Inland river bidirectional channel ship driving route autonomous collision avoidance optimization method based on BAS algorithm
Patent Information
- Application Number
- CN202511102973.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-07
- Publication Date
- 2025-11-14
Smart Images
Figure CN120949771A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of ship navigation and autonomous collision avoidance technology, and more specifically to an autonomous collision avoidance optimization method for ship navigation routes in inland waterways based on the BAS algorithm. Background Technology
[0002] Inland waterways allow vessels to navigate simultaneously in both directions, but they have high requirements for parameters such as width and turning radius. Vessels are significantly affected by ship-to-ship effects and bank effects. In the complex environment of inland waterways, vessels are prone to collisions due to natural factors (such as strong winds and high waves leading to reduced visibility), waterway factors (such as inadequate navigational aids and insufficient width), vessel factors (such as excessive traffic flow and inappropriate speed), and human factors (such as illegal navigation and communication abnormalities), resulting in casualties and property damage.
[0003] Currently, autonomous collision avoidance optimization methods for ship navigation routes mainly include methods based on MPC (Model Predictive Control), methods based on swarm intelligence optimization algorithms, methods based on artificial intelligence technology, and methods based on neural network sliding mode. However, these methods have many shortcomings: the collision avoidance model constructed by the MPC execution method is easily affected by changes in random search direction, leading to inconsistencies between the navigation state and the ideal state; methods based on swarm intelligence optimization algorithms are easily affected by smoothing parameters, resulting in deviations between the navigation state and the ideal state; methods based on artificial intelligence technology are easily affected by changes in rudder angle, resulting in excessive deviations between the navigation state and the ideal state; and methods based on neural network sliding mode are easily affected by relative velocity deviation problems, resulting in a mismatch between the navigation state and the ideal state.
[0004] Based on this, the present invention proposes an autonomous collision avoidance optimization method for vessel navigation routes in inland waterways based on the BAS algorithm to solve the above problems. Summary of the Invention
[0005] To overcome the aforementioned deficiencies in the prior art, this invention provides an autonomous collision avoidance optimization method for vessel routes in inland waterways based on the BAS algorithm, in order to solve the problems existing in the background art.
[0006] This invention provides the following technical solution: an autonomous collision avoidance optimization method for vessel routes in inland waterways based on the BAS algorithm, comprising the following steps:
[0007] S1: Determine the relative speed and heading angle of the vessel to determine the dynamic navigation risk level;
[0008] S2: Search in continuous spatial coordinates and combine the fitness function to determine the update position of the objective function value;
[0009] S3: Use the BAS algorithm to compare fitness states, update the pheromone matrix, and determine the historical optimal target value;
[0010] S4: Generate a set of reference points based on the locations of ship conflicts, and use the Ackley / Rosenbrock function to perform a decision search;
[0011] S5: Under a finite state machine, complete the dynamic scene transition, generate autonomous collision avoidance decision constraints, correct the deviation parameters of route uncertainty, and complete the autonomous collision avoidance optimization of the inland waterway vessel route.
[0012] By coordinating dynamic risk assessment, BAS algorithm optimization search, and decision constraint generation, autonomous collision avoidance optimization of vessel routes in inland waterways is achieved, ensuring that the collision avoidance route always deviates from obstructions, and that rudder angle, heading, and distance are all within safe ranges. This solves the problem of inconsistency between the actual operating state and the ideal state in traditional methods, and improves the safety and reliability of vessel navigation.
[0013] As a further aspect of the present invention: In S1, the specific process of determining the dynamic navigation risk degree is as follows: the distance between the two ships is calculated by using the velocity vector components of the ships, and the dynamic navigation risk degree is determined by combining the risk membership weights; wherein, the calculation formula for the velocity vector components is:
[0014]
[0015] Among them, V o δ0 represents the preset initial speed of the ship, V represents the ship's heading, and V represents the initial speed. T δ represents the target ship's speed. T Represents the target ship's heading;
[0016] The formula for calculating the distance between two ships is:
[0017]
[0018] Among them, X T Y T X0 and Y0 represent the x and y coordinates of the target point, respectively;
[0019] The formula for calculating the risk membership weight is:
[0020] CRI=W dcp +W tcp +W D +W B +W K
[0021] Among them, W dcp W represents the DCPA metric. tcp Representing the TCPA metric, WD W represents the relative distance index. B W represents a relative orientation indicator. K This represents the ship speed ratio.
