A dynamic task allocation and path planning method for unmanned vehicle cluster

Abstract

The application discloses a kind of unmanned vehicle cluster dynamic task allocation and path planning method, comprising: establishing a virtual data field representing operating environment;Task target is defined as virtual information source capable of generating scalar concentration gradient;Make each unmanned vehicle in cluster independently calculate its motion vector based on its local perceived gradient;And according to the information interaction frequency between unmanned vehicle, dynamically adjust the weight of communication link connecting them.The present application introduces the aquatic mycelium network core constructed based on mycelium network growth and information transmission model, converts task allocation into pheromone gradient optimization problem, and uses communication interaction history as the basis for network topology adaptive growth, thereby converting communication uncertainty into the key dimension driving the cluster to realize self-organization, self-healing and emergent intelligence, significantly improving the task execution efficiency, system resilience and autonomous collaboration ability of unmanned vehicle cluster in communication limited, high dynamic environment.

Description

Technical Field

[0001] This invention relates to the field of unmanned vehicle swarm control technology, and more specifically to a dynamic task allocation and adaptive path planning method and system for unmanned surface vessel swarms in communication-constrained, highly dynamic, and large-scale aquatic environments. Background Technology

[0002] Unmanned surface vessel (USV) swarms, with their advantages of wide coverage, low operating costs, and strong risk avoidance capabilities, have shown great application potential in fields such as marine environmental monitoring, waterway security patrols, emergency search and rescue, and underwater topographic mapping.

[0003] However, achieving efficient and robust collaborative operations in large-scale unmanned surface vessel swarms has always faced a series of deep-rooted technical challenges. Among them, the three core contradictions between swarm autonomy and task determinism, individual efficiency and system resilience, and the high intelligence of the swarm and the unreliability of underwater communication channels are particularly prominent.

[0004] In existing technologies, for example, a surface unmanned vessel swarm control system is disclosed in patent application CN117519163A published by the Chinese Patent Office. This patent is a scheme that uploads massive amounts of terminal data to a cloud platform for centralized analysis, and then the central controller issues optimization instructions. However, such a centralized architecture has inherent technical defects in the application of unmanned swarm control due to data link delays, bandwidth bottlenecks, and central processing bottlenecks, resulting in delayed control response. Especially in environments with poor communication conditions, such as waterways, the loss of the central node or communication interruption will directly lead to the paralysis of the entire swarm system.

[0005] To overcome the drawbacks of centralized architectures, those skilled in the art have developed various distributed control algorithms, such as negotiation algorithms based on auction mechanisms and formation control algorithms based on consensus theory. However, existing distributed solutions generally suffer from technical problems such as decision failure or system collapse when communication is limited due to over-reliance on global information consistency. For example, the efficiency and success rate of task allocation in consensus-based auction protocols heavily depend on the reliable transmission of key information such as broadcasting, bidding, and winning bids. In underwater acoustic communication environments with high packet loss rates, the protocol may fall into an endless cycle of retransmission and confirmation, or even fail to achieve consensus, resulting in task allocation failure and an inability to maintain operational effectiveness in dynamic and adversarial aquatic environments.

[0006] The root of the aforementioned dilemma lies in a deep-seated technological bias developed by those skilled in the art within the theoretical framework of classical cybernetics and computer network protocols. This bias holds that swarm intelligence must be designed and implemented through precise mathematical modeling and reliable, explicit information exchange. This bias leads research and development to consistently focus on overcoming communication unreliability, such as developing more robust physical layer modulation and demodulation techniques, designing more complex network layer routing protocols, or constructing more accurate environmental prediction models to reduce communication requirements. These methods are essentially still improvements within the framework of command-execution and explicit communication, treating the dynamic uncertainties of communication links as system noise and control obstacles that need to be suppressed or eliminated, rather than fundamentally utilizing the inherent characteristics of communication links. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art, this invention provides a control system and method for aquatic mycelial network clusters that can achieve large-scale unmanned cluster resilience, high efficiency and controllable collaboration in unreliable communication environments, thereby solving the problems of control response lag, decision failure and insufficient system resilience in high dynamic and communication-limited environments.

