Ship-shore cooperative energy-saving collision avoidance berthing method and system based on adaptive time-domain MPC

Abstract

The application relates to a ship-shore cooperative energy-saving collision avoidance berthing method and system based on adaptive time-domain network model predictive control (MPC). In the model predictive control, an adaptive prediction time domain is designed. The existing objective function of the model predictive control is improved. Energy consumption is constructed as one of the objective functions in the nonlinear model predictive control framework. Based on the framework, a new actual control and state constraint algorithm is designed to ensure energy saving of the berthing framework. Therefore, the improved emperor penguin optimizer is used for conflict avoidance decision. At each sampling time, based on a nonlinear ship maneuvering model, a finite time domain optimal control problem is established. Under different disturbances, including direct berthing and turning approach berthing technology, different berthing methods are realized by using robust model predictive control. Through the control strategy, close connection between ship maneuvering control and energy consumption is realized. Due to its predictability, the MPC ensures collision detection and avoidance as well as path planning and tracking.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of ship berthing control, in particular to a ship-land cooperative energy-saving collision avoidance berthing method and system based on adaptive time-domain MPC. BACKGROUND

[0002] With the increasing demand for energy, as well as the dual constraints of environment and enterprise survival, the energy saving and emission reduction of ships has attracted more and more attention. In previous studies, energy consumption was rarely considered in path planning. In the berthing process, due to low-speed sailing, turning efficiency is reduced, and maneuverability is limited. The water area of the wharf parking lot is a forbidden zone, and the traffic flow of nearby ships is more complex, which greatly increases the difficulty and risk of berthing. Therefore, in order to develop unmanned ship technology and improve berthing safety, it is imperative to study automatic berthing control. Although great progress has been made in ship collision avoidance and control, the following problems still exist:

[0003] (1) The model of target and constraint in the process of collision avoidance is not linear, but it involves time-varying nonlinear functions such as geometry and triangle;

[0004] (2) The feasible set of collision avoidance optimization problem is non-convex, which belongs to NP-hard problem of uncertain polynomial time. The commonly used methods are mixed integer programming and non-convex quadratic constraint quadratic programming. However, in the feasible region of non-convex optimization, the solution is not necessarily a global solution;

[0005] (3) The energy device that unmanned ships can carry is limited. In order to take energy consumption as one of the goals of the control algorithm, it is very important to improve the endurance of these ships. In addition, so far, most collision avoidance and control algorithms have not considered energy consumption.

[0006] (4) In the meeting of multiple ships, if a ship is considered to have the responsibility of giving way and standing at the same time, there is an identity conflict, and it may be difficult to handle such a situation. In most studies on collision avoidance decision, the focus is on the collision avoidance opportunity when the ship is a giving way ship, and less research is done when the ship is a standby ship. When the giving way ship does not take collision avoidance action, how to determine the collision avoidance opportunity. SUMMARY

[0007] Therefore, the purpose of the present application is to provide a ship-land cooperative energy-saving collision avoidance berthing method and system based on adaptive time-domain MPC, which aims to solve the above problems.

[0008] To achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0009] The application discloses a ship-shore cooperative energy-saving collision avoidance berthing method and system based on adaptive time domain MPC, which comprises an underactuated ship predictive control unit, a collision avoidance coordination unit and an automatic berthing control unit.

[0010] Further, the control strategy of the underactuated ship predictive control unit is specifically as follows:

[0011] Step S1: based on a preset berthing berth position and an expected ship berthing speed, a system model, system constraints and optimization indexes, an MPC constraint optimization problem is constructed with energy consumption as one of the targets;

[0012] Step S2: a motion reference trajectory of the ship is acquired, an underactuated ship motion model is constructed, motion control data is predicted according to the motion reference trajectory, and the motion control data is used to control the motion state to obtain motion data of the underactuated ship;

[0013] Step S3: the offset size of the underactuated ship motion data and the motion reference trajectory is compared, and motion adjustment data is generated according to the comparison result of the offset size, the motion data is used to adjust the motion state, and the offset between the underactuated ship motion data and the motion reference trajectory is calculated;

[0014] Step S4: input force and torque of the controller are generated according to the comparison result of the offset.

