Under-actuated unmanned ship distributed multi-task allocation method for sea rescue

Abstract

The invention discloses an under-actuated unmanned ship distributed multi-task distribution method for sea rescue, and the method comprises the following steps: carrying out the clustering and grouping through employing a nearest neighbor clustering algorithm of a greedy strategy based on the position information of a person falling into water at sea, building a task cluster package, and carrying out the clustering and grouping of the task cluster package; constructing a target function of a target allocation problem of the rescue task of the unmanned ship at sea by taking the shortest search and rescue distance as a constraint condition, and taking the optimal time satisfaction as a rescue target; and solving the target function of the target allocation problem of the rescue tasks of the maritime unmanned ships by adopting a clustering consensus consistency packet algorithm to realize task allocation of the plurality of maritime search and rescue unmanned ships carrying the plurality of search and rescue unmanned ships. Compared with an existing unmanned ship task allocation method, the allocation method designed by the invention fully considers the kinematics constraint problem caused by the underactuation characteristic of the unmanned ship, so that the allocation result is closer to the path distance from the unmanned ship to the target point, and the task performability is improved.

Description

Technical Field

[0001] The present invention relates to the field of unmanned boat cluster task allocation, and in particular to a distributed multi-task allocation method for under-actuated unmanned boats for maritime rescue. Background Art

[0002] With the advancement of science and technology, unmanned aerial vehicles (UAVs) are increasingly equipped with collaborative capabilities, intelligent autonomy, and advanced operational capabilities, enabling them to perform dangerous tasks beyond the physiological limits of the human body. Currently, UAVs are being extensively researched for in-depth exploration and application of marine resources, being used for tasks such as environmental monitoring, maritime patrols, island mapping, and maritime rescue. Maritime rescue is a critical line of defense for ensuring maritime safety, and its efficiency is crucial to the safety of personnel at sea. Therefore, research on methods for allocating rescue tasks to UAVs can help improve their intelligence and thus better ensure overall maritime safety.

[0003] Many relevant research results have been achieved in the research on the multi-task allocation problem of unmanned boats at home and abroad. Li Yan proposed an improved auction algorithm based on ship benefits and cost attributes for the allocation of multi-agent maritime search tasks. Nunes proposed the use of sequential single-item auctions for allocation, based on the ability of each unmanned boat to maintain its own independent local time network, which effectively reduced the communication cost. Ye proposed an extended CBBA with task coupling constraints for the multi-task allocation problem with task coupling constraints in heterogeneous multi-unmanned boat systems. Chen proposed a local replanning algorithm based on CBBA on how to achieve task redistribution in a dynamic environment. Gao Cheng proposed task package construction and conflict resolution rules in the extended consistency package algorithm to eliminate deadlock in the execution of distributed tasks. Choi HL, based on the auction algorithm, relaxed the limit on the number of unmanned boats for each task, added cooperation constraints for unmanned boats, and proposed a new scoring mechanism for tasks that require unmanned boats to cooperate. These algorithms are crucial for target allocation in the process of maritime search and rescue. However, the existing unmanned boat target allocation methods still have the following shortcomings:

[0004] First, with existing centralized task allocation strategies, agents rely heavily on the computing power and communication capabilities of central nodes. A central node failure or communication interruption severely impacts the entire task and prevents its completion. Furthermore, centralized approaches are slow to respond to dynamic environmental changes, potentially reducing system adaptability and leading to potentially adverse consequences.

[0005] Second, existing consistent distributed auction task allocation strategies only consider the reward value of a single task for an agent, ignoring the coupling between agents performing multiple tasks. Existing CBBA distributed auction task allocation strategies allow agents to converge on conflict-free allocations, but they require extensive computation to process bids and frequent communication to ensure that the allocations converge to the global optimum.

[0006] Third, most existing multi-agent task allocations only consider the straight-line path distance for fully driven unmanned boats, which cannot solve the multi-agent task allocation problem with kinematic constraints. Summary of the Invention

[0007] In order to solve the above problems, the technical solution adopted by the present invention is: a distributed multi-task allocation method for under-actuated unmanned boats for maritime rescue, comprising the following steps:

[0008] Based on the location information of people who fall into the water at sea, a greedy nearest neighbor clustering algorithm is used to perform clustering and task clustering. The target allocation problem of the unmanned boat rescue mission at sea is constructed with the shortest search and rescue distance as the constraint condition, and the optimal time satisfaction is set as the rescue goal.

[0009] The consistency bag algorithm of clustering consensus is used to solve the objective function of the target allocation problem of maritime unmanned boat rescue missions, and the task allocation of multiple catamaran search and rescue unmanned boats carrying multiple small search and rescue unmanned boats is realized.