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

[0009] A method for dynamic task allocation and path planning of unmanned surface vessel (USV) swarms includes the following steps:

[0010] Step S1: Establish a virtual data field representing the operating environment, and define the task objective or environmental event in the field as a virtual information source capable of generating scalar concentration gradients;

[0011] Step S2: Each unmanned surface vessel in the cluster independently calculates the velocity and heading vector driving its own motion based on the local gradient of the virtual data field it perceives locally.

[0012] Step S3: Dynamically adjust the logical weight of the communication link connecting any two unmanned vessels. The weight is mainly determined by the information interaction frequency and historical data between the two unmanned vessels. High-frequency and effective interaction will strengthen the link weight, while long-term idle or inefficient links will be weakened or even logically disconnected.

[0013] Step S4, the macro-level behavior of the cluster, such as global path planning, network topology, and coordinated response to tasks, emerges from the simple, local, and decentralized individual rules mentioned above from the bottom up.

[0014] As a further preferred embodiment of the present invention, the dynamic adjustment of the weight of the communication link includes: increasing the weight of the link when the information interaction frequency exceeds a preset strengthening threshold; and decreasing the weight of the link when the information interaction frequency is lower than a preset weakening threshold.

[0015] As a further preferred embodiment of the present invention, reducing the weight of the link includes at least one of the following: reducing the priority of the link in the data routing protocol, or logically disconnecting the link.

[0016] As a further preferred embodiment of the present invention, the overall path planning and network topology of the cluster emerge from the independent motion of each unmanned surface vessel based on local gradient calculation and the dynamically adjusted communication link weights.

[0017] As a further preferred embodiment of the present invention, it further includes: receiving an operator instruction and setting the position, intensity, or type of the virtual information source in the virtual data field according to the instruction, wherein the type includes at least attractive information sources and repulsive information sources.

[0018] As a further preferred embodiment of the present invention, each unmanned surface vessel is modeled as a tip of a mycelial network, the communication link is modeled as a mycelium connecting the tip, and the mission objective is modeled as a nutrient source, thereby transforming the cluster control problem into a problem of simulating biological network growth and resource optimization.

[0019] A dynamic task allocation and path planning system for an unmanned surface vessel (USV) swarm, the system comprising:

[0020] Multiple unmanned surface vessels (USVs) are provided, each equipped with a communication module and an onboard processor. The onboard processor is configured to perform the following operations: receive information from a virtual data field representing the operating environment and containing one or more virtual information sources; generate drive commands to control the movement of the USV based on the local gradient of the virtual data field at the USV's own position; and dynamically adjust the state parameters of the communication links with other USVs based on historical data of information interaction with other USVs, wherein the adjustment of the state parameters takes precedence over or combines with physical layer signal indicators.

[0021] As a further preferred embodiment of the present invention, the onboard processor is configured to dynamically adjust the state parameters in the following manner: when the information interaction frequency with another unmanned surface vessel meets an increasing condition, the priority of the corresponding communication link in the routing table is increased; and when the information interaction frequency meets an decreasing condition, the priority of the corresponding communication link is decreased or it is marked as inactive.

[0022] As a further preferred embodiment of the present invention, it also includes a remote control terminal, which is configured to allow an operator to create, delete or modify the virtual information source in the virtual data field in order to macroscopically guide the overall behavior of the unmanned surface vessel cluster, rather than directly controlling individual unmanned surface vessels.

[0023] The beneficial effects of this invention are as follows:

[0024] This invention introduces a mycelial network growth and information transmission model to construct a cluster control architecture based on the aquatic mycelial paradigm. This transforms the complex task allocation problem into a simple, local gradient climbing problem, thereby significantly improving the task execution efficiency, system resilience, and autonomous coordination capabilities of unmanned surface vessel (USV) clusters in communication-constrained and highly dynamic environments. It also solves the decision failure problem caused by over-reliance on communication reliability in existing technologies.