[0015] Further, the MPC constraint optimization problem is specifically as follows: at a current time t k The objective function of the i-th ship is set as:

[0016]

[0017] Wherein T p is a prediction time domain, X e (t;t k ) is a prediction of the error state of the system at time t k , U e (t;t k ) is a prediction of the error input of the system at time t k , Q is a weighted matrix of the error state, F is a weighted matrix of the input error, and P is a weight matrix of the terminal error state; the objective function minimizes the distance between the ship position in the prediction interval and the corresponding point on the reference trajectory, the size of the control amount and the energy consumption; the input U, the expected input U d, input error U e and the coefficient matrix K as follows:

[0018] U = [τ u , τ r ] T (1)

[0019]

[0020] U d = [τ ud , τ rd ] T (3)

[0021]

[0022] Let η be the position and orientation vector, υ be the velocity vector, u be the forward velocity, v be the sway velocity, r be the yaw velocity, x be the forward position, y be the sway position, and Ψ be the yaw angle, then the underactuated ship motion model is:

[0023]

[0024] η = [x y ψ] T (6)

[0025] υ = [u v r] T (7)

[0026] where R is the rotation matrix:

[0027]

[0028] The dynamics model of the underactuated ship is calculated as follows:

[0029]

[0030] The velocity equation is expressed as:

[0031]

[0032] where τ· u represents the control input along the forward direction, τ· r represents the control input along the yaw direction; m 11 , m 22 , and m 33 represent the first, second, and third elements on the diagonal of the inertia matrix, respectively; d 11 , d 22 , and d 33 represent the first, third, and second elements on the diagonal of the linear fluid dynamic damping parameter matrix, respectively; X represents the position and velocity vector, Fwx represents the wind disturbance along the forward direction, and Fwy is the wind disturbance in the yaw direction, N w is the wind disturbance in the yaw direction, F cx is the current disturbance in the forward direction, F cy is the flow field disturbance in the yaw direction, N c is the current disturbance in the yaw direction. X d is expressed as the desired position and velocity vector, whose components are given by

[0033] X = [x y ψ u v r] T (11)

[0034] X d = [x d y d ψ d u d v d r d ] T (12)

[0035] where T p is the prediction horizon, X e (t; t k ) is the prediction of the error state of the system at time t k , U e (t; t k ) is the prediction of the error input of the system at time t k , Q is the weighting matrix of the error state, F is the weighting matrix of the input error, P is the weight matrix of the terminal error state; the objective function minimizes the distance between the ship position in the prediction interval and the corresponding point on the reference trajectory, as well as the size and energy consumption of the control quantity;

[0036] Further, the energy consumption is specifically:

[0037]

[0038] Further, the step S3 is specifically:

[0039] where x d is the desired forward position, y d is the desired yaw position, ψ d is the desired yaw angle, u d is the desired forward speed, v d is the desired yaw speed, and r d is the desired yaw angle speed.

[0040] In addition, let x e be the forward position error, y e be the yaw position error, and ψ e be the yaw angle error.is the yaw angle error, u e is the forward speed error, v e is the roll speed error, r e is the yaw angle speed error, then X e is expressed as follows

[0041]

[0042] The error dynamics are written as:

[0043]

[0044] Further, the MPC constraint optimization problem is also provided with an adaptive prediction range, specifically:

[0045]

[0046] where T pmin is the lower limit value of the prediction time domain, T pmax is the upper limit value of the prediction time domain, and e represents the exponential function.

[0047] Further, the collision avoidance coordination unit is based on the improved emperor penguin optimization algorithm for collision avoidance decision, specifically

[0048] A1: considering the parameters of DCPA, TCPA, turning distance, distance to the original route, etc., a collision risk assessment model is established:

[0049]

[0050] where x1 represents the steering angle, x2 is the recovery time, w1, w2, w3 and w4 are weights; the objective function f is to maximize DCPA and minimize the distance to the original route;

[0051] A2: the quantum particle swarm based on Sobol random sequence is introduced into EPO to obtain the improved EPO, and the optimal collision avoidance scheme is obtained based on the improved EPO solving formula (36).

