energy to the requesting underwater robot when there is sufficient remaining energy. The output of the data sensing module is electrically connected to the input of the dynamic networking module; the output of the dynamic networking module is electrically connected to the input of the energy prediction module; the output of the energy prediction module is electrically connected to the input of the cooperation request module; and the output of the cooperation request module is electrically connected to the input of the energy transmission module.

[0012] Furthermore, the data sensing module includes a heterogeneous data acquisition unit and a spatiotemporal encapsulation unit; the heterogeneous data acquisition unit is used to collect the underwater robot's energy, speed, sensor load power and its own status data, as well as task progress data such as the area of ​​the completed detection area and the remaining task duration; the spatiotemporal encapsulation unit is used to add timestamps to the data set, embed three-dimensional position coordinates, and package the data set and timestamps into node data.

[0013] Furthermore, the dynamic networking module includes a leader election unit and a subnet data synchronization unit; the leader election unit is used for each underwater robot to calculate its own comprehensive indicators in real time, select a temporary leader node through a consensus mechanism, and re-elect a leader node after its energy falls below a threshold; the subnet data synchronization unit is used to synchronize the node data in each subnet to the entire underwater robot network through the leader node, forming a distributed data set. The energy prediction module includes a feature extraction unit and a hybrid modeling unit. The feature extraction unit is used to parse the sailing speed sequence, energy change, and task consumption intensity from the node data, calculate the mean speed, standard deviation of energy, and task completion rate within the window, and construct a feature vector. The hybrid modeling unit is used to capture the time series periodicity of energy changes using an ARIMA model, learn the nonlinear influence of sailing speed and task load on energy through an LSTM network, establish a joint prediction model, and plot the energy change curve.

[0014] Furthermore, the collaboration request module includes a three-dimensional neighborhood unit and a target request unit; the three-dimensional neighborhood unit is used to construct a three-dimensional adjacency graph containing a depth dimension and determine the neighborhood set of each node; the target request unit is used to sort the neighborhood nodes in descending order according to the multi-objective function and send energy transfer requests to the node with the highest function value first. The energy transmission module includes a path planning unit and an adaptive transmission unit. The path planning unit is used to search for the shortest path from the response node to the requesting node in the three-dimensional neighborhood graph, avoiding obstacles, and enabling multi-hop transmission mode when relay nodes exist. The adaptive transmission unit is used to calculate the mapping relationship between transmission rate and water depth, dynamically adjust the transmission rate according to the current water depth, and receive the data packet containing the energy received and transmission time sent by the requesting node after the transmission is completed.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention collects energy, speed and mission load data through multi-dimensional data acquisition and spatiotemporal encapsulation, constructs a vector model and embeds three-dimensional coordinates and timestamps to form node data, providing basic information for energy prediction and mission scheduling.

[0016] 2. This invention uses a dynamic networking mechanism to elect a temporary leader node by calculating comprehensive indicators for each robot. Non-leader nodes form subnets. When the leader node's energy is insufficient, a new election is held, which enables global synchronization of node data and ensures dynamic stability of the network and efficient data interaction.

[0017] 3. This invention constructs a three-dimensional adjacency graph with depth dimension through a three-dimensional neighborhood perception and intelligent request strategy. It sorts neighborhood nodes based on a multi-objective function of energy, distance, and task conflict, and prioritizes sending requests to nodes with higher comprehensive values ​​to ensure the rationality and efficiency of energy transmission response. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an adaptive perception-computation scheduling method for underwater robots according to the present invention. Figure 2 This is a schematic diagram of the structure of an adaptive perception computing scheduling underwater robot system according to the present invention. Detailed Implementation

[0019] 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.