[0025] This invention defines each unmanned surface vessel (USV) as the tip of a mycelium network, considers communication links as the mycelium connecting nodes, and abstracts mission objectives as nutrient sources. This integrates path planning and network topology maintenance for the USV swarm. When multiple USVs move toward the same nutrient source, their communication links are strengthened due to frequent information exchange, forming an efficient backbone. Conversely, they are weakened, thereby achieving self-healing capabilities at the system level and ensuring mission continuity in the event of individual loss.

[0026] This invention adopts a decentralized control strategy. The overall path planning and network topology of the cluster emerge from the independent motion of each unmanned surface vessel based on local gradient calculation and the dynamically adjusted communication link weights. This eliminates the dependence on a central control unit or global information consistency, and transforms the uncertainty of communication into a key dimension that drives the cluster to achieve self-organization and self-healing. It effectively solves the technical problem of system paralysis caused by the failure of the central node in the prior art. Attached Figure Description

[0027] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0028] Figure 1 A schematic diagram of the modular architecture of an aquatic mycelium network cluster control system provided in an embodiment of the present invention;

[0029] Figure 2 This is an overall flowchart of a method for dynamic task allocation and path planning of unmanned surface vessel swarms provided in an embodiment of the present invention;

[0030] Figure 3 This is a schematic diagram of a virtual pheromone concentration field and its gradient guiding the motion of an unmanned surface vessel in one embodiment of the present invention;

[0031] Figure 4 This is a schematic diagram illustrating the dynamic strengthening and weakening of a communication link based on the interaction frequency in one embodiment of the present invention. Detailed Implementation

[0032] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings.

[0033] Example 1

[0034] This embodiment describes in detail a rapid response and pollution source detection system applied in a high-risk nuclear leakage environment. In this embodiment, the emergency response task process follows a macro-guided-autonomous-emergent-network-reinforced model.

[0035] First, the operator delineates an initial circular search area with a radius of 5.2 kilometers on the electronic nautical chart using a Human-Machine Interface (HMI-MP) for task planning and interaction. The HMI-MP module encodes the operation command into a Command Packet, where "intensity" is set to 0.25, "profile" to "ATTRACT", and "ttl" to 7200.0 seconds. This packet is then broadcast to the cluster via a boat-to-shore data link. The aquatic mycelium network core AMNC deployed within the onboard computing unit of each unmanned surface vessel in the cluster receives this external pheromone source. .

[0036] When any unmanned surface vessel in the cluster detects that the gamma-ray spectrometer in its multimodal sensing module (MMP) exceeds a preset threshold for the count rate of cesium-137 characteristic gamma rays in the water, the MMP immediately sends an Event_Trigger internal process call to its local AMNC module. The event_type is RADIONUCLIDE_SPIKE_DETECTED, and the significance parameter is quantized to 0.98. Upon receiving this trigger, the AMNC immediately models its own position as an endogenous pheromone source with an intensity of 0.98 during the solution of the cluster guidance field evolution equation.

[0037] The pheromone field management submodule within the AMNC module of the nearby unmanned surface vessel is used in the next iteration calculation, specifically, the iteration period in the iteration calculation is 100.2 milliseconds.

[0038] Sensing a rapidly changing local pheromone concentration gradient A Control_Vector pointing towards the pheromone source is generated based on the gradient, and this Control_Vector is passed to the Physical Execution Platform (PEP). Simultaneously, due to the surge in information interaction and stable communication quality between the two unmanned surface vessels, the mycelial intensity between them... According to the network topology adaptive growth and apoptosis kinetic model, it begins to grow exponentially. This process spreads in the cluster, eventually causing the entire cluster to autonomously emerge from a decentralized search state and autonomously converge at a high speed toward the pheromone source location.