[0052] Further, let the speed of the main ship v A = (v Ax , v Ay ), the speed of the target ship v B = (v Bx , v By ), and the flow velocity vector of the target ship relative to the main ship is:

[0053]

[0054] Relative heading:

[0055]

[0056] Alpha is calculated as follows:

[0057]

[0058] Where R T represents the distance, the position of the primary ship is The position of the target ship is Alpha T Refers to the azimuth angle, and the calculation method is as follows:

[0059]

[0060]

[0061] The distance to the closest point of approach DCPA, which reflects the highest acceptable navigation risk, is calculated as follows:

[0062] DCPA=R T Sin theta (22)

[0063] The time to the closest point of approach TCPA is calculated as follows:

[0064]

[0065] The conditions of ship encounter are defined as follows:

[0066]

[0067] Step S5: The prediction control increases constraint conditions, including state change constraints and ship collision avoidance. The emperor penguin optimization is adopted, and the optimization problem with constraints is iteratively solved until the optimal solution is converged.

[0068] Step S6: The shore station monitors whether the ship has a collision danger. According to the navigation conditions of the ship, the anti-collision instruction is generated, and the real-time interaction, dynamic adjustment, reasonable division of labor and cooperative control between the shore and the ship are realized.

[0069] Further, the berthing includes direct berthing and turning to approach berthing.

[0070] Compared with the prior art, the present application has the following beneficial effects:

[0071] The present application effectively improves the efficiency of ship-shore cooperation, reduces the collision probability of ship berthing, reduces the energy consumption of the ship, and realizes automatic berthing. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 Is the Sobo l sequence in an embodiment of the present application;

[0073] Figure 2is a schematic diagram of relative motion of a ship in an embodiment of the present application;

[0074] Figure 3 is a ship motion model in an embodiment of the present application

[0075] Figure 4 is non-convexity of collision constraint using Goodwin ship domain in an embodiment of the present application

[0076] Figure 5 is a priority-based network energy-saving nonlinear MPC framework in an embodiment of the present application

[0077] Figure 6 is a flowchart of the present application. DETAILED DESCRIPTION

[0078] The present application will be further described below in conjunction with the accompanying drawings and embodiments.

[0079] Please refer to Figures 1-5 The present application provides a ship-shore cooperative energy-saving collision avoidance berthing method and system based on adaptive time domain MPC, which comprises an underactuated ship predictive control unit, a collision avoidance coordination unit and an automatic berthing control unit. The underactuated ship predictive control unit is based on an energy-saving robust nonlinear model predictive control of adaptive prediction horizon to handle state constraints in actual control and berthing process based on ship-shore cooperation. The collision avoidance coordination unit is based on an improved emperor penguin optimizer to realize conflict avoidance decision. The automatic berthing control unit realizes different berthing methods under different disturbances by using robust model predictive control.

[0080] In this embodiment, the parameters and definitions are shown in the following table: In this embodiment, the control strategy of the underactuated ship predictive control unit is specifically:

[0081] Step S1: Based on the preset berthing berth position and expected ship berthing speed, combined with the system model, system constraints and optimization index, an MPC constraint optimization problem is constructed with energy consumption as one of the targets.

[0082] Step S2: Obtain the motion reference trajectory of the ship, build the underactuated ship motion model, and predict the motion control data according to the motion reference trajectory, and use the motion control data to control the motion state to obtain the motion data of the underactuated ship.

[0083] Step S3: Compare the offset size of the underactuated ship motion data and the motion reference trajectory, and generate motion adjustment data according to the comparison result of the offset size, the motion data is used to adjust the motion state, and the offset between the underactuated ship motion data and the motion reference trajectory is calculated.

[0084] Step S4: Generate the input force and torque of the controller according to the comparison result of the deviation.

[0085] Step S5: Predictive control adds constraints, including state change constraints and ship collision avoidance. Emperor Penguin optimization is used to iteratively solve the constrained optimization problem until the optimal solution is converged.