[0022] By quantifying ship speed vectors, distance, and multi-dimensional risk indicators (DCPA, TCPA, etc.), a precise assessment of dynamic navigation risk is achieved, providing a scientific basis for subsequent collision avoidance decisions, avoiding subjectivity in risk assessment, ensuring early identification of collision hazards, and improving the pertinence and timeliness of collision avoidance decisions.
[0023] As a further aspect of the present invention: In S2, the specific process of determining the update position of the objective function value is as follows: assuming an n-dimensional search space, determining the orientation and change state of the longhorn beetle's antennae head, obtaining a random search direction, and then determining the positions of the longhorn beetle's left and right antennae, combining the fitness function to determine the update position of the objective function value; wherein, the calculation formula for the random search direction is:
[0024]
[0025] Where M represents the original search location;
[0026] The formula for calculating the update position of the objective function value is:
[0027] x n+1 =x n -S n ·dir·sign[f(x l )-f(x r )]
[0028] Where, x n S represents the direction of the beetle's center of mass. n f(x) represents random fitness. l f(x) represents the odor intensity of the left antennae. r () represents the odor intensity of the right antenna.
[0029] By simulating the beetle's antennae search mechanism using the BAS algorithm, the target function update position is efficiently located in continuous space. Combined with the fitness function, the collision avoidance path is dynamically optimized, avoiding the blindness of traditional search methods and improving the efficiency and accuracy of path search, thus laying the foundation for finding the optimal collision avoidance route.
[0030] As a further aspect of the present invention: In S3, the specific process of determining the historical optimal target value is as follows: using the BAS algorithm to search, comparing fitness states, updating the pheromone matrix, determining the termination condition, calculating the fitness factor, and combining the fitness factor to determine the historical optimal target value; wherein, the formula for calculating the fitness factor is:
[0031]
[0032] Where α represents the target update factor, θ represents the number of iterations, and n represents the iteration step size factor.
[0033] By comparing fitness states and updating the pheromone matrix using the BAS algorithm, the historical optimal target value is tracked, effectively overcoming the shortcomings of traditional methods that are prone to getting trapped in local optima. This ensures that the final collision avoidance path is the global optimal solution, thus improving the reliability and stability of path optimization.
[0034] As a further aspect of the present invention: In S4, the specific process of using the Ackley / Rosenbrock function for decision search is as follows: A set of reference points is generated based on the ship conflict positions; the constraint step size is determined using the Ackley / Rosenbrock function; the original heading boundary is identified; and the ship navigation coordinate function and heading expression are obtained. The calculation formula for the ship navigation coordinate function is:
[0035] ξ=xcosγ+ycosγ+μsinγ
[0036] Where x and y represent the real-time navigation coordinates of ships in a two-way channel, μ represents the encounter angle, and γ represents the wave direction angle;
[0037] The formula for calculating the heading expression is:
[0038]
[0039] Among them, V O VK represents the decision barrier angle, and VK represents the spatial velocity reference value.
[0040] By combining the characteristics of the Ackley / Rosenbrock function, a precise decision search is performed on the reference point of the ship conflict position, clarifying the course boundary and constraint step size, which improves the adaptability of collision avoidance decision in complex navigation environments (such as changes in encounter angle and wave direction angle), and ensures the scientific and rational nature of the decision results.
[0041] As a further aspect of the present invention: In S5, the specific process of generating autonomous collision avoidance decision constraints is as follows: when facing a head-on collision navigation scenario, ideal line-of-sight processing is performed, dynamic scene transitions are performed using a finite state machine, autonomous collision avoidance decision constraints are generated, and the deviation parameters of the route uncertainty are corrected in combination with the constraints.
[0042] By using finite state machines to achieve real-time adaptation to dynamic scenarios, targeted autonomous collision avoidance decision constraints are generated, effectively correcting course uncertainty deviations and ensuring that ships can maintain a consistent sailing state with the ideal state even in complex encounter situations (such as head-on or cross encounters), thereby improving the stability and robustness of the collision avoidance process.