[0020] Example 1: As Figure 1As shown, the present invention provides a technical solution, an adaptive perception computing scheduling method for underwater robots, the underwater robot method comprising the following steps: Step S1: Collect the underwater robot's own status data and task progress to form a data set, generate the corresponding timestamp, and package the data set and timestamp into a node, while embedding three-dimensional position coordinates to form node data; The self-state data includes energy, speed, and mission load, constructing a vector S=[E, V, P]. sen E represents energy, V represents speed, and P represents speed. sen The task load; the task progress data includes the progress of completed tasks and the remaining task duration, constructing a vector T=[A done T left A done For the progress of the completed task, T left The remaining task duration is denoted as (x, y, z) using an inertial navigation system and underwater acoustic positioning technology. The three-dimensional coordinates are obtained using an inertial navigation system and underwater acoustic positioning technology, and are denoted as (x, y, z) with a timestamp t added. Here, x is the horizontal coordinate, y is the vertical coordinate, z is the water depth where the underwater robot is located, and t is the time when the three-dimensional coordinates were obtained.

[0021] Step S2: Construct an underwater robot network, collect node data of all underwater robots, elect a temporary leader node, and form a distributed data set; Step S2-1: Each underwater robot calculates its own comprehensive index in real time. The calculation formula is as follows: ; Where C represents the underwater robot's overall performance indicators, and E represents the underwater robot's current energy. Where α is the task load rate, D is the communication distance of the node, and α, β, and γ are weighting coefficients. A temporary leader node is elected through a consensus mechanism. The temporary leader node is the elected temporary network manager. Non-temporary leader nodes are connected with adjacent underwater robot nodes to form subnets. The subnets are local networks automatically formed by non-leader nodes according to their communication radius. When the energy of the temporary leader node falls below a threshold, a new leader node is elected. Step S2-2: Node data within each subnet is synchronized to the entire underwater robot network through a temporary leader node, forming a distributed data set covering all underwater robots.

[0022] Step S3: Based on the underwater robot's own state data and mission progress in the node data, construct an energy change curve; Step S3-1: Parse the sailing speed sequence {V} from the node data. t Energy change ΔE t =E t -Et−1 The task consumption intensity λ, where the sailing speed sequence is a set of sailing speeds in a discrete time series; a sliding window technique is introduced to calculate the average speed μ within the window. V Energy standard deviation σ E Task completion rate ρ, construct feature vector F t =[μ V ,σ E ,ρ,λ]; Where the energy standard deviation σ E The calculation formula is: ; in the formula Let be the energy value at time t within the window. Let n be the average energy within the window, and n be the number of points within the window. The mean velocity μ within the window V The calculation formula is: ; in the formula Let n be the speed at time t within the window, and n be the number of points within the window. The formula for calculating the task consumption intensity λ is: ; in the formula The task is complete, and t represents time. The formula for calculating the task completion rate ρ is: ; in the formula The task has been completed. For the overall task; Step S3-2: Use the ARIMA model to capture the time-series periodicity of energy changes, and use an LSTM network to learn the nonlinear effects of airspeed and mission load on energy; establish a joint prediction model: ; in To predict the remaining energy, ARIMA is an autoregressive integral moving average model. for LSTM output characteristics , ,..., Let be the node energy at times t-1, t-2, ..., tn; Step S3-3: Construct a time window, Arrange them in chronological order and plot the energy change curves.

[0023] Step S4: Collect the location information of all underwater robots. When the energy of any underwater robot is lower than the preset threshold, send an energy transfer request to the temporary leader node. Step S4-1: Construct a 3D adjacency graph including the depth dimension, mark the neighboring nodes of a given node as neighboring nodes, and obtain all neighboring nodes to form a neighborhood set: ; in Let be the Euclidean distance between nodes i and j, h be the depth tolerance, and z be the depth coordinates of the underwater robot. For communication radius; Step S4-2: Sort the neighboring nodes in the neighborhood set in descending order according to the comprehensive value and store them in the temporary leader node. A node obtains the neighboring node order and sends requests in descending order. The node that sends the request is marked as the requesting node, and the neighboring nodes that receive the request are marked as the responding nodes. The comprehensive value is calculated. The formula is: ; Where w is the weighting coefficient. The current energy of the neighboring nodes. Let be the Euclidean distance between node i and node j. This is a function to indicate task conflicts. Step S4-3: Obtain the remaining energy of the response node itself. Remaining task duration The transferable energy of neighboring nodes is calculated using the following formula: ; in In response to the current energy of the neighboring nodes, In response to the remaining task time of neighboring nodes, Total task duration for the response node; like If >0, then accept the request; if If the value is less than 0, then calculate the transferable energy of the next node.