[0010] Furthermore, the objective function of the target allocation problem of the unmanned boat rescue mission at sea is constructed as follows:

[0011] Because N u The set of catamaran unmanned boats A = {α i |i=1,2…N u}, by N t The set of tasks to be executed is set as T = {β j |j=1,2…N t The decision-making goal of the allocation is to achieve a feasible and conflict-free task allocation to maximize the objective function. The mathematical expressions constructed are shown in formulas (1)-(4):

[0012]

[0013] where x ij ∈{0,1} is the decision vector, when x ij =1, it means that the task point S j Mission Beta j Assigned to catamaran unmanned boat α i ; Represents the catamaran unmanned boat α iThe maximum number of tasks allowed to be executed; Represents the catamaran unmanned boat α i The task sequence to be executed, if the catamaran unmanned boat α i The kth task executed is β j , then p i The kth element of is j, if the catamaran unmanned boat α i There is no task to execute, then the ordered task set is

[0014] The profit function c in formula (1) ij (x ij ,p i ) is a decision vector x ij and task sequence p i A non-negative function of the catamaran unmanned boat α i According to p i Complete Mission β j The benefit obtained when x ij =0, c ij (x ij ,p i )=0;

[0015] Formula (2) represents the catamaran unmanned boat α i The number of tasks performed does not exceed

[0016] Formula (3) indicates that each mission can only be performed by one catamaran unmanned boat;

[0017] Formula (4) indicates that the mission volume completed by all catamaran unmanned boats cannot exceed the total mission volume.

[0018] Furthermore, the process of constructing the optimal goal of time satisfaction is as follows:

[0019] The starting point of time is 0. According to the change of the survival probability of the trapped person over water with time, it is believed that the survival probability of the person in water decreases exponentially with time. The survival probability function of a single individual is formula (5):

[0020]

[0021] Where λ<1 means the number of people falling into the water is β j Time discount factor, t ij (p i ) is a catamaran unmanned boat α i Along the path β j Arrival at the location of the person who fell into the water j The time required, is the current danger level of the drowning person, and the situation is divided into three levels: "low risk 3", "medium risk 2" and "high risk 1". The higher the survival probability, the higher the satisfaction. The time satisfaction of the drowning person regarding the rescue time is considered to be positively correlated with the survival probability of the drowning person. In the case of multiple drowning people, the constraint condition of time satisfaction is formula (6):

[0022]

[0023] Where R j Represents task β j The initial satisfaction value.

[0024] Furthermore, the process of constructing the constraint condition of the shortest search and rescue distance is as follows:

[0025] The Dubins path algorithm is used to calculate the shortest path that satisfies the kinematic constraints of the catamaran unmanned boat from the state (x1, y1, θ1) (x1, y1 represent the horizontal and vertical coordinates of the catamaran unmanned boat at the position, and θ1 represents the heading of the catamaran unmanned boat at the position) to the state (x2, y2, θ2) (x2, y2 represent the horizontal and vertical coordinates of the catamaran unmanned boat at the position, and θ2 represents the heading of the catamaran unmanned boat at the position). When the minimum turning radius is known, for the multi-task assignment of each catamaran unmanned boat, the route to the first target point can be directly calculated using the Dubins path algorithm through the initial position of the catamaran unmanned boat and the position of the target. The subsequent target points need to be based on the position of the catamaran unmanned boat when it reaches the first target point and the position of the second target point as the input of this Dubins path. The center O2 of the turning circle of the catamaran unmanned boat passes through the initial position of the catamaran unmanned boat and the minimum turning radius, and is obtained:

[0026]

[0027] The coordinates of tangent points B and C are:

[0028]

[0029] The optimal trajectory of the catamaran unmanned boat to reach the target point 1 is composed of arc AB, straight line BC and arc CD, and the length is:

[0030]

[0031] The length of the chord l corresponding to the central angle ε is:

[0032]

[0033] Then the position coordinates of the catamaran unmanned boat when it reaches the first target point are D(x D ,y D )for:

[0034]

[0035] The heading angle θ2 when the catamaran unmanned boat reaches the first target point is.

[0036]

[0037] Furthermore, the clustering grouping process of the greedy strategy nearest neighbor clustering algorithm is as follows:

[0038] In the initialization stage, for each sample point, the number of adjacent points within the distance threshold is calculated;

[0039] Then, the sample points with the most adjacent points are selected as cluster centers. These cluster centers will serve as representative locations of subsequent task clusters and can more accurately represent the core of the cluster area.

[0040] The positions of people who fall into the water that are closer to the cluster center are grouped into the same task cluster. By selecting the cluster center, the positions of people who fall into the water within the task cluster can better reflect the actual clustering situation;

[0041] The center of each task cluster, i.e., the cluster center, will serve as the target location of the task cluster; the unmanned boat will use this target location as a guide to proceed to the rescue;

[0042] When the unmanned boat reaches the target location, which is the center of the cluster, multiple small unmanned boats will be released at the same time to carry out the rescue mission.