[0039] The Human-Machine Interface (HMI-MP) is used to provide operators with task planning and situational awareness functions. It is specifically deployed on a Panasonic Toughbook CF-33 military-grade ruggedized laptop, and the command and control software running on it is developed based on the Qt 5.15.2 framework.

[0040] The multimodal perception module (MMP) of each unmanned surface vessel is used to perceive the environment and trigger endogenous pheromone sources. Its core sensor is an Ortec an Gammastream-P underwater high-purity germanium HPGe gamma-ray spectrometer. The preprocessing software module built into the MMP is used to smooth the raw energy spectrum and find peaks. When the count rate of the identified cesium-137 characteristic peaks exceeds a dynamic threshold, an event is triggered. The value of significance is precisely quantified through a logarithmic mapping function.

[0041] The onboard computing unit is used to provide computing resources for the AMNC and other onboard software, and it adopts the NVIDIA Jetson AGX Xavier embedded computing platform.

[0042] The core of the aquatic mycelial network runs as a multi-threaded C++17 application, internally divided into three key sub-modules:

[0043] The pheromone field management submodule maintains a two-dimensional floating-point grid representing the virtual pheromone field in memory and iteratively solves the cluster guidance field evolution equation on the graphics processor using the finite difference method. The virtual diffusion coefficient is set to 0.25 m² / s, and the pheromone decay rate is set to... The evolution equation of the cluster guidance field is:

[0044]

[0045] The chemotactic motion control submodule is used to implement the hyphal tip chemotactic motion model. It first uses the Sobel operator to calculate the gradient of the pheromone concentration grid, and then combines the chemotactic sensitivity coefficient of 1.25, the random exploration weight coefficient of 0.25, and the path enhancement function to calculate the target velocity vector. The hyphal tip chemotactic motion model is as follows:

[0046]

[0047] The network topology dynamics submodule is used to execute the network topology adaptive growth and apoptosis dynamics model. It implements a distributed hash table based on the Kademlia protocol to maintain neighbor node information and hyphal strength. The hyphal growth rate constant is set to 0.65, and the hyphal apoptosis rate constant is set to... The network topology adaptive growth and apoptosis kinetic model is as follows:

[0048]

[0049] The physical execution platform is used to receive control vector commands from the onboard computing unit and directly control the underlying power and servo motors. It uses BlueRobotics T200 thrusters and Pixhawk 4 flight controllers. The underlying communication hardware is used to realize communication between the submarine and between the submarine and the shore. Its inter-submarine communication uses EvoLogics S2C R 7 / 17 underwater acoustic communication device.

[0050] Example 2

[0051] As a preferred embodiment, the present invention can also be applied to wide-area maritime search and rescue scenarios. The core difference between this embodiment and the first specific embodiment is that a predictive nutrient source generation submodule is further integrated into the core module of the aquatic mycelium network.

[0052] The predictive nutrient source generation submodule functions as follows: when at least three unmanned surface vessels (USVs) in the cluster discover valid clues and generate endogenous pheromone sources, it collects the location information of the discovered clues and inputs it as nodes into a pre-trained graph convolutional network model. The graph convolutional network model is used to predict a heatmap of the probability distribution of other potential targets based on the discrete discovery points. This heatmap is used by the predictive nutrient source generation submodule to dynamically modify the external pheromone source term in the cluster guidance field evolution equation, thereby superimposing a data-driven, dynamically updated predictive nutrient background field on top of the original macroscopic guidance field, guiding USVs that have not yet discovered targets to move to areas with higher probability in advance.

[0053] Furthermore, to adapt to this implementation method, the technical implementation of each system module has been adjusted as follows.

[0054] The core sensor of the multimodal sensing module has been replaced with a Teledyne FLIR SeaFLIR 280-HD photoelectric infrared turret, and a target detection model trained based on YOLOv7 has been integrated.

[0055] The onboard computing unit has been upgraded to NVIDIA Jetson AGX Orin to provide additional computing resources for the graph convolutional network model and the YOLOv7 model.

