Vehicle path planning method and system in narrow and long deck space of roll-on-roll-off ship

Abstract

The invention belongs to the technical field of path planning, and discloses a vehicle path planning method and system in a narrow and long deck space of a roll-on-roll-off ship, and the method comprises the steps: inputting vehicle information, a starting position and a target position after setting algorithm parameters, alternately expanding a starting tree and a target tree through employing an improved Bi-RRT algorithm, and obtaining a path of a vehicle in a narrow and long deck space of the roll-on-roll-off ship; and node optimization and path smoothing are carried out to finally generate a vehicle path. According to the invention, based on a target bias sampling strategy of obstacle vertex extension, the obstacle avoidance capability in the search process can be enhanced; redundant nodes are screened out twice, so that the node quality is effectively ensured, and the rapid generation of a vehicle path can be ensured; the expansion search capability of the node is improved by utilizing a variable step length (large step length, medium step length and small step length) mode; finally, the path is smoothed in an arc optimization mode, and it is guaranteed that the path conforms to vehicle driving characteristics. The method can improve the algorithm search speed and the obstacle avoidance capability, and enables the vehicle to quickly generate a smooth path from the starting point to the target point.

Description

Technical Field

[0001] The present invention belongs to the technical field of path planning, and particularly relates to a vehicle path planning method and system in the narrow and long deck space of a ro-ro ship. Background Art

[0002] Path planning is an important means to improve the vehicle loading efficiency of a ro-ro ship, ensure loading safety, and optimize the loading path. Inside a ro-ro ship, due to the narrow and long deck space and the presence of obstacles such as columns and fixed equipment, path planning faces many problems. The narrow and long space limits the turning radius and driving flexibility of vehicles, making route design more complex; columns and equipment, as obstacles, increase the need for obstacle avoidance, and it is necessary to ensure bypassing obstacles and maintaining a safe distance. At the same time, the narrow passage requires precise control of the vehicle's position and direction to avoid collisions. Vehicle path planning in the narrow and long space of a ro-ro ship mainly ensures the ability of the vehicle to quickly generate a safe path to near the specified position in the narrow and long deck space environment, so as to improve the real-time performance and safety of the planning.

[0003] The prior art speeds up the search by means of a bias sampling strategy or a regional constraint strategy, but the utilization rate of the search space is low, and it is easy to fall into a dead zone in a complex and narrow area; a RRT algorithm with a dynamic step size is proposed, although the obstacle avoidance ability of the algorithm is improved, but only a simple obstacle environment is considered. The path after the above optimization still has problems such as a long search time and low path quality during the generation process, mainly due to the influence of the sampling method, the expansion step size, and node optimization, and still needs to be further improved. At the same time, the research object of the prior art has relatively few deck vehicles, and the non-holonomic constraints of the vehicles are not fully considered.

[0004] The Bidirectional Rapid Search Random Tree (Bi-RRT algorithm) is an improvement of the Rapid Search Random Tree. This algorithm grows and expands from the start point and the end point respectively through two trees and considers the vehicle inflated into a circle, and searches in an unknown space environment, greatly improving the vehicle path planning efficiency, and is very suitable for unknown and complex narrow and long deck complex environments.

[0005] The idea of the Bi-RRT algorithm is to grow and expand nodes from the start point and the end point respectively through two trees. If an obstacle is encountered, a new node is reselected, that is, a new sampling point is selected within the reach of the search space to avoid the obstacle. Until the distance between the nodes of the two trees reaches a certain value, the nodes of the two trees are connected and the path planning is completed. However, due to the randomness of the expansion method, the fixedness of the expansion node step size, and the incomplete node optimization, the current algorithm has problems such as low search efficiency, slow node expansion, and redundant nodes, seriously affecting the real-time performance and accuracy of path planning.

[0006] At present, although a series of improvements have been made to the existing Bi-RRT algorithm, it is less relevant to the problems and vehicles and thus inapplicable. For example, the existing patent CN114545931A discloses a path planning method for a surface unmanned boat based on an improved Bi-RRT algorithm guided by an artificial potential field method, which mainly targets surface unmanned boats and uses a target bias sampling strategy and an artificial potential field method for a set obstacle area to improve the expansion efficiency. However, vehicles have higher requirements for path smoothness and obstacle distance accuracy compared to unmanned boats, and land obstacles are more complex, which is different from the problem analysis of the present invention; the existing patent CN116784975A discloses a flexible puncture needle path planning method based on an improved Bi-RRT algorithm, which mainly targets a medical puncture path model and considers expanding the obstacle by 1.5 times. However, this method is not applicable to vehicle path planning because obstacle inflation will cause a significant decrease in accuracy. The existing patent CN118672273A discloses a robot and its adaptive path planning method and device, control method and device, which mainly analyzes robots and introduces the APF algorithm. Although it has certain improvements compared to traditional algorithms, the vehicle path planning problem needs to consider factors such as turning radius, so it is different from the present invention.