[0086] Step S6: The shore station monitors whether the ship is in danger of collision. According to the navigation conditions of the ship, the anti-collision command is generated to realize the real-time interaction, dynamic adjustment, reasonable division of labor and cooperative control between the shore and the ship.

[0087] In this embodiment, the EPO specifically includes setting the selected position range of the Emperor Penguin during the curling heating process within the grid range of a polygon. During the collection process, each Emperor Penguin is at least adjacent to two others, and the selection of neighbors is random. During the Emperor Penguin cluster process, the range boundary is an irregular polygon; therefore, the wind gradient around the Emperor Penguin cluster is used to represent the boundary of the entire cluster, and the wind speed is defined by γ. Its gradient is α and β. The cluster boundary is μ. In order to restore, the wind speed and the cluster boundary can be represented by equations (25) and (26), respectively, as follows:

[0088] γ = Δα (25)

[0089] μ = α + iβ (26)

[0090] The harsh environment in the polar region causes Emperor Penguins to gather in cold weather during migration to keep warm. If the current gathering radius d is greater than 0.5, the temperature W is zero. However, when d is less than 0.5, the temperature W is equal to 1. The temperature gradient curve can be described as follows:

[0091]

[0092] where t max is the maximum number of iterations. x is the current iteration number, and the expression for temperature W is:

[0093]

[0094] The distance between Emperor Penguins within the group is represented as the distance between the individual and the group center Emperor Penguin. The cluster distance formula is as follows:

[0095] L ep = ||F(Γ)O best (x)-IO ep (x)|| (29)

[0096]

[0097] P acc = ||O best (x)-O ep (x)|| (31)

[0098] I = Random([0, 1]) (32)

[0099]

[0100] where L ep is the distance from the penguin to the center. x represents the current iteration number. Γ and I form the influence vector factor for defining the penguin volume to avoid conflicts between individuals. Obest(x) is the optimal solution in x iteration. O ep (x) represents the position vector of the current penguin. As for F, it defines the social status of the penguin and is responsible for distinguishing the best individual from the ordinary individual. Bmove is the moving step, which can be set to 2.5. Pacc represents the accuracy of the polygon grid by comparing the difference with the best value. Finally, ξ and are control parameters in the range of (2, 3) and (1.5, 2), respectively, which can lead to better results.

[0101] Individuals in the penguin colony update their position information by moving in the direction of the colony center penguin. The position update formula is as follows

[0102] O ep (x+1) = O best (x) - ΓL ep (34)

[0103] where O ep (x+1) indicates the updated position of the next generation of penguins. During the iteration process, once the mover is repositioned, the above parameters of the penguin will be recalculated.

[0104] In this embodiment, reference Figure 1 is made to Sobol sequence is a quasi-random sequence. The sampling of Sobol sequence method is more uniform than the sampling of uniform distribution random sequence. It can be extended to any dimension and is a high coverage and stable random sequence.

[0105] There is currently a defect in quasi-random number sequences, that is, the sample distribution is inconsistent with the true distribution. This event mainly occurs in the following two cases: on the one hand, the sampling cost is too high, the sample quantity is small, and on the other hand, the spatial dimension is high. Quasi-random sequences, also known as low-discrepancy sequences, are used to replace uniform distribution random sequences, of which Sobol random sequences are an example of the latter. It has good distribution uniformity and less time consumption. The sampling of Sobol sequence is more uniform than the sampling of uniform distribution random sequence. Figure 1 A scatter plot based on Sobol random sequence is shown in

[0106] Preferably, in the present embodiment, the velocity v A of the host ship Ax = (v Ay , v B ), the velocity v Bx of the target ship By , and the flow velocity vector of the target ship relative to the host ship are as follows:

[0107]

[0108] Relative heading:

[0109]

[0110] α is calculated as follows:

[0111]

[0112] where R T represents the distance, the position of the host ship is the position of the target ship is α T refers to the azimuth angle, and the calculation method is as follows:

[0113]

[0114]

[0115] The distance to the closest point of approach (DCPA) reflects the highest acceptable risk of navigation, and the calculation of DCPA is as follows:

[0116] DCPA = R T sinθ (40)

[0117] The time to the closest point of approach (TCPA) is calculated as follows:

[0118]

[0119] The conditions for ship encounter are defined as follows:

[0120]

[0121] Preferably, in the present embodiment, step S2 is specifically:

[0122] Let η be the position and attitude vector, υ be the velocity vector, u be the forward speed, v be the sway speed, r be the yaw speed, x be the surge position, y be the sway position, and ψ be the yaw angle, and the under-actuated ship motion model is as follows:

[0123]

[0124] η = [x y ψ] T (20)

[0125] υ = [u v r] T (21)

[0126] where R is the rotation matrix:

[0127]

[0128] The dynamics model of the underactuated vessel is calculated as follows:

[0129]

[0130] The velocity equation is expressed as follows:

[0131]

[0132] where τ u represents the control input along the forward direction, τ r represents the control input along the yaw direction; m 11 , m 22 and m 33 represent the first, second and third elements on the diagonal of the inertia matrix, respectively; d 11 , d 22 and d 33 represent the first, third and second elements on the diagonal of the linear fluid dynamic damping parameter matrix, respectively; X represents the position and velocity vector, Fwx represents the wind disturbance along the forward direction; F wy is the wind disturbance along the yaw direction, N w is the wind along the yaw, F cx is the current disturbance along the forward direction, F cy is the flow field disturbance along the yaw, N c is the current disturbance along the yaw.

[0133] Preferably, in the present embodiment, step S3 is specifically:

[0134] X d is expressed as the required position and velocity vector, the components of which are shown in equation (26)

[0135] X = [x y ψ u v r] T (25)

[0136] X d = [x d y d ψ d u d v d r d ]T (26)

[0137] where x d is the desired forward position, y d is the desired yaw position, ψ d is the desired yaw angle, u d is the desired surge speed, v d is the desired yaw speed, r d is the desired yaw angle speed.

[0138] In addition, let x e , y e , ψ e , u e , v e and r e be the errors of the position and speed vectors, surge position, yaw position, yaw angle, surge speed, yaw speed and yaw angle speed, respectively, then X e is expressed as follows

[0139]

[0140] The error dynamics are written as:

[0141]

[0142] In this embodiment, the MPC constraint optimization problem is also provided with an adaptive prediction range, specifically:

[0143]

[0144] where T pmin is the lower limit value of the prediction time domain, T pmax is the upper limit value of the prediction time domain, and e represents the exponential function.

[0145] In this embodiment, the shore control center monitors whether the ship is in danger of collision by analyzing and processing fusion data. If there is a collision danger, it will generate an anti-collision instruction according to the sailing conditions of the ship, and send the anti-collision instruction to the ship, so that the ship can control its own heading and speed, and optimize the local route. The core of ship-shore cooperation is to establish a complete and systematic decision-making system to optimize the allocation and balance of ship-shore information resources. It emphasizes real-time interaction and dynamic adjustment between the shore and the ship, and realizes reasonable division of labor and cooperative control between the ship and the shore. Due to its predictability, MPC can ensure collision detection and avoidance as well as path planning and tracking.

[0146] To predict the future position of the ships, their dynamic model is used. The objective function is to track the reference trajectory and avoid collisions on the trajectory. Computing the appropriate turning angle will avoid collisions, as each ship will deviate from its original reference orbit as little as possible. In the event of an imminent collision, the deviation between the new trajectory and the reference trajectory should be as small as possible if the trajectory needs to be maintained to avoid a collision. After avoiding a collision, the ship returns to the original reference trajectory. The control objective can only be achieved through collective behavior, and the feasible state set of each ship is a function of its own and other ship states.

[0147] In addition, constraints are added in the predictive control. These constraints include the ship dynamics model, control input and its variation constraints, and collision avoidance between ships. The constraints are a function of the ship and its neighbor state subset. The constrained optimization problem is solved iteratively until it converges to the optimal solution.