[0043] The technical effects and advantages of this invention are as follows:
[0044] This invention, by introducing the global optimization capability of the BAS algorithm and combining dynamic risk assessment with a continuous space efficient search mechanism, achieves precise autonomous collision avoidance optimization of vessel routes in inland waterways. Specific technical effects and advantages are as follows:
[0045] (1) Effectively improves collision avoidance accuracy and safety: By dynamically judging the relative speed and heading angle of the ship to determine the navigation risk, and combining the efficient search capability of the BAS algorithm in continuous space, the collision avoidance path can be accurately planned. Experimental results show that after applying this method, the collision avoidance route always deviates from the navigation obstacle's route, and the rudder angle, heading, and distance are all kept outside the safe range of the navigation obstacle, ensuring that there is no risk of collision during the ship's navigation.
[0046] (2) Overcoming the shortcomings of traditional methods and ensuring that the driving state is consistent with the ideal state: In view of the problems that traditional MPC methods are easily affected by random search direction and swarm intelligence optimization algorithms are easily affected by smoothing parameters, this invention uses the BAS algorithm to compare fitness states, update the pheromone matrix and track historical optimal target values, effectively avoiding the local optimum trap. At the same time, the Ackley / Rosenbrock function is used to assist decision search and dynamic scene transition of finite state machine to accurately correct the deviation parameters of route uncertainty, so that the ship's driving state is consistent with the ideal state.
[0047] (3) Enhance the robustness and adaptability of the method: The longhorn beetle whisker search mechanism is innovatively introduced into the field of ship collision avoidance. Through pheromone matrix update and historical optimal target value guidance, the adaptability to complex inland waterway two-way navigation environment (such as the coexistence of dynamic and static obstacles and the changing encounter situation) is improved. Stable and reliable collision avoidance trajectory can be generated regardless of whether it is a dynamic obstacle or a static obstacle.
[0048] (4) Optimize decision-making efficiency and path rationality: Generate a set of reference points based on the location of ship conflicts, and perform decision search using the Ackley / Rosenbrock function. This can quickly determine the constraint step size and heading boundary. Combined with a finite state machine, autonomous decision-making constraint generation in dynamic scenarios can be achieved, ensuring the real-time nature of collision avoidance decisions and the optimality of the path, while taking into account both collision avoidance safety and driving efficiency. Attached Figure Description
[0049] Figure 1 This is a ship encounter situation diagram according to the present invention.
[0050] Figure 2 This is a graph showing the trajectory of the fitness function change in this invention.
[0051] Figure 3This is the decision search graph for the Ackley / Rosenbrock function of the present invention.
[0052] Figure 4 This is a diagram of the integrated display interface for ship navigation data according to the present invention.
[0053] Figure 5 This is a diagram illustrating the collision avoidance trajectory of a vessel navigating a two-way waterway, as described in this invention.
[0054] Figure 6 This is a diagram showing the dynamic collision avoidance results of the ship's navigation route according to the present invention.
[0055] Figure 7 This is a diagram showing the optimization results of autonomous collision avoidance based on the coupling of dynamic and static obstacles in this invention.
[0056] Figure 8 This is a map showing the positional information of vessels and obstructions in the navigation channel according to the present invention.
[0057] Figure 9 This is a flowchart illustrating the invention process of this invention. Detailed Implementation
[0058] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention will be further described in detail below. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.
[0059] Example 1:
[0060] The invention will now be further described with reference to the accompanying drawings.
[0061] 1.1 Constructing an Autonomous Collision Avoidance Model for Ship Navigation Routes Based on the BAS Algorithm
[0062] The general process of ship collision avoidance can be described as follows: target vessel detection - target vessel information analysis - collision risk assessment - collision avoidance liability determination - collision avoidance maneuver implementation - resuming navigation. This allows for a comprehensive analysis of ship encounter situations, classifying them according to scenarios such as head-on encounters and cross-encounters. The ship encounter situation is then categorized as follows: Figure 1 As shown.
[0063] Depend on Figure 1 It can be seen that when sailing to area B, this ship is a yielding ship. When a ship comes from area F, a head-on situation will occur, and all ships need to perform avoidance operations. For ships coming from other areas, this ship has no avoidance task, but needs to maintain a safe distance. At this time, the ship's navigation state can be assumed, and the velocity vector components are as follows (1).
[0064]
[0065] In formula (1), V oδ0 represents the preset initial speed of the ship, V represents the ship's heading, and V represents the initial speed. T δ represents the target ship's speed. T The target ship's heading represents the distance D between the two ships at this point. oT As shown in (2) below.
[0066]
[0067] In formula (2), X T Y T X0 and Y0 represent the horizontal and vertical coordinates of the target point, respectively. From this, the relative speed and heading angle of the ship can be determined, and the dynamic navigation risk can be determined. The risk membership weight CRI at this time is as follows (3).