[0007] Through the above analysis, the problems and defects existing in the prior art are as follows:

[0008] (1) The Bi-RRT path planning algorithm has too much randomness in the sampling method, which will cause problems such as blind expansion of nodes and poor obstacle avoidance ability. Although some current studies have been improved by combining methods such as the artificial potential field method, the target bias sampling strategy, and setting obstacle areas, the following problems exist: it is easy to fall into dead zones in complex environments, the space utilization rate is low, and infeasible paths are easily generated. At the same time, most of the research objects are not vehicles, and conditions such as dynamic constraints, distance accuracy, and non-holonomic constraints are not comprehensively considered, so it is not applicable to the deck environment. Therefore, the sampling method of the Bi-RRT algorithm still needs to be further improved in vehicle path planning.

[0009] (2) During the node search and expansion process of the Bi-RRT path planning algorithm, the fixed expansion step size results in low expansion ability of nodes in different regions. The deck space of a roll-on / roll-off ship is relatively special, with a long and narrow internal space and a large number of obstacles such as columns and fixed facilities distributed. There is relatively little research and improvement on variable step sizes for vehicle path planning within the deck of a roll-on / roll-off ship. It is necessary to improve the node expansion step size in vehicle path planning to increase the search speed.

[0010] (3) There are often too many redundant nodes among the path nodes obtained after the search is completed, and the effect after the paths are connected is poor. Although there are currently methods to optimize by reselecting the parent node and reconnecting the nodes, this method is relatively one-sided, and the effect of this method is not obvious after the algorithm is improved. Therefore, node optimization needs to be studied and improved according to specific improvement strategies. Summary of the Invention

[0011] To overcome the problems existing in the related art, the disclosed embodiments of the present invention provide a vehicle path planning method and system in the narrow and long deck space of a ro-ro ship. The technical problems solved by the present invention are as follows: blind expansion of nodes, poor obstacle avoidance ability; slow expansion speed of nodes in different regions; screening of redundant nodes, and path smoothing problems.

[0012] The technical solution is as follows: The present invention is implemented as a vehicle path planning method in the narrow and long deck space of a ro-ro ship, including the steps of:

[0013] S1, after setting the algorithm parameters, input vehicle information, as well as the starting position and the target position;

[0014] S2, based on the set parameters, conduct path node search, and use the improved Bi-RRT algorithm to alternately expand the starting tree and the target tree; combined with the Bi-RRT algorithm search, a heuristic sampling strategy based on the local expansion method of the obstacle vertex outward expansion is proposed to solve the problems of blind node expansion and poor obstacle avoidance ability; a variable step size strategy is adopted to solve the problem of slow expansion speed of nodes in different regions;

[0015] S3, perform primary node optimization, secondary node optimization, and arc optimization to smooth the path and generate the vehicle path.

[0016] In step S1, the vehicle information V has a total of 4 parameters, which are successively represented as:

[0017] Starting point (x start , y start ), where x start is the abscissa of the starting point, and y start is the ordinate of the starting point;

[0018] End point (x goal , y goal ), where x goal is the abscissa of the end point, and y goal is the ordinate of the end point;

[0019] Vehicle expansion circle radius r;

[0020] Vehicle minimum turning circle radius R;

[0021] The above parameters are all scalars, all length units are m, and all coordinates are (x, y).

[0022] In step S2, when conducting path node search, it includes: when the number of iterations is even, expand the starting tree T start , and when it is odd, expand the target tree T goal , and select the step size H for expansion;

[0023] The step size H is divided into three categories: large, medium, and small. When the distance d between the vehicle center and the obstacle is between r - 2r, the small step size h is used; when the distance d between the vehicle center and the obstacle is between 2r - 3r, the medium step size 2h is used; when the distance d between the vehicle center and the obstacle is greater than 3r, the large step size 3h is used. The expression is as follows:

[0024]

[0025] In the formula, h is the set step size, 2h is twice the set step size, 3h is three times the set step size, r is the radius of the vehicle's expansion circle, and d is the distance between the vehicle center and the obstacle;

[0026] When d < r, it indicates that the expanded node collides with the obstacle, and the node expansion is restarted.