[0024] Step S5: The underwater robot that receives the energy transfer request calculates the required energy transfer efficiency based on its own location and generates a data packet for energy transfer.

[0025] The shortest path from the response node to the request node is searched in the 3D neighborhood graph. The mapping relationship between transmission rate and water depth is calculated, and the transmission rate is adjusted according to the current water depth. The calculation formula is as follows: ; in The energy transfer rate at the current water depth, denoted as , where k is the surface energy transfer rate, k is the seawater attenuation coefficient, and z is the current water depth. After the transmission is complete, the requesting node sends a data packet containing the amount of energy received and the transmission time to the responding node.

[0026] For example, in an open sea area at a depth of 50 meters, three underwater robots are deployed, each equipped with an inertial navigation system, an underwater acoustic positioning module, and an energy transmission device.

[0027] The underwater robot's own data is recorded with energy E = 80%, speed V = 2 m / s, and sensor load power P. sen =50W, vector S=[0.8, 2, 50]; 30 square meters of task completed, 6 hours of task remaining, vector T=[30, 8]; 3D coordinates are (100, 200, 50), timestamp t=12:00:00; Calculate the comprehensive index: C A =12.3, C B =8.7, C C =17.4, elect robot C as the temporary leader node; robots A and B join the subnet with C as the leader, and the data of each node is synchronized to the global network through C to form a distributed data set.

[0028] Calculate the feature vector F of robot A t =[2.0, 1.5, 0.15, 0.05]; Predicting periodic changes based on historical electricity consumption; LSTM network F t The impact of learning speed and load on energy; predicted remaining energy E=65%.

[0029] E t Arranged by time, an energy change curve for the next 8 hours is generated, showing that A's energy will be lower than the preset threshold in the 6th hour; The combined values ​​of neighboring nodes B and C of robot A are 0.66 and 0.58, respectively. Therefore, the request is sent to B first.

[0030] Calculate the transferable energy based on the state of robot B. =0.36>0, B accepts the request.

[0031] Searching for the shortest path between A and B in a 3D neighborhood graph, with a water depth of z = 45-50 meters, no obstacles, and direct transmission. The transmission rate calculation formula is as follows: =0.41.

[0032] When the transmission is complete, A receives 20% of the energy, which takes about 50 seconds, and sends a data packet to B containing the received amount of 20% and the time taken of 50 seconds.

[0033] Example 2, as Figure 2 As shown, the present invention provides an adaptive perception computing scheduling underwater robot system. The underwater robot system further includes an adaptive perception computing scheduling underwater robot system, wherein the underwater robot system includes a data perception module, a dynamic networking module, an energy prediction module, a cooperation request module, and an energy transmission module. The data sensing module is used to collect the underwater robot's own status data and task progress data, generate timestamps, and embed three-dimensional position coordinates to form node data; the dynamic networking module is used to construct an underwater robot network, collecting all node data to form a distributed dataset; the energy prediction module is used to construct an energy consumption prediction curve based on the underwater robot's operating characteristics and task progress in the node data; the cooperation request module is used to collect the position information of all underwater robots, and send an energy transfer request to adjacent underwater robots when the energy of any underwater robot is lower than a preset threshold; the energy transfer module is used to receive the energy transfer request from the underwater robot, and transfer energy to the requesting underwater robot when there is sufficient remaining energy. The output of the data sensing module is electrically connected to the input of the dynamic networking module; the output of the dynamic networking module is electrically connected to the input of the energy prediction module; the output of the energy prediction module is electrically connected to the input of the cooperation request module; and the output of the cooperation request module is electrically connected to the input of the energy transmission module.