[0056] The core of the aquatic mycelial network is newly equipped with a predictive nutrient source generation submodule, and in the chemotactic motion control submodule, the chemotactic sensitivity coefficient is increased to 1.50; in the network topology dynamics submodule, the mycelial growth rate constant is fine-tuned to 0.70.

[0057] Example 3

[0058] As a preferred embodiment, the present invention can also be applied to underwater topographic mapping and resource exploration scenarios. The core innovation of this embodiment lies in modifying the chemotactic movement model of the hyphal tips by introducing an array maintenance force term. The modified model is as follows:

[0059]

[0060] in, The array maintenance force weighting coefficient is used to calculate a resultant force based on the relative position of UAV i with its neighboring UAVs and the hyphal strength. Specifically, this resultant force is a function weighted by hyphal strength and based on the Lennard-Jones potential. This function generates a repulsive force on nearby neighbors and an attractive force on distant neighbors, within a preset ideal distance. When equilibrium is reached, the final motion decision of the unmanned surface vessel is the result of the dynamic balance between the chemotactic force, the array maintenance force, and the random exploration force, thereby enabling the unmanned surface vessel swarm to autonomously form and maintain a precise observation array with a specific geometric configuration for a long period of time.

[0061] Furthermore, to adapt to this implementation method, the technical implementation of each system module has been adjusted as follows.

[0062] The control software of the human-computer interaction and task planning interface has added an array planning module to generate an external pheromone source containing periodic attraction lines and repulsion boundaries.

[0063] In the pheromone field management submodule of the core of the aquatic mycelium network, the pheromone decay rate is reduced to The virtual diffusion coefficient was reduced to 0.10 m² / s. In the chemotactic motion control submodule, the calculation of the array maintenance force term was introduced, and the array maintenance force weight was set to 0.80, the chemotactic force was reduced to 0.80, and the random exploration weight was reduced to 0.02. In the network topology dynamics submodule, the hyphal apoptosis rate was reduced to... In order to maintain a stable network topology that is highly consistent with the physical array shape.

[0064] To verify the technical effect of the present invention, a series of simulation comparison experiments were conducted. The experimental scenario simulated a cluster of 20 unmanned surface vessels performing wide-area search and target approach missions in a 100-square-kilometer sea area with multiple communication-restricted areas. The average communication packet loss rate in the sea area was set to 50%.

[0065] The control group adopted a distributed consensus auction protocol, experimental group 1 adopted the basic aquatic mycelium network architecture described in this invention, and experimental group 2, based on experimental group 1, introduced a predictive nutrient source guidance mechanism. The underlying cluster collaboration and task allocation algorithm used by the three groups is the only variable.

[0066] Performance Comparison Experiment Report of Unmanned Surface Vessel Swarms in Complex Communication Environments

[0067]

[0068] The experimental data show that, compared with the control group, experimental groups 1 and 2, which adopt the aquatic mycelium network architecture of the present invention, have achieved significant improvements in key performance indicators such as average task completion time, total navigation redundancy rate, task recovery time after single point of failure, task success rate under high communication packet loss rate, and average network communication load.

[0069] Specifically, the average task completion time of Experiment Group 1 was shortened by 29.6%, its total navigation redundancy rate decreased from 34.6% to 18.2%; its single-point fault recovery time was 16.1% of that of the control group; its task success rate increased to 94.8% in the high packet loss environment; and its average network communication load decreased by 60.7%.

[0070] Furthermore, based on the high resilience and high communication efficiency of Experimental Group 1, Experimental Group 2, by introducing the predictive guidance mechanism, shortened its average task completion time by 40.6% compared with the control group, and reduced its navigation redundancy rate to 11.8%.

[0071] The above description is merely a preferred embodiment of the present invention, and therefore should not be construed as limiting the scope of the present invention. Any simple equivalent changes and modifications made in accordance with the scope of the patent application and the contents of the specification of the present invention should still fall within the scope of the patent of the present invention.