[0148] In cooperative MPC, each ship has its own controller, which not only considers its own interests, but also considers the interests of other ships and the dynamic changes around the ships. Each ship needs to communicate and exchange its current measured or estimated state, predicted reference trajectory and model parameters, reference trajectory and time stamp of each sampling state with other ships in the encounter situation. These ships send new predictions to each other, share the future states and positions of other ships, and know the paths of other ships that must be avoided when starting to avoid maneuvers. In a parallel computing distributed architecture, a time-varying coupling graph based on local neighborhood is used. The coupling graph is fully connected, because each ship must avoid all other ships. The chart provides relevant environmental data for MPC, which can solve part of the overall optimization problem.

[0149] When a collision is detected, the order is determined according to the reference trajectory of the ship and the DCPA of other ships. Priority-based cooperation can reduce the computation time. The designed distributed strategy update mechanism makes it work efficiently. The priority-based distributed predictive control framework combines the decentralization of sensor perception access strategy with the model predictive control algorithm based on predicted sequence by designing access threshold and priority mechanism.

[0150] When the ship is sailing normally, the heading controller designed by the model predictive control method controls the ship to travel along the established path, and calculates the collision risk in real time according to the motion information of the ship and other ships using the collision risk model. When the collision risk is higher than the threshold, the ship enters the anti-collision state. The collision avoidance process of the ship is calculated by using the rule-based collision avoidance model, and then the controller designed by the model predictive control method is used to control the ship to avoid collision and turn to ensure the safety of the ship sailing. The priority-based network energy-saving nonlinear MPC framework is as shown in Figure 5 .

[0151] In this embodiment, preferably, the energy consumption is also considered, specifically:

[0152]

[0153] where the input U, the desired input U d , the input error U e and the coefficient matrix K are as follows:

[0154] U = [τ u ,τ r ] T (31)

[0155]

[0156] U d = [τ ud ,τ rd ] T (33)

[0157]

[0158] In this embodiment, preferably, the MPC constraint optimization problem is specifically: at the current time t k The objective function of the i-th ship is set as:

[0159]

[0160] where T p is the prediction horizon, X ei (t;t k ) is the prediction of the error state of the system at time t k , U ei (t;t k ) is the prediction of the error input of the system at time t k , Q is the weighting matrix of the error state, F is the weighting matrix of the input error, and P is the weight matrix of the terminal error state; the objective function minimizes the distance between the ship position in the prediction interval and the corresponding point on the reference trajectory, as well as the size of the control quantity and the energy consumption.

[0161] The general tasks performed by the Net MPC in the ship collision avoidance and control method are as follows:

[0162] Step 1: Set the reference trajectory and the initial state Xd(0), generate the reference trajectory using the surface ship model, and set the initial state of the actual surface ship as X(0);

[0163] Step 2: Generate the prediction horizon Tp and the sampling interval of MPC iteration. Considering the sampling mechanism of the actual system, the control input is defined as the state value of formula (24), which is solved by the fourth-order or fifth-order Runge-Kutta algorithm;

[0164] Step 3: Set the objective function of the MPC constrained optimization problem for the i-th ship as the current tk:

[0165]

[0166] where Tp is the prediction horizon, Xe(t; tk) is the prediction of the error state of the system at time tk, Ue(t; tk) is the prediction of the error input of the system at time tk, Q is the weighting matrix of the error state, F is the weighting matrix of the input error and P is the weight matrix of the terminal error state. The objective function minimizes the distance between the ship position in the prediction interval and the corresponding point on the reference trajectory, as well as the size of the control and the energy consumption.

[0167] Step 4: Set the state, input and terminal constraints of the MPC optimization problem based on ship collision avoidance. The dynamic behavior of the collision avoidance ship is constrained; the state is x, y, u, v, r. The input refers to the input force τ u and the input torque τ r . The terminal constraint is the dynamic behavior of the collision avoidance ship, such as the DCPA between the own ship and the surrounding target ship in formula (18) being greater than 0.5 nautical miles.

[0168] Step 5: Solve the MPC constrained optimization problem of tk. Calculate formula (35) and minimize J. Thus, the optimal control sequence predicted at time tk is obtained;

[0169] Step 6: When the new sampling time tk+1 is reached, the new sampling state is used as the initial state to construct the MPC constrained optimization problem at this new sampling time;

[0170] Step 7: Apply the optimal input to the ship;

[0171] Step 8: Communicate relevant data. The ships send their new predictions among them. Each ship knows which ships will track according to the received predictions. According to the assigned priority, the ships take action strategies to maintain tracking or avoid collision;

[0172] Step 9: The iteration is completed until the tracking control process is completed.