[0068] CRI=W dcp +W tcp +W D +W B +W K (3)
[0069] In formula (3), W dcp W represents the DCPA metric. tcp Representing the TCPA metric, W D W represents the relative distance index. B W represents a relative orientation indicator. K The BAS algorithm is a variable sensing algorithm that can search in continuous spatial coordinates
[11] . It can assume an n-dimensional search space and determine the orientation and change state of the longhorn beetle's antennae based on the objective function solution. The random search direction dir is as shown in (4).
[0070]
[0071] In formula (4), M represents the original search position, which determines the positions of the left and right antennae of the longhorn beetle. Combined with the fitness function, the target function value update position x is determined. n+1 As shown in (5) below.
[0072] x n+1 =x n -S n ·dir·sign[f(x l )-f(x r (5)
[0073] In formula (5), x n S represents the direction of the beetle's center of mass. n f(x) represents random fitness. l f(x) represents the odor intensity of the left antennae. rThe ) represents the odor intensity of the right antenna, from which the step size can be updated
[14] . After returning, the iteration continues, and the trajectory of its fitness function change is as follows: Figure 2 As shown.
[0074] Depend on Figure 2 It can be seen that the two sides of the trapezoid represent the directions of the longhorn beetle's tentacles, and the bottom of the trapezoid represents the distance between the tentacles. This complete curve can reflect the trajectory of the fitness function. The autonomous collision avoidance modeling steps are as follows: quantify the actual problem into an effective model, perform parameter initialization iteration, search the effective space range, perform fitness evaluation, use the BAS algorithm for search, compare the fitness states, update the pheromone matrix, then determine the termination condition, and calculate the fitness factor at this time.
[0075] β, as shown in (6).
[0076]
[0077] In formula (6), α represents the target update factor, θ represents the number of iterations, and n represents the iteration step size factor. Combined with the fitness factor, the historical optimal target value can be determined. The constructed autonomous collision avoidance model W for ship navigation route is shown in (7).
[0078]
[0079] In formula (7), R E The parameter representing the centroid coordinate update, combined with the collision avoidance model, can determine the output beetle position and make it the optimal solution, serving as a reference for subsequent obstacle avoidance decision constraints.
[0080] 1.2 Solving the Autonomous Collision Avoidance Decision Scheme for Vessel Navigation Routes in Inland Two-Way Waterways
[0081] During collision avoidance on a ship's course, the influence of multiple convergence effects may lead to control dynamics problems. The autonomous obstacle avoidance model constructed in section 1.1 can be used to determine absolute velocity obstacles and impose autonomous obstacle avoidance decision constraints. First, a set of reference points can be generated based on the ship's collision location. Then, the Ackley / Rosenbrock function is used for decision search, as follows: Figure 3 As shown.
[0082] Depend on Figure 3 As can be seen, the constraint step size can be determined by combining the function, and the original course boundary can be identified. The ship navigation coordinate function ξ at this time is shown in (8).
[0083] ξ=xcosγ+ycosγ+μsinγ 8)
[0084] In formula (8), x and y represent the real-time navigation coordinates of ships in a two-way channel, μ represents the encounter angle, and γ represents the wave direction angle. The resulting heading expression C0 is shown in (9).
[0085]
[0086] In formula (9), V O VK represents the decision obstacle angle, and VK represents the spatial velocity reference value. In the face-to-face navigation scenario, ideal line-of-sight processing can be performed. A finite state machine is used to handle dynamic scene transitions and generate an autonomous collision avoidance decision scheme, which satisfies the autonomous collision avoidance decision constraint U. rout As shown in (10) below.
[0087]
[0088] In formula (10), l0 represents the collision avoidance decision constraint offset weight, which, combined with the constraint formula, can correct the deviation parameter of the route uncertainty.
[0089] 2 Experiments
[0090] This experiment fully considers the characteristics of ship navigation models. By setting ship navigation data and the position information of ships and obstacles, it analyzes the ship's collision avoidance trajectory, dynamic collision avoidance effect, and autonomous collision avoidance optimization effect, and verifies the effectiveness of the autonomous collision avoidance optimization method for ship navigation routes in inland waterways based on the BAS algorithm.
[0091] The system uses devices such as GPS receivers, speedometers, compasses, and anemometers to collect ship navigation data, which is then displayed on a comprehensive interface. Figure 4 As shown.