[0043] Furthermore: the task cluster package construction process is as follows:

[0044] Defining an index set During the mission cluster package construction phase, five bidding lists are created based on the mission information obtained by each catamaran unmanned boat through wireless communication equipment, and the five bidding list information are defined:

[0045] 1). Task cluster list b i , Represents the task cluster set α obtained by the bidding of the catamaran unmanned boat i ,The assigned task clusters are sorted according to the order of joining the task clusters;

[0046] 2). Task cluster sequence p i , Indicates the allocation to the catamaran unmanned boat α i The execution order of the task clusters;

[0047] 3). Task cluster completion list pn i , Indicates the number of people rescued from drowning corresponding to each cluster;

[0048] 4). List of winning bidders i , There is a catamaran unmanned boat ij ∈y i (j=1,2…N t ), y ij Represents the catamaran unmanned boat α i In local state awareness, task cluster σ j The highest bid;

[0049] 5). Winners list i , That is, there exists z ij ∈z i (j=1,2…N t ) indicates a bid of y ij The catamaran unmanned boat,

[0050] In the consensus pack algorithm based on cluster consensus, the value c of the CBBA bid is ij The score is calculated based on the marginal gain decreasing function; the marginal gain decreasing function indicates that the score value of the task cluster will not increase as a new task cluster is added to the existing task cluster set, that is, formula (14):

[0051]

[0052] Among them, b i Represents the catamaran unmanned boat α i The current task cluster set represents any task cluster added to the tail of the current task cluster set, c ij Indicates greedily adding task clusters according to the score value, where the task clusters with higher scores are added first, c ij (b i ) is as shown in equations (15)-(17):

[0053]

[0054] In formula (15), p i Indicates that it is assigned to a catamaran unmanned boat i The task cluster sequence of formula (16) represents any execution task cluster sequence p i The total score of t ij Indicates a catamaran unmanned boat i Execution task cluster σ j time; Formula (17) represents the catamaran unmanned boat α i Execution task cluster σ j Benefits obtained.

[0055] Furthermore, the consistency package algorithm using cluster consensus solves the objective function of the target allocation question of the maritime unmanned boat rescue mission, and the process of realizing the task allocation of multiple maritime search and rescue unmanned boats equipped with multiple search and rescue unmanned boats is as follows:

[0056] Assume that each catamaran unmanned boat adopts a synchronous communication mechanism in the consistency stage and updates the three vectors (y i , z i and timestamps i ) information to obtain the latest allocation information, based on communication exchanges, catamaran unmanned boat a i For task cluster σ j There are three possible operational definitions, as shown in Equation (18), and the decision rules are as follows:

[0057]

[0058]

[0059]

[0060] The UAVs use the above approach to determine whether it should update, reset, or leave the bid by comparing what receiver i and sender k know about task j, and the last time each UAV received communication from the UAV they believe is assigned to task j, communicating and negotiating until they converge to a conflict-free solution where all tasks have a unique assigner and there is no overlapping or conflicting UAVs.

[0061] The present invention provides a distributed multi-task allocation method for under-actuated unmanned boats for maritime rescue, which fully considers the urgency of the rescue process and the dynamic changes of the environment, explores the problem of maritime rescue task allocation, takes time satisfaction as the optimization goal, and applies the nearest neighbor clustering algorithm based on the greedy strategy to the unmanned boat maritime rescue task target allocation problem on the basis of the traditional CBBA algorithm. The positions of people who fall into the water are grouped into the same task cluster, and the center of the task cluster is used as the target position (cluster center). When the catamaran unmanned boat arrives at the target position, multiple small unmanned boats are released synchronously to carry out the rescue, thereby avoiding the time loss caused by multiple releases of small unmanned boats. In addition, in view of the kinematic constraints of the under-actuated unmanned boat, a cost function based on the Dubins curve is designed to more accurately calculate the actual distance from the unmanned boat to the target point. Finally, considering the contradiction that the amount of tasks in the task cluster exceeds the limited load of the unmanned boat, the maximum load of the task cluster is set.

[0062] It has the following advantages:

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] First, compared to the centralized communication methods widely used in existing maritime unmanned boat search and rescue systems, this paper proposes an innovative distributed communication strategy. This strategy eliminates the reliance on centralized planning and unified computing communication nodes, reduces hardware resource consumption, and improves the accuracy and flexibility of search and rescue operations. Under this strategy, each unmanned boat acts as an independent communication node, enabling information exchange and collaborative operations with other unmanned boats through intelligent algorithms, thereby ensuring the efficiency and stability of search and rescue operations.

[0065] Second, compared with the existing application of CBAA in the problem of allocating unmanned boat cluster maritime search and rescue tasks, this paper proposes Cluster-CBBA for the target allocation problem of unmanned boat maritime rescue tasks. It is verified that the designed Cluster-CBBA is superior to the commonly used CBBA in solving the problem of drowning rescue. It fully considers the coupling between tasks and can effectively improve the rescue efficiency and cost-effectiveness, thereby reducing casualties while increasing the success rate of rescue operations.

[0066] Third, compared with the existing unmanned boat task allocation method, the allocation method designed in the present invention fully considers the kinematic constraint problem caused by the under-actuated characteristics of the unmanned boat, making the allocation result closer to the path distance from the unmanned boat to the target point, thereby improving the executability of the task. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0068] Figure 1 This is a schematic diagram of a search and rescue scenario.