[0034] The data sensing module includes a heterogeneous data acquisition unit and a spatiotemporal encapsulation unit. The heterogeneous data acquisition unit is used to collect the underwater robot's energy, speed, sensor load power, and its own status data, as well as task progress data such as the area of ​​the completed detection area and the remaining task duration. The spatiotemporal encapsulation unit is used to add timestamps to the data set, embed three-dimensional position coordinates, and package the data set and timestamps into node data.

[0035] The dynamic networking module includes a leader election unit and a subnet data synchronization unit. The leader election unit is used for each underwater robot to calculate its own comprehensive indicators in real time, select a temporary leader node through a consensus mechanism, and re-elect a leader node after its energy falls below a threshold. The subnet data synchronization unit is used to synchronize the node data in each subnet to the entire underwater robot network through the leader node, forming a distributed data set. The energy prediction module includes a feature extraction unit and a hybrid modeling unit. The feature extraction unit is used to parse the sailing speed sequence, energy change, and task consumption intensity from the node data, calculate the mean speed, standard deviation of energy, and task completion rate within the window, and construct a feature vector. The hybrid modeling unit is used to capture the time series periodicity of energy changes using an ARIMA model, learn the nonlinear influence of sailing speed and task load on energy through an LSTM network, establish a joint prediction model, and plot the energy change curve.

[0036] The collaboration request module includes a three-dimensional neighborhood unit and a target request unit; the three-dimensional neighborhood unit is used to construct a three-dimensional adjacency graph containing a depth dimension and determine the neighborhood set of each node; the target request unit is used to sort the neighborhood nodes in descending order according to the multi-objective function and send energy transfer requests to the node with the highest function value first. The energy transmission module includes a path planning unit and an adaptive transmission unit. The path planning unit is used to search for the shortest path from the response node to the requesting node in the three-dimensional neighborhood graph, avoiding obstacles, and enabling multi-hop transmission mode when relay nodes exist. The adaptive transmission unit is used to calculate the mapping relationship between transmission rate and water depth, dynamically adjust the transmission rate according to the current water depth, and receive the data packet containing the energy received and transmission time sent by the requesting node after the transmission is completed.

[0037] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An adaptive perception-computation scheduling method for underwater robots, characterized in that: The underwater robot method includes the following steps: Step S1: Collect the underwater robot's own status data and task progress to form a data set, generate the corresponding timestamp, and package the data set and timestamp into a node, while embedding three-dimensional position coordinates to form node data; Step S2: Construct an underwater robot network, collect node data of all underwater robots, elect a temporary leader node, and form a distributed data set; Step S3: Based on the underwater robot's own state data and mission progress in the node data, construct an energy change curve; Step S4: Collect the location information of all underwater robots. When the energy of any underwater robot is lower than the preset threshold, send an energy transfer request to the temporary leader node. Step S5: The underwater robot that receives the energy transfer request calculates the required energy transfer efficiency based on its own location and generates a data packet for energy transfer.

2. The adaptive perception-computation scheduling method for underwater robots according to claim 1, characterized in that: In step S1, the self-state data includes energy, speed, and mission load, constructing a vector S=[E, V, P]. sen E represents energy, V represents speed, and P represents speed. sen The task load; the task progress data includes the progress of completed tasks and the remaining task duration, constructing a vector T=[A done T left A done For the completed task progress, T left The remaining task duration is denoted as (x, y, z) using an inertial navigation system and underwater acoustic positioning technology. The three-dimensional coordinates are obtained using an inertial navigation system and underwater acoustic positioning technology, and are denoted as (x, y, z) with a timestamp t added. Here, x is the horizontal coordinate, y is the vertical coordinate, z is the water depth where the underwater robot is located, and t is the time when the three-dimensional coordinates were obtained.