[0027] Furthermore, according to the selected step size H, a new node N node is expanded from the node N new , and the target bias sampling strategy is used to expand the new node. The target bias sampling strategy is defined as:

[0028]

[0029] In the formula, q rand is a random point, p rand is a random number, q is the expansion direction point, and p 0 is the bias sampling probability;

[0030] A random point q rand is generated in each iteration cycle, and a random number p rand is generated. The probability is judged. When the random number p rand is less than or equal to the set bias sampling probability p 0 , the node is expanded in the direction of the random point q rand ; when the random number p rand is greater than the set bias sampling probability p 0 , if the starting tree T start is expanded at this time, the expansion direction of the node N node is in the direction of the end point for node expansion, q = (x goal , y goal ); if the target tree T goal is expanded at this time, then the expansion direction of the node N node is in the direction of the starting point for node expansion, q = (x start , y start );

[0031] The new node N new =(x new , y new)Perform collision detection according to the size of the vehicle expansion circle with a set radius of r, where x new is the abscissa of the new node, and y new is the ordinate of the new node.

[0032] Furthermore, perform collision detection according to the size of the vehicle expansion circle with a set radius of r. If the new node N new =(x new , y new ) has no collision with the obstacle set O = [(x 1l , y 1l , x 1r , y 1r )…(x il , y il , x ir , y ir )…(x ol , y ol , x or , y or ), where x new is the abscissa of the new node, y new is the ordinate of the new node; i = 1, 2…o, x il is the abscissa of the lower left corner of the i-th obstacle rectangle, y il is the ordinate of the lower left corner of the i-th obstacle rectangle, x ir is the abscissa of the upper right corner of the i-th obstacle rectangle, y ir is the ordinate of the upper right corner of the i-th obstacle rectangle, and o is the total number of obstacles; determine whether the starting tree T start and the target tree T goal are connected, and determine whether the distance dis between the latest node of the starting tree and the latest node of the target tree is less than or equal to the radius r of the vehicle expansion circle. If it is less than or equal to r, it means they can be connected; otherwise, it means they cannot be connected;

[0033] If the new node N new =(x new , y new ) has no collision with the obstacle set O = [(x 1l , y 1l , x 1r , y 1r )…(x il , y il , x ir , y ir )…(x ol , y ol , x or , y or)] If there is a collision, the target bias sampling strategy based on the obstacle vertex extension is performed to recalculate the new node N new =(x new ,y new ).

[0034] Furthermore, the target bias sampling strategy based on the obstacle vertex extension includes:

[0035] Get the new node N new =(x new ,y new ) when the coordinates (x il ,y il ,x ir ,y ir ), calculate the coordinates of the obstacle that collides with (x il ,y il ,x ir ,y ir ) from node N node The coordinates of the nearest vertex (x f ,y f ), where x f is the horizontal coordinate of the vertex, y f is the vertical coordinate of the vertex; taking this vertex as the center of the circle, randomly select a point within the rectangular range of the vehicle's minimum turning radius R-1.5R and directly use it as the new node N new =(x new ,y new ); In the starting tree T start and the target tree T goal During the expansion process, obstacle vertices are used only once.

[0036] In step S3, a node optimization is performed, including:

[0037] Set an empty node set S after the first node optimization. 1 , the starting point (x start ,y start )Add to set S 1 And as the starting point for judgment; judging the path node set S = [(x 1 ,y 1 )…(x n ,y n )] and the starting point (x start ,y start ) Whether there is an obstacle on the line, where x n is the horizontal coordinate of the nth node in the path node set, y n is the ordinate of the nth node in the path node set, and n is the total number of path nodes; if there is no obstacle, the next node is determined until the following two situations occur:

[0038] The first method: When there is an obstacle in the connection between the node (x i , y i ) and the starting point (x start , y start ), the previous node (x i , y i ) of this node (x i-1 , y i-1 ) is added to the new path node set S 1 , and the node (x i-1 , y i-1 ) is used as the new starting point to repeat the optimal selection determination process; where, x i represents the abscissa of the i-th node, and y i represents the ordinate of the i-th node; x i-1 represents the abscissa of the (i - 1)-th node, and y i-1 represents the ordinate of the (i - 1)-th node;

[0039] The second method: When the node is a new node N new = (x new , y new ) extended from the vertex of the obstacle, directly add N new = (x new , y new ) to the new path node set S 1 , and use N new = (x new , y new ) as the new starting point to repeat the optimal selection determination process; where, x new is the abscissa of the new