[0069] Figure 2 is the Dubins curve path model, (a) is RSL, (b) is LSR, (c) is RSr, (d) is LSR, (e) is RLR, and (f) is LRL;

[0070] Figure 3 It is the shortest Dubins path analysis graph.

[0071] Figure 4 It is the overall task execution roadmap.

[0072] Figure 5 It is the overall flow chart of the rescue mission.

[0073] Figure 6 are four snapshots of the iterative process of the five USVs in the consensus phase, where (a) t = 5, (b) t = 10, (c) t = 10, and (d) t = 15;

[0074] Figure 7 It is the mission schedule of the catamaran unmanned boat. DETAILED DESCRIPTION

[0075] It should be noted that, unless there is any conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0076] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is in no way intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0077] This paper proposes a distributed collaborative target allocation strategy for the target allocation of catamaran unmanned boats in maritime rescue missions. First, taking global time satisfaction optimization as the rescue goal and fully considering the dynamic changes in the rescue process, a distributed unmanned boat rescue task allocation algorithm that does not require a communication center is selected. Building on the traditional consistency bag algorithm (CBBA), a nearest neighbor clustering algorithm based on a greedy strategy is applied to the target allocation problem of maritime unmanned boat rescue missions, and a cluster-based consistency bag algorithm (Cluster-CBBA) is constructed. This algorithm groups people with similar positions in the water into the same task cluster, with the center of the task cluster as the target location (cluster center). When the catamaran unmanned boat arrives at the target location, multiple small unmanned boats are released simultaneously for rescue, thus avoiding the time loss caused by multiple releases of small unmanned boats. This method not only reduces communication bandwidth and task load, but also improves allocation efficiency. Furthermore, based on the actual kinematic characteristics of underactuated unmanned boats, a cost function based on the Dubins curve is designed to quantify the distance cost between the unmanned boat and the target point, resolving the kinematic constraints that are ignored by the cost function based on a straight line.

[0078] A distributed multi-task allocation method for underactuated unmanned boats for maritime rescue, comprising the following steps:

[0079] S1: Based on the location information of people who fall into the water at sea, a greedy nearest neighbor clustering algorithm is used to perform clustering and task clustering. The objective function of the target allocation problem of the unmanned boat rescue task at sea is constructed with the shortest search and rescue distance as the constraint condition, and the optimal time satisfaction is used as the rescue goal.

[0080] S2: The cluster consensus consistency package algorithm is used to solve the objective function of the target allocation problem of the maritime unmanned boat rescue mission, and realize the task allocation of multiple maritime search and rescue unmanned boats carrying multiple search and rescue unmanned boats.

[0081] Steps S1 / S2 are executed sequentially;

[0082] The present invention provides two main types of unmanned boats designed for rescue missions: a medium-sized catamaran search and rescue unmanned boat equipped with a lifting frame and a small search and rescue unmanned boat, featuring high speed and large capacity. In the event of an accident, this unmanned boat can quickly reach the accident site and effectively release the small search and rescue unmanned boat for rescue; the other type is a small search and rescue unmanned boat, capable of providing precise rescue for lone or unconscious individuals.

[0083] At the search and rescue site, a search and rescue drone and a catamaran-style unmanned search and rescue boat equipped with multiple small unmanned search and rescue boats operate simultaneously. First, the search and rescue drone locates the person in the water from the air. This location information is then transmitted to the catamaran-style unmanned search and rescue boats via wireless communication. The unmanned boats continuously receive the person's location information and, using the rescue task allocation algorithm designed in this paper, make decisions, quickly travel to the location, and release the small unmanned search and rescue boats for rescue.

[0084] Description of unmanned boat mission assignment

[0085] The catamaran UAV maritime rescue model can be viewed as a mathematical model within the Single Robot Task-Multiple Robot Task-Time Delay Assignment (ST-MR-TA) framework. Each UAV in the catamaran swarm can perform multiple tasks, and there are timing issues between these tasks. The order in which tasks are executed is closely related to the utility of completing them.

[0086] For ease of understanding: Let the subject of the task be N u The set of catamaran unmanned boats A = {α i |i=1,2…N u}, by N t The set of tasks to be executed is set as T = {β j |j=1,2…N t The decision-making goal of the allocation is to achieve a feasible and conflict-free task allocation to maximize the objective function. The mathematical expressions constructed are shown in formulas (1)-(4):

[0087]

[0088] where x ij ∈{0,1} is the decision vector, when x ij =1, it means that the task point S j Mission Beta j Assigned to catamaran unmanned boat α i ; represents the catamaran unmanned boat α i The maximum number of tasks allowed to be executed; represents the catamaran unmanned boat α i The task sequence to be executed, if the catamaran unmanned boat α i The kth task executed is β j , then p i The kth element of is j, if the catamaran unmanned boat α i There is no task to execute, then the ordered task set is

[0089] The profit function c in formula (1) ij (x ij ,p i ) is a decision vector x ij and task sequence p i A non-negative function of the catamaran unmanned boat α i According to p i Complete Mission β j The benefit obtained when x ij =0, c ij (x ij ,p i )=0;

[0090] Formula (2) represents the catamaran unmanned boat α i The number of tasks performed does not exceed

[0091] Formula (3) indicates that each mission can only be performed by one catamaran unmanned boat;

[0092] Formula (4) indicates that the mission volume completed by all catamaran unmanned boats cannot exceed the total mission volume.