[0092] Depend on Figure 4 As can be seen, the central section displays information such as rudder angle and thrust, reflecting the current operational status of the vessel. Collision avoidance maneuvers are achieved by adjusting the rudder angle, course, and distance. In the initial state, the positional information of the vessel and the obstruction in the channel is as follows: Figure 8 As shown.
[0093] like Figure 8 As shown, case-i, case-ii, case-iii, case-iiii, case-iiiii, and case-iiiiii represent the positions of the rudder angle and heading adjusted four times. Based on the initial position, berthing position, and the position of the obstruction, the ship's route is predicted, thereby achieving early collision avoidance.
[0094] The safe distance between the vessel and the obstruction was set at 70m. Different starting positions were selected to analyze collision avoidance trajectories in a two-way waterway. Figure 5 As shown.
[0095] Depend on Figure 5It can be seen that A, B, and C are three types of obstacles to navigation. (a) is the obstacle avoidance trajectory of a two-way channel for a dynamic obstacle; (b) is the obstacle avoidance trajectory of a channel for a static obstacle. After the ship's navigation route collision avoidance optimization, the ship can stay away from the obstacle and find a safer collision avoidance trajectory, with good collision avoidance effect.
[0096] Under experimental conditions, the present invention plans the ship's route in the lower left corner of the grid map, and sets the target point in the upper right corner. In the experiment, the optimal route for ships in the waterway is planned, and the dynamic collision avoidance of the route is determined. Figure 6 As shown.
[0097] Depend on Figure 6 It can be seen that the gray grid represents dynamically moving obstacles, and the orange grid represents static obstacles. The red curve represents the optimal driving route, and the green curve represents the actual driving route. (1) is the optimal driving route map; (2) is the route for the first approach to the dynamic obstacle; (3) is the route for the second approach to the dynamic obstacle; (4) is the route for the third approach to the dynamic obstacle. There were no collisions with obstacles or navigational obstructions along the route from the starting point to the ending point, and the collision avoidance effect was good.
[0098] In a ship's navigation path, distance, heading, and rudder angle reflect the effectiveness of autonomous collision avoidance optimization. Based on changes in the distance between the ship and obstructions, the rudder angle and heading are adjusted before exceeding a safe distance. The effectiveness of autonomous collision avoidance optimization during this process is analyzed. Figure 7 As shown.
[0099] Depend on Figure 7 As can be seen, this invention utilizes the BAS algorithm to simulate the way a longhorn beetle senses its environment and searches for food. In complex navigation environments, it searches for the closest or optimal collision avoidance path from the current position to the target, ensuring the safety of ship navigation. The final optimization results show that the optimized collision avoidance route always deviates from the obstruction's path, with rudder angle, heading, and distance all deviating from the obstruction's safe range, demonstrating good autonomous collision avoidance optimization of the ship's navigation path.
[0100] Example 2:
[0101] In this embodiment, two ships traveling in opposite directions (the ship itself and the target ship) in a two-way inland waterway are taken as the research object. The initial position of the ship itself is (393, 625, 1 rad), and the position of the target ship (obstacle) is (712, 523). The safe distance is set to 70 m. The specific collision avoidance optimization steps are as follows:
[0102] S1: Determine the ship's relative speed and heading angle to ascertain the dynamic navigation risk level.
[0103] Collect real-time navigation data of this ship and the target ship: the original speed of this ship V o=10m / s, heading δ0=1rad; target ship speed V T =8m / s, heading δ T = -1 rad.
[0104] Calculate the velocity components using the velocity vector component formula:
[0105] Own ship speed component: V ox =ν o ·cos(δ0)=10×cos(1)≈5.403m / s, V oy =ν o ·sin(δ0)=10×sin(1)≈8.415m / s;
[0106] Target ship velocity component: V Tx =ν T ·cos(δ T V = 8 × cos(-1) ≈ 4.322 m / s Tx =ν T ·cos(δ T =8×sin(-1)≈-6.732m / s.
[0107] Calculate the real-time distance D between the two ships. oT Substituting the coordinates of our ship (393, 625) and the coordinates of the target ship (712, 523), we get...
[0108]
[0109] Calculate the Risk Membership Weight (CRI): Combining DCPA (Distance to Nearest Encounter), TCPA (Time to Nearest Encounter), relative distance WD, relative bearing WB, and speed ratio WK, we get CRI = W dcp +W tcp +W D +W B +W K =0.3+0.2+0.15+0.2+0.15=1.0, the current risk level is determined to be medium, and the collision avoidance mechanism needs to be activated.