3. The adaptive perception-computation scheduling method for underwater robots according to claim 2, characterized in that: The specific steps of step S2 are as follows: Step S2-1: Each underwater robot calculates its own comprehensive index in real time. The calculation formula is as follows: ； Where C represents the underwater robot's overall performance indicators, and E represents the underwater robot's current energy. Where α is the task load rate, D is the communication distance of the node, and α, β, and γ are weighting coefficients. A temporary leader node is elected through a consensus mechanism. The temporary leader node is the elected temporary network manager. Non-temporary leader nodes are connected with adjacent underwater robot nodes to form subnets. The subnets are local networks automatically formed by non-leader nodes according to their communication radius. When the energy of the temporary leader node falls below a threshold, a new leader node is elected. Step S2-2: Node data within each subnet is synchronized to the entire underwater robot network through a temporary leader node, forming a distributed data set covering all underwater robots.

4. The adaptive perception-computation scheduling method for underwater robots according to claim 3, characterized in that: The specific steps of step S3 are as follows: Step S3-1: Parse the sailing speed sequence {V} from the node data. t Energy change ΔE t =E t -E t−1 The task consumption intensity λ, where the sailing speed sequence is a set of sailing speeds in a discrete time series; a sliding window technique is introduced to calculate the average speed μ within the window. V Energy standard deviation σ E Task completion rate ρ, construct feature vector F t =[μ V ,σ E ,ρ,λ]; Where the energy standard deviation σ E The calculation formula is: ； in the formula Let be the energy value at time t within the window. Let n be the average energy within the window, and n be the number of points within the window. The mean velocity μ within the window V The calculation formula is: ； in the formula Let n be the speed at time t within the window, and n be the number of points within the window. The formula for calculating the task consumption intensity λ is: ； in the formula The task is complete, and t represents time. The formula for calculating the task completion rate ρ is: ； in the formula The task has been completed. For the overall task; Step S3-2: Use the ARIMA model to capture the time-series periodicity of energy changes, and use an LSTM network to learn the nonlinear effects of airspeed and mission load on energy; establish a joint prediction model: ； in To predict the remaining energy, ARIMA is an autoregressive integral moving average model. for LSTM output characteristics , ,..., Let be the node energy at times t-1, t-2, ..., tn; Step S3-3: Construct a time window, Arrange them in chronological order and plot the energy change curves.

5. The adaptive perception-computation scheduling method for underwater robots according to claim 4, characterized in that: The specific steps of step S4 are as follows: Step S4-1: Construct a 3D adjacency graph including the depth dimension, mark the neighboring nodes of a given node as neighboring nodes, and obtain all neighboring nodes to form a neighborhood set: ； in Let be the Euclidean distance between nodes i and j, h be the depth tolerance, and z be the depth coordinates of the underwater robot. For communication radius; Step S4-2: Sort the neighboring nodes in the neighborhood set in descending order according to the comprehensive value and store them in the temporary leader node. A node obtains the neighboring node order and sends requests in descending order. The node that sends the request is marked as the requesting node, and the neighboring nodes that receive the request are marked as the responding nodes. The comprehensive value is calculated. The formula is: ； Where w is the weighting coefficient. The current energy of the neighboring nodes. Let be the Euclidean distance between node i and node j. This is a function to indicate task conflicts. Step S4-3: Obtain the remaining energy of the response node itself. Remaining task duration The transferable energy of neighboring nodes is calculated using the following formula: ； in In response to the current energy of the neighboring nodes, In response to the remaining task time of neighboring nodes, Total task duration for the response node; like If >0, then accept the request; if If the value is less than 0, then calculate the transferable energy of the next node.