[0093] Furthermore, the process of constructing the optimal goal of time satisfaction is as follows:

[0094] In the rescue process, the sooner the rescued person is rescued, the better. The starting time is 0. According to the change of the survival probability of the trapped person over water with time, it is assumed that the survival probability of the person in the water decreases exponentially with time. The survival probability function of a single individual is formula (5):

[0095]

[0096] Where λ<1 means the number of people falling into the water is β j Time discount factor, t ij (p i ) is a catamaran unmanned boat α i Along the path β j Arrival at the location of the person who fell into the water j The time required, is the current danger level of the drowning person, and the situation is divided into three levels: "low risk 3", "medium risk 2" and "high risk 1". The higher the survival probability, the higher the satisfaction. The time satisfaction of the drowning person regarding the rescue time is considered to be positively correlated with the survival probability of the drowning person. In the case of multiple drowning people, the constraint condition of time satisfaction is formula (6):

[0097]

[0098] Where R j Represents task β j The initial satisfaction value.

[0099] Furthermore, the process of constructing the constraint condition of the shortest search and rescue distance is as follows:

[0100] Taking into account the influence of turning radius on the motion of catamaran unmanned boat, the Dubins path algorithm is used to calculate the shortest path for the catamaran unmanned boat to reach the state (x2, y2, θ2) (x2, y2 are the horizontal and vertical coordinates of the catamaran unmanned boat at the position, θ2 is the heading of the catamaran unmanned boat at the position) from the state (x1, y1, θ1) (x1, y1 are the horizontal and vertical coordinates of the catamaran unmanned boat at the position, and θ1 is the heading of the catamaran unmanned boat at the position) to meet the kinematic constraints. When the minimum turning radius is known, for the multi-task assignment of each catamaran unmanned boat, the route to the first target point can be directly calculated by the Dubins path algorithm through the initial position of the catamaran unmanned boat and the position of the target. The subsequent target points need to be based on the position of the catamaran unmanned boat when it reaches the first target point and the position of the second target point as the input of this Dubins path. The center O2 of the turning circle of the catamaran unmanned boat passes through the initial position of the catamaran unmanned boat and the minimum turning radius, and is obtained:

[0101]

[0102] The coordinates of tangent points B and C are:

[0103]

[0104] The optimal trajectory of the catamaran unmanned boat to reach the target point 1 is composed of arc AB, straight line BC and arc CD, and the length is:

[0105]

[0106] The length of the chord l corresponding to the central angle ε is:

[0107]

[0108] Then the position coordinates of the catamaran unmanned boat when it reaches the first target point are D(x D ,y D )for:

[0109]

[0110] The heading angle θ2 when the catamaran unmanned boat reaches the first target point is.

[0111]

[0112] Furthermore, the clustering grouping process of the greedy strategy nearest neighbor clustering algorithm is as follows:

[0113] In the initialization stage, for each sample point, the number of adjacent points within the distance threshold is calculated;

[0114] Then, the sample points with the most adjacent points are selected as cluster centers. These cluster centers will serve as representative locations of subsequent task clusters and can more accurately represent the core of the cluster area.

[0115] The positions of people who fall into the water that are closer to the cluster center are grouped into the same task cluster. By selecting the cluster center, the positions of people who fall into the water within the task cluster can better reflect the actual clustering situation;

[0116] The center of each task cluster, i.e., the cluster center, will serve as the target location of the task cluster; the unmanned boat will use this target location as a guide to proceed to the rescue;

[0117] When the unmanned boat reaches the target location, the center of the cluster, multiple small unmanned boats are released simultaneously to carry out the rescue mission. This method can rescue multiple people in the water at the same time, avoiding the time loss of multiple releases of small unmanned boats in traditional methods, thus significantly improving rescue efficiency.

[0118] Through this approach, UAVs can efficiently allocate tasks and avoid time delays caused by frequent UAV dispatch during the rescue process. The rational selection of task cluster centers allows rescue operations within each task cluster to be centralized, improving the coordination and timeliness of the entire rescue process.

[0119] With the assistance of drones, the distribution information of the positions of people who fall into the water on the X-axis and Y-axis can be determined. In order to improve the efficiency of the catamaran unmanned boat in releasing small unmanned boats for rescue, an optimized strategy is adopted, that is, similar positions of falling into the water are classified into the same task group, and the center point of the group is set as the target position (i.e., the cluster center). When the catamaran unmanned boat arrives at this target position, multiple small unmanned boats will be released simultaneously to carry out rescue. This method greatly shortens the time consumed by repeatedly releasing small unmanned boats. In order to minimize the number of clusters generated, the present invention, based on the original nearest neighbor clustering algorithm based on distance threshold, no longer randomly selects sample points as cluster centers, but selects sample points with the most adjacent points within the threshold as cluster centers. Similar positions of people who fall into the water are classified into the same task cluster, and the center of the task cluster is used as the target position (cluster center). When the unmanned boat arrives at the target position, multiple small unmanned boats are released at the same time to rescue the people who fall into the water, thereby avoiding the shortcomings that may be caused by randomly selecting sample points.