[0173] In this embodiment, the collision avoidance coordination unit comprehensively evaluates the DCPA, TCPA, turning distance, distance to the original course and the parameter of prohibiting the ship from turning to the side of the wind farm, and establishes a collision risk assessment model. By using the risk function established by the above factors, the risk degree related to the target ship is objectively and effectively evaluated, and reasonable suggestions are provided for safe collision avoidance.

[0174] Based on the five parameters described above, the population fitness is defined as follows:

[0175]

[0176] where x1 represents the turning angle, x2 is the recovery time, w1, w2, w3 and w4 are weights. The objective function f is to maximize the DCPA and minimize the distance from the original route.

[0177] In order to increase the population diversity, quantum particle swarm based on Sobol random sequence is introduced in EPO. Sobol sequence has shorter calculation period, faster sampling speed and higher efficiency in processing high-dimensional sequence. Therefore, Sobol sequence is used to draw the initial population. Let the value range of the optimal solution be [xmin, xmax], then the random number kn∈[0, 1] generated by Sobol sequence is applied as follows:

[0178] x n =x min +k n (x max -x min ) (37)

[0179] Firstly, Sobol sequence is used to initialize the population to ensure its diversity. Secondly, when the particles fly across the boundary and fly out of the search domain, the diversity of the population will be reduced with the increase of particles at the boundary, which will inevitably affect the global search ability of the algorithm; therefore, boundary mutation is applied to the particles. Random numbers generated by Sobol sequence. After this processing, the diversity of the population is improved again to a greater extent; thus, the global search ability of the algorithm is improved.

[0180] The above only describes the preferred embodiments of the present application, and any equivalent changes and modifications made within the scope of the patent application of the present application shall be covered by the present application.

Claims

1. A ship-shore cooperative energy-saving collision avoidance berthing system with adaptive time-domain MPC, characterized in that, The underactuated ship prediction control unit, the collision avoidance coordination unit and the automatic berthing control unit; the underactuated ship prediction control unit is based on an energy-saving robust nonlinear model prediction control of an adaptive prediction horizon, so as to process actual control and state constraints in a berthing process based on ship-shore cooperation; the collision avoidance coordination unit is based on an improved emperor penguin optimizer, so as to realize conflict avoidance decision; and the automatic berthing control unit realizes different berthing methods by using a robust model prediction control under different disturbances; The control strategy of the underactuated ship prediction control unit is specifically as follows: Step S1: based on a preset berthing position and an expected ship berthing speed, in combination with a system model, system constraints and optimization indexes, an MPC constraint optimization problem is constructed with energy consumption as one of the targets; Step S2: a ship motion reference trajectory is obtained, an underactuated ship motion model is constructed, motion control data are predicted according to the motion reference trajectory, and the motion control data are used to control the motion state to obtain motion data of the underactuated ship; Step S3: the offset size of the underactuated ship motion data and the motion reference trajectory is compared, and motion adjustment data are generated according to the comparison result of the offset size, the motion data are used to adjust the motion state, and the offset between the underactuated ship motion data and the motion reference trajectory is calculated; Step S4: input force and torque of the controller are generated according to the comparison result of the offset; Step S5: constraint conditions are added in the prediction control, including state change constraints and ship collision avoidance; an emperor penguin optimization is used to iteratively solve the optimization problem with constraints until the optimal solution is converged; Step S6: whether the ship has a collision risk is monitored by the shore station; anti-collision instructions are generated according to the navigation conditions of the ship, so as to realize real-time interaction, dynamic adjustment, reasonable division of work and cooperative control between the shore and the ship; The collision avoidance coordination unit is based on an anti-collision decision of an improved emperor penguin optimization algorithm EPO, and specifically comprises the following steps: A1: a collision risk assessment model is established by considering a closest approach distance DCPA, a closest approach time TCPA, a turning distance and a distance to the original route: (16) Wherein x1 represents a turning angle, x2 is a recovery time, w1, w2, w3 and w4 are weights; and a target function f is to maximize the DCPA and minimize the distance to the original route; A2: a quantum particle swarm based on a Sobol random sequence is introduced into the EPO to obtain an improved EPO, and an optimal anti-collision scheme is obtained based on the improved EPO solving formula (16).