6. The adaptive perception-computation scheduling method for underwater robots according to claim 5, characterized in that: The shortest path from the response node to the request node is searched in the 3D neighborhood graph. The mapping relationship between transmission rate and water depth is calculated, and the transmission rate is adjusted according to the current water depth. The calculation formula is as follows: ； in The energy transfer rate at the current water depth, denoted as , where k is the surface energy transfer rate, k is the seawater attenuation coefficient, and z is the current water depth. After the transmission is complete, the requesting node sends a data packet containing the amount of energy received and the transmission time to the responding node.

7. An adaptive perception-computation scheduling underwater robot system, applied to the adaptive perception-computation scheduling underwater robot method according to any one of claims 1-6, characterized in that: The underwater robot system includes a data sensing module, a dynamic networking module, an energy prediction module, a cooperation request module, and an energy transmission module. The data sensing module is used to collect the underwater robot's own status data and task progress data, generate timestamps and embed three-dimensional position coordinates to form node data; the dynamic networking module is used to build an underwater robot network, collect all node data to form a distributed data set; the energy prediction module is used to construct an energy consumption prediction curve based on the underwater robot's operating characteristics and task progress in the node data. The collaboration request module is used to collect the location information of all underwater robots and send an energy transfer request to the adjacent underwater robots when the energy of any underwater robot is lower than a preset threshold; the energy transfer module is used to receive the energy transfer request from the underwater robot and transfer energy to the requesting underwater robot when the remaining energy is sufficient. The output of the data sensing module is electrically connected to the input of the dynamic networking module; the output of the dynamic networking module is electrically connected to the input of the energy prediction module; the output of the energy prediction module is electrically connected to the input of the cooperation request module; and the output of the cooperation request module is electrically connected to the input of the energy transmission module.

8. An adaptive perception-computation scheduling underwater robot system according to claim 7, characterized in that: The data sensing module includes a heterogeneous data acquisition unit and a spatiotemporal encapsulation unit. The heterogeneous data acquisition unit is used to collect the underwater robot's energy, speed, sensor load power, and its own status data, as well as task progress data such as the area of ​​the completed detection area and the remaining task duration. The spatiotemporal encapsulation unit is used to add timestamps to the data set, embed three-dimensional position coordinates, and package the data set and timestamps into node data.

9. An adaptive perception-computation scheduling underwater robot system according to claim 7, characterized in that: The dynamic networking module includes a leader election unit and a subnet data synchronization unit. The leader election unit is used for each underwater robot to calculate its own comprehensive indicators in real time, select a temporary leader node through a consensus mechanism, and re-elect a leader node after its energy falls below a threshold. The subnet data synchronization unit is used to synchronize the node data in each subnet to the entire underwater robot network through the leader node, forming a distributed data set. The energy prediction module includes a feature extraction unit and a hybrid modeling unit; the feature extraction unit is used to parse the sailing speed sequence, energy change, and task consumption intensity from the node data, calculate the mean speed, energy standard deviation, and task completion rate within the window, and construct a feature vector. The hybrid modeling unit is used to capture the time-series periodicity of energy changes using an ARIMA model, learn the nonlinear effects of sailing speed and mission load on energy through an LSTM network, establish a joint prediction model, and plot the energy change curve.

10. An adaptive perception-computation scheduling underwater robot system according to claim 7, characterized in that: The collaboration request module includes a three-dimensional neighborhood unit and a target request unit; the three-dimensional neighborhood unit is used to construct a three-dimensional adjacency graph containing a depth dimension and determine the neighborhood set of each node; the target request unit is used to sort the neighborhood nodes in descending order according to the multi-objective function and send energy transfer requests to the node with the highest function value first. The energy transmission module includes a path planning unit and an adaptive transmission unit. The path planning unit is used to search for the shortest path from the response node to the requesting node in the three-dimensional neighborhood graph, avoiding obstacles, and enabling multi-hop transmission mode when relay nodes exist. The adaptive transmission unit is used to calculate the mapping relationship between transmission rate and water depth, dynamically adjust the transmission rate according to the current water depth, and receive the data packet containing the energy received and transmission time sent by the requesting node after the transmission is completed.

Images