[0120] Furthermore: the task cluster package construction process is as follows:

[0121] Defining an index set During the mission cluster package construction phase, five bidding lists are created based on the mission information obtained by each catamaran unmanned boat through wireless communication equipment, and the five bidding list information is defined:

[0122] 1). Task cluster list b i , Represents the task cluster set α obtained by the bidding of the catamaran unmanned boat i ,The assigned task clusters are sorted according to the order of joining the task clusters;

[0123] 2).Task cluster sequence p i , Indicates the allocation to the catamaran unmanned boat α i The execution order of the task clusters;

[0124] 3). Task cluster completion list pn i , Indicates the number of people rescued from drowning corresponding to each cluster;

[0125] 4). List of winning bidders i , There is a catamaran unmanned boat ij ∈y i (j=1,2…N t ), y ij represents the catamaran unmanned boat α i In local state awareness, task cluster σ j The highest bid;

[0126] 5). Winners listi , That is, there exists z ij ∈z i (j=1,2…N t ) indicates a bid of y ij The catamaran unmanned boat,

[0127] In the Cluster-CBBA algorithm, the value of the CBBA bid is c ij The score is calculated based on the marginal gain decreasing function; the marginal gain decreasing function indicates that the score value of the task cluster will not increase as a new task cluster is added to the existing task cluster set, that is, formula (14):

[0128]

[0129] Among them, b i represents the catamaran unmanned boat α i The current task cluster set represents any task cluster added to the tail of the current task cluster set, c ij Indicates greedily adding task clusters according to the score value, where the task clusters with higher scores are added first, c ij (b i ) is as shown in equations (15)-(17):

[0130]

[0131] In formula (15), p i Indicates that it is assigned to a catamaran unmanned boat i The task cluster sequence of formula (16) represents any execution task cluster sequence p i The total score of t ij Indicates a catamaran unmanned boat i Execution task cluster σ j time; Formula (17) represents the catamaran unmanned boat a i Execution task cluster σ j Benefits obtained.

[0132] Furthermore, the consistency package algorithm using cluster consensus solves the objective function of the target allocation question of the maritime unmanned boat rescue mission, and the process of realizing the task allocation of multiple maritime search and rescue unmanned boats equipped with multiple search and rescue unmanned boats is as follows:

[0133] During the mission cluster package construction phase, each catamaran UAV performs mission planning based on its local information. Therefore, a single mission cluster may be assigned to multiple catamaran UAVs, and conflicts between the catamaran UAVs need to be resolved through data exchange to achieve a global conflict-free mission.

[0134] In the conflict resolution phase, the catamaran unmanned boat uses a consensus strategy to coordinate the allocation of global task clusters.

[0135] Assume that each catamaran unmanned boat adopts a synchronous communication mechanism in the consistency stage and updates the three vectors (y i , z i and timestamps i ) information to obtain the latest allocation information, based on communication exchanges, catamaran unmanned boat a i For task cluster σ j There are three possible operational definitions, as shown in Equation (18), and the decision rule is as follows

[0136] As shown in Table 1:

[0137]

[0138] Table 1 Conflict coordination methods in communication

[0139]

[0140]

[0141] The UAVs use lookup table 1 to determine if it should update, reset, or leave the bid, communicating and negotiating by comparing what receiver i and sender k know about task j and the last time each UAV received communication from the UAV they believe is assigned to task j, until they converge to a conflict-free solution where all tasks have a unique assigner and there is no overlapping or conflicting UAVs. Figure 1 A schematic diagram of the search and rescue scenario for an under-actuated unmanned boat is shown. The search and rescue mission is carried out in the following three steps:

[0142] 1): The search and rescue drone locates the position of the person who fell into the water in the air.

[0143] 2): The positioning information is then transmitted to the catamaran search and rescue unmanned boat via wireless communication.

[0144] 3): The double-headed unmanned boat continuously receives the location information of the person who fell into the water, makes dynamic decisions through the rescue task allocation algorithm designed in this paper, quickly drives to the falling point and releases a small search and rescue unmanned boat for rescue.

[0145] Figure 2The path models of the Dubins curve are shown in Figures 1 and 2. (a) RSL, (b) LSR, (c) RSr, (d) LSR, (e) RLR, and (f) LRL. Given a minimum turning radius, the shortest path between two states can be either circle-line-circle or circle-circle-circle. The circle refers to the UAV's minimum turning radius, and the line refers to the tangent of the circle. These two types of paths include six path patterns: {RSL, LSR, RSR, LSL, RLR, LRL} (L represents a left-turning circle, R represents a right-turning circle, and S represents a straight line).