2. The ship-shore cooperative energy-saving collision avoidance berthing system with adaptive time-domain MPC according to claim 1, characterized in that, The MPC constrained optimization problem, in particular, is: k The objective function of the ship is set as: where T p is the prediction horizon, is the time t k the prediction of the error state of the system, is the prediction of the error input of the system at time t k Q is the weighting matrix of the error state, F is the weighting matrix of the input error, and P is the weight matrix of the terminal error state; The objective function minimizes the distance between the ship position in the prediction horizon and the corresponding point on the reference trajectory, as well as the magnitude of the control and the energy consumption; input U, desired input U d , input error U e and the coefficient matrix K as follows: (1) (2) (3) (4) Let η be the position and attitude vector, is the velocity vector, u is the forward velocity, v is the sway velocity, r is the yaw velocity, x is the forward position, y is the sway position, and Ψ is the yaw angle, then the underactuated ship motion model is: (5) (6) (7) Wherein R is a rotation matrix: (8) The dynamics model of the underactuated ship is calculated as follows: (9) The velocity equation is expressed as follows: (10) where τu represents the control input in the forward direction, τr represents the control input in the yaw direction; m 11 , m 22 , and m 33 represent the first, second, and third elements on the diagonal of the inertia matrix, respectively; d 11 , d 22 , and d 33 represent the first, third, and second elements on the diagonal of the linear hydrodynamic damping parameter matrix, respectively; X represents the position and velocity vector, Fwx represents the wind disturbance in the forward direction; F wy is the wind disturbance in the yaw direction, N w is the wind in the yaw, F cx is the current disturbance in the forward direction, F cy is the current disturbance in the yaw, N c is the current disturbance in the yaw; X d is expressed as the desired position and velocity vector, the components of which are given by the following equations (11) (12)。 3. The ship-shore cooperative energy-saving collision avoidance berthing system with adaptive time-domain MPC according to claim 2, characterized in that, The energy consumption is specifically as follows: (13)。 4. The ship-shore cooperative energy-saving collision avoidance berthing system with adaptive time-domain MPC according to claim 1, characterized in that, The step S3 is specifically as follows: x d is the desired forward position, y d is the desired yaw position, is the desired yaw angle, u d is the desired forward velocity, v d is the desired yaw velocity, r d is the desired yaw angle velocity; Further, let x e be the forward position error, y e be the roll position error, be the yaw angle error, u e be the forward velocity error, v e be the roll velocity error, r e be the yaw angular velocity error, then X e is given by (14) The error dynamics is written as: (15)。 5. The ship-shore cooperative energy-saving collision avoidance berthing system with adaptive time-domain MPC according to claim 1, characterized in that, The MPC constraint optimization problem is further provided with an adaptive prediction range, and specifically comprises the following steps: where T pmin is a lower limit value of the prediction time domain, T pmax is an upper limit value of the prediction time domain, and e represents an exponential function.

6. The ship-shore cooperative energy-saving collision avoidance berthing system with adaptive time-domain MPC according to claim 1, characterized in that, Let the velocity of the host ship be , the velocity of the target ship be , and the flow velocity vector of the target ship relative to the host ship be (17) The relative heading: (18) The calculation is as follows: (19) wherein R T represents a distance, the position of the main ship is , the position of the target ship is , is meant by azimuth is calculated as follows: (20) (21) The distance DCPA to the closest approach point, reflecting the highest acceptable navigation risk, is calculated as follows: (22) The time TCPA to the closest approach point is calculated as follows: (23) The conditions of ship encounter are defined as follows: (24)。 7. The ship-shore cooperative energy-saving collision avoidance berthing system with adaptive time-domain MPC according to claim 1, characterized in that, The berthing includes direct berthing and turning approach berthing.

Images