Claims

1. A method for dynamic task allocation and path planning in an unmanned surface vessel (USV) swarm, characterized in that, Includes the following steps: In at least one computing processor, a virtual data field representing the operating environment is established; In the virtual data field, one or more task objectives are defined as virtual information sources capable of generating scalar concentration gradients; The motion vector of each unmanned surface vessel is independently calculated based on the scalar concentration gradient sensed locally by each unmanned surface vessel. And dynamically adjust the weight of the communication link connecting any two unmanned vessels based on the information exchange frequency between the unmanned vessels, wherein the weight adjustment is independent of or combined with the physical signal quality of the link; The method further includes: by executing independent motion of each unmanned surface vessel based on local gradient calculation, and combining it with the dynamic adjustment of the communication link weight, the overall path planning of the cluster and the self-organizing construction of the network topology are realized; Each unmanned surface vessel (USV) is modeled as the tip of a mycelial network, the communication link is modeled as the mycelium connecting the tip, and the mission objective is modeled as a nutrient source. This transforms the dynamic task allocation and path planning problem of the USV swarm into a problem of simulating biological network growth and resource optimization.

2. The method for dynamic task allocation and path planning of an unmanned surface vessel swarm according to claim 1, characterized in that, The dynamic adjustment of the weight of the communication link includes: increasing the weight of the link when the information interaction frequency exceeds a preset strengthening threshold; and decreasing the weight of the link when the information interaction frequency is lower than a preset weakening threshold.

3. The method for dynamic task allocation and path planning of an unmanned surface vessel swarm according to claim 2, characterized in that, The reduction of the weight of the link includes at least one of the following: reducing the priority of the link in the data routing protocol, or logically disconnecting the link.

4. The method for dynamic task allocation and path planning of an unmanned surface vessel swarm according to claim 1, characterized in that, Also includes: Receive an operator instruction and set the location, intensity, or type of the virtual information source in the virtual data field according to the instruction, wherein the type includes at least attractive information sources and repulsive information sources.

5. A dynamic task allocation and path planning system for an unmanned surface vessel swarm, characterized in that, The system includes: Multiple unmanned surface vessels (USVs), each equipped with a communication module and an onboard processor; wherein the onboard processor is configured to perform the following operations: Receive information from a virtual data field that represents the operating environment and contains one or more virtual information sources; Based on the local gradient of the virtual data field of the unmanned surface vessel's own position, drive commands are generated to control the movement of the unmanned surface vessel; And based on historical data of information interaction with other unmanned surface vessels, dynamically adjust the state parameters of the communication link with the other unmanned surface vessels, wherein the adjustment of the state parameters takes precedence over or is combined with physical layer signal indicators. The system is configured to achieve overall path planning and self-organizing construction of the network topology by executing independent motions of each unmanned surface vessel based on local gradient calculations and combining them with dynamic adjustments to the communication link state parameters. Logically, the system models each unmanned surface vessel (USV) as the tip of a mycelial network, the communication link as the mycelium connecting the tip, and the virtual information source as a nutrient source, thereby transforming the dynamic task allocation and path planning problem of the USV swarm into a problem of simulating biological network growth and resource optimization.

6. The dynamic task allocation and path planning system for an unmanned surface vessel swarm according to claim 5, characterized in that, The onboard processor is configured to dynamically adjust the state parameters in the following ways: when the information interaction frequency with another unmanned surface vessel meets an increasing condition, the priority of the corresponding communication link in the routing table is increased; and when the information interaction frequency meets an decreasing condition, the priority of the corresponding communication link is decreased or it is marked as inactive.

7. The dynamic task allocation and path planning system for an unmanned surface vessel swarm according to claim 5, characterized in that, It also includes a remote control terminal configured to allow an operator to create, delete, or modify the virtual information source in the virtual data field to macroscopically guide the overall behavior of the unmanned surface vessel swarm, rather than directly controlling individual unmanned surface vessels.

Images