[0146] Figure 3 The shortest Dubins path analysis diagram of the catamaran unmanned boat is described, which clearly reveals the motion characteristics of the under-actuated unmanned boat and makes it closer to the path distance from the unmanned boat to the corresponding target point.

[0147] Figure 4 The overall search and rescue mission execution roadmap is displayed.

[0148] Figure 5 The total rescue route obtained by Cluster-CBBA algorithm is shown. Figure 5 It can be seen that after clustering the falling points, 30 falling points form 15 task clusters. The search and rescue UAV locates the position of the cluster center of the 15 task clusters in the air. The positioning information is then transmitted to the five catamaran search and rescue unmanned boats through wireless communication. The catamaran unmanned boats continuously receive the position information of the cluster center and make dynamic decisions through the rescue task allocation algorithm Cluster-CBBA designed in this paper to generate a non-contradictory allocation result. At the same time, considering the influence of the turning radius on the movement of the catamaran unmanned boat, the minimum turning radius R is adopted. min The shortest path that satisfies the dynamic constraints is calculated by using the 4m Dubins curve for the catamaran unmanned boat from the initial position to each cluster center, and it quickly travels to the cluster center and releases a small search and rescue unmanned boat to rescue the people who fall into the water.

[0149] Figure 6 These are four snapshots of the iterative process of the five USVs in the consensus phase, where (a) t = 5, (b) t = 10, (c) t = 10, and (d) t = 15. Figure 6 The iterative process of the catamaran unmanned boat at the 1st, 5th, 10th and 15th times in the consensus stage is shown.

[0150] Figure 7 The mission schedule of the catamaran unmanned boat is displayed.

[0151] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distributed multi-task allocation method for underactuated unmanned boats for maritime rescue, characterized by: The following steps are involved: Based on the location information of people who fall into the water at sea, the nearest neighbor clustering algorithm with a greedy strategy is used for clustering and task clustering. The target allocation problem of the unmanned boat rescue mission at sea is constructed with the shortest search and rescue distance as the constraint condition, and the optimal time satisfaction is taken as the rescue goal. The cluster consensus consistency bag algorithm is used to solve the objective function of the target allocation problem of the maritime unmanned boat rescue mission, and the task allocation of multiple catamaran search and rescue unmanned boats carrying multiple small search and rescue unmanned boats is realized.

2. The distributed multi-task allocation method for underactuated unmanned boats for maritime rescue according to claim 1, characterized in that: The construction process of the objective function of the target allocation problem of the unmanned boat rescue mission at sea is as follows: Because N u The set of unmanned catamaran boats A = {α i |i=1,2…N u }, by N t The set of tasks to be executed is set as T = {β j |j=1,2…N t }, the decision-making goal of the allocation is to achieve a feasible and conflict-free task allocation to maximize the objective function. The mathematical expressions constructed are shown in equations (1)-(4): where x ij ∈{0,1} is the decision vector, when x ij =1, it means that the task point S j Mission Beta j Assigned to Catamaran Unmanned Boat α i ; Represents the unmanned catamaran α i The maximum number of tasks allowed to be executed; Represents the unmanned catamaran α i The task sequence to be executed is as follows: i The kth task executed is β j , then p i The kth element of is j, if the catamaran unmanned boat α i There is no task to execute, then the ordered task set is The profit function c in formula (1) ij (x ij ,p i ) is a decision vector x ij and task sequence p i A non-negative function of i According to p i Completed Mission β j The benefit obtained is when x ij =0, c ij (x ij ,p i )=0; Formula (2) represents the catamaran unmanned boat α i The number of tasks to be performed does not exceed Formula (3) indicates that each task can only be performed by one catamaran unmanned boat; Formula (4) indicates that the mission volume completed by all catamaran unmanned boats cannot exceed the total mission volume.

3. The distributed multi-task allocation method for underactuated unmanned boats for maritime rescue according to claim 1, characterized in that: The construction process of the optimal goal of time satisfaction is as follows: The starting point of time is 0. According to the change of the survival probability of the trapped person over time, it is considered that the survival probability of the person in the water decreases exponentially over time. The survival probability function of a single individual is formula (5): In the formula, λ<1 means the number of people falling into the water is β j The time discount factor, t ij (p i ) is a catamaran unmanned boat α i Along the path β j Arrival at the location of the drowning person j The time required, is the current danger level of the drowning person, and the situation is divided into three levels: "low risk 3", "medium risk 2" and "high risk 1". The higher the survival probability, the higher the satisfaction. The time satisfaction of the drowning person about the rescue time is considered to be positively correlated with the survival probability of the drowning person. In the case of multiple drowning people, the constraint condition of time satisfaction is formula (6): In the formula, R j Represents task β j The initial satisfaction value.