[0110] S2: Search in continuous spatial coordinates and determine the update position of the objective function value by combining the fitness function.
[0111] Assuming the search space is two-dimensional (n=2), the initial position of the longhorn beetle's antennae is the original search position M=(393,625), and a random search direction is generated.
[0112] Calculate the positions of the longhorn beetle's left and right antennae: Left antenna x l =(393+0.3×S) n 625+0.4×Sn ), right antenna x r =(393-0.3×S) n 625-0.4×S n (S) n =5 is the initial step size.
[0113] The fitness function was used to evaluate the "odor intensity" of the antennae: the fitness of the left antenna was f(xl) = 0.8, and the fitness of the right antenna was f(x). r Given x = 0.6, and based on the objective function update position formula, we obtain x. n+1 =x n -S n ·dir·sign[f(x l )-f(x r )]=(393,625)-5×(0.3,0.4)×1=(391.5,623), that is, the objective function update position is (391.5,623).
[0114] S3: Use the BAS algorithm to compare fitness states, update the pheromone matrix, and determine the historical optimal target value.
[0115] Iterative search: with x n+1 Starting from a new point, repeat step S2. After each iteration, compare the fitness state. If the current fitness is higher than the historical best, update the pheromone matrix (pheromone concentration is positively correlated with fitness).
[0116] Calculate the fitness factor: When the number of iterations θ = 10, the iteration step size factor n = 2, and the target update factor α = 5, we get
[0117] Determine the historical optimal target value: When the iteration reaches the 20th iteration, the fitness function converges, and the position corresponding to the historical optimal target value is (410, 650). This position is 350m away from the target ship, and the risk level drops to 0.2.
[0118] S4: Generate a set of reference points based on the locations of ship conflicts, and perform decision search using the Ackley / Rosenbrock function.
[0119] Generate a set of reference points: Based on the historical best position and the target ship's position, the potential conflict point is predicted to be (550, 580), and a set of reference points {(540, 570), (560, 590), (550, 585)} is generated.
[0120] The decision boundary is searched using the Ackley / Rosenbrock function: After function iteration, the constraint step size is determined to be 10m, and the original heading boundary is ±0.5rad. According to the ship's navigation coordinate function, substituting the real-time coordinates x=410, y=650, the encounter angle μ=0.3rad, and the wave direction angle γ=0.2rad, we get ξ=xcosγ+ycisγ+μsinγ=(410+650)×0.980+0.3×0.198≈1039.0+0.06=1039.06.
[0121] Calculate the heading expression: Decision obstacle angle m = 3, spatial velocity reference value VK = 9m / s, get C0 = ∑(k = 1 to 3)V0VK = 3 × 10 × 9 = 270, determine the optimal heading adjustment range as 1 rad ± 0.3 rad.
[0122] S5: Complete dynamic scene transitions under a finite state machine, generate autonomous collision avoidance decision constraints, and correct for deviations in trajectory uncertainty parameters.
[0123] Dynamic scene transition: The finite state machine switches the current scene from "normal navigation" to "collision avoidance", triggering the collision avoidance decision mechanism.
[0124] Generate decision constraints: Combining the reference point and the heading boundary, generate the constraint |Δδ|≤0.3rad (Δδ is the heading adjustment amount) to ensure that the heading adjustment does not exceed the safe range.
[0125] Correction of deviation parameters: Real-time monitoring showed that the course deviated from the preset path by 2m due to water flow interference. The rudder angle was adjusted by 5° according to the constraint. After correction, the deviation between the course and the target path was less than 0.5m. Finally, the collision avoidance route was always kept at a safe distance (70m) from the target ship.
[0126] In this invention, the assessment weights of dynamic driving risk (such as the weight ratios of DCPA and TCPA indicators), the search parameters of the BAS algorithm (such as the value of the initial step size Sn and the iteration termination threshold), the decision boundaries of the Ackley / Rosenbrock function (such as the adjustment range of the constraint step size), and the scenario switching conditions of the finite state machine (such as the risk trigger threshold) are all adjustable parameterized settings. Those skilled in the art can adapt the invention by modifying relevant parameters (such as the safety distance threshold and the maximum heading adjustment angle) according to the actual environment of the inland waterway (such as channel width and water flow speed), vessel type (such as cargo ships and passenger ships), and navigation density (such as peak or off-peak periods). All such adjustments fall within the protection scope of this invention.