4. The distributed multi-task allocation method for underactuated unmanned boats for maritime rescue according to claim 1, characterized in that: The construction process of the constraint condition of the shortest search and rescue distance is as follows: The Dubins path algorithm is used to calculate the shortest path that satisfies the kinematic constraints of the catamaran unmanned boat from the state (x1, y1, θ1) to the state (x2, y2, θ2), where x1 and y1 represent the horizontal and vertical coordinates of the catamaran unmanned boat at the position, and θ1 represents the heading of the catamaran unmanned boat at the position; x2 and y2 represent the horizontal and vertical coordinates of the catamaran unmanned boat at the position, and θ2 represents the heading of the catamaran unmanned boat at the position; When the minimum turning radius is known, for each catamaran unmanned boat in the multi-task assignment, the route to the first target point can be directly calculated by the initial position of the catamaran unmanned boat and the position of the target using the Dubins path algorithm. The subsequent target points need to be based on the position of the catamaran unmanned boat when it reaches the first target point and the position of the second target point as the input of this Dubins path. The center O2 of the turning circle of the catamaran unmanned boat passes through the initial position and minimum turning radius of the catamaran unmanned boat and is obtained: The coordinates of the tangent points B and C are: The optimal trajectory of the catamaran unmanned boat to reach the target point 1 is composed of arc AB, straight line BC and arc CD, and the length is: The length of the chord length l corresponding to the central angle ε is: Then the position coordinates of the catamaran unmanned boat when it reaches the first target point are D(x D ,y D )for: The heading angle θ2 when the catamaran unmanned boat reaches the first target point is.

5. The distributed multi-task allocation method for underactuated unmanned boats for maritime rescue according to claim 1, characterized in that: The clustering grouping process of the nearest neighbor clustering algorithm of the greedy strategy is as follows: In the initialization stage, for each sample point, the number of adjacent points within the distance threshold is calculated; Then, the sample points with the most neighboring points are selected as cluster centers. These cluster centers will serve as representative locations of subsequent task clusters and can more accurately represent the core of the clustered area. The positions of people who fall into the water that are closer to the cluster center are grouped into the same task cluster. By selecting the cluster center, the positions of people who fall into the water in the task cluster can better reflect the actual aggregation situation. The center of each task cluster, i.e., the cluster center, will be used as the target location of the task cluster; the unmanned boat will use the target location as a guide to go to the rescue; When the unmanned boat reaches the target location, which is the center of the cluster, multiple small unmanned boats will be released at the same time to carry out the rescue mission.

6. The distributed multi-task allocation method for underactuated unmanned boats for maritime rescue according to claim 1, characterized in that: The task cluster package construction process is as follows: Defining an index set In the task cluster package construction phase, five bidding lists are created based on the task information obtained by each catamaran unmanned boat through the wireless communication device, and the five bidding list information are defined: 1). Task cluster list b i , Represents the task cluster set α obtained by the bidding of the catamaran unmanned boat i ,The assigned task clusters are sorted according to the order of joining the task clusters; 2). Task cluster sequence p i , Indicates the allocation to the catamaran unmanned boat α i The execution order of the task clusters; 3). Task cluster completion list pn i , Indicates the number of people rescued from drowning corresponding to each cluster; 4). List of winning bidders i , That is, there is a catamaran unmanned boat ij ∈y i (j=1,2…N t ), y ij Represents the unmanned catamaran α i In local state awareness, task cluster σ j The highest bid; 5). Winners list i , That is, there exists z ij ∈z i (j=1,2…N t ) indicates the bid is y ij Catamaran unmanned boat In the consensus pack algorithm based on cluster consensus, the value c of the CBBA bid is ij The score is calculated based on the marginal gain decreasing function; the marginal gain decreasing function indicates that the score value of the task cluster will not increase as a new task cluster is added to the existing task cluster set, that is, formula (14): Among them, b i Represents the unmanned catamaran α i The current task cluster set represents any task cluster added to the tail of the current task cluster set, c ij Indicates greedily adding task clusters according to the score value, where the task clusters with higher scores are added first, c ij (b i ) is as follows (15)-(17): In formula (15), p i Indicates that it is assigned to a catamaran unmanned boat i The task cluster sequence of p; Formula (16) represents any execution task cluster sequence p i The total score value of t ij Indicates a catamaran unmanned boat i Execute task cluster σ j time; Formula (17) represents the catamaran unmanned boat α i Execute task cluster σ j Benefits obtained.

7. The distributed multi-task allocation method for underactuated unmanned boats for maritime rescue according to claim 1, characterized in that: The consistent packet algorithm using cluster consensus solves the objective function of the target allocation question of the maritime unmanned boat rescue mission, and the process of realizing the task allocation of multiple maritime search and rescue unmanned boats carrying multiple search and rescue unmanned boats is as follows: Assume that each catamaran unmanned boat adopts a synchronous communication mechanism in the consistency stage and updates the three vectors (y i , z i and timestamps i ) information to obtain the latest allocation information, based on communication exchanges, catamaran unmanned boat a i For task cluster σ j Three possible operational definitions are adopted, as shown in equation (18), and the decision rules are as follows: The UAVs use the above approach to determine if it should update, reset, or leave the bid, communicating and negotiating by comparing what receiver i and sender k know about task j and the last time each UAV received communication from the UAV they think is assigned to task j, until they converge to a conflict-free solution, where all tasks have a unique assignor and there is no overlap or conflict between the UAVs.

Images