[0127] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. An autonomous collision avoidance optimization method for vessel navigation routes in inland waterways based on the BAS algorithm, characterized by: Includes the following steps: S1: Determine the relative speed and heading angle of the vessel to determine the dynamic navigation risk level; S2: Search in continuous spatial coordinates and combine the fitness function to determine the update position of the objective function value; S3: Use the BAS algorithm to compare fitness states, update the pheromone matrix, and determine the historical optimal target value; S4: Generate a set of reference points based on the locations of ship conflicts, and use the Ackley / Rosenbrock function to perform a decision search; S5: Under a finite state machine, complete the dynamic scene transition, generate autonomous collision avoidance decision constraints, correct the deviation parameters of route uncertainty, and complete the autonomous collision avoidance optimization of the inland waterway vessel route.
2. The autonomous collision avoidance optimization method for vessel navigation routes in inland waterways based on the BAS algorithm according to claim 1, characterized in that: In S1, the specific process for determining the dynamic navigation risk level is as follows: The distance between two ships is calculated using the ship's velocity vector components, and the dynamic navigation risk level is determined by combining this with the risk membership weights; wherein, the formula for calculating the velocity vector components is: Among them, V o δ0 represents the preset initial speed of the ship, V represents the ship's heading, and V represents the initial speed. T δ represents the target ship's speed. T Represents the target ship's heading; The formula for calculating the distance between two ships is: Among them, X T Y T X0 and Y0 represent the x and y coordinates of the target point, respectively; The formula for calculating the risk membership weight is: CRI=W dcp +W tcp +W D +W B +W K Among them, W dcp W represents the DCPA metric. tcp Representing the TCPA metric, W D W represents the relative distance index. B W represents a relative orientation indicator. K This represents the ship speed ratio.
3. The autonomous collision avoidance optimization method for vessel navigation routes in inland waterways based on the BAS algorithm according to claim 1, characterized in that: In S2, the specific process for determining the update position of the objective function value is as follows: Assuming an n-dimensional search space, determine the orientation and change state of the longhorn beetle's antennae head to obtain a random search direction, then determine the positions of the longhorn beetle's left and right antennae, and combine this with the fitness function to determine the update position of the objective function value; wherein, the formula for calculating the random search direction is: Where M represents the original search location; The formula for calculating the update position of the objective function value is: x n+1 =x n -S n ·dir·sign[f(x l )-f(x r )] Where, x n S represents the direction of the beetle's center of mass. n f(x) represents random fitness. l f(x) represents the odor intensity of the left antennae. r () represents the odor intensity of the right antenna.
4. The autonomous collision avoidance optimization method for vessel navigation routes in inland waterways based on the BAS algorithm according to claim 1, characterized in that: In S3, the specific process for determining the historical optimal target value is as follows: The BAS algorithm is used for searching, fitness states are compared, the pheromone matrix is updated, termination conditions are determined, the fitness factor is calculated, and the historical optimal target value is determined based on the fitness factor; wherein, the formula for calculating the fitness factor is: Where α represents the target update factor, θ represents the number of iterations, and n represents the iteration step size factor.
5. The autonomous collision avoidance optimization method for vessel navigation routes in inland waterways based on the BAS algorithm according to claim 1, characterized in that: In S4, the specific process of using the Ackley / Rosenbrock function for decision search is as follows: A set of reference points is generated based on the ship's conflict location; the constraint step size is determined using the Ackley / Rosenbrock function; the original heading boundary is identified; and the ship's navigation coordinate function and heading expression are obtained. The formula for calculating the ship's navigation coordinate function is: ξ=xcosγ+ycosγ+μsinγ Where x and y represent the real-time navigation coordinates of ships in a two-way channel, μ represents the encounter angle, and γ represents the wave direction angle; The formula for calculating the heading expression is: Where VO represents the decision barrier angle and VK represents the spatial velocity reference value.
6. The autonomous collision avoidance optimization method for vessel navigation routes in inland waterways based on the BAS algorithm according to claim 1, characterized in that: In S5, the specific process of generating autonomous collision avoidance decision constraints is as follows: when facing a head-on collision navigation scenario, ideal line-of-sight processing is performed, dynamic scene transitions are performed using a finite state machine, autonomous collision avoidance decision constraints are generated, and the deviation parameters of the route uncertainty are corrected in combination with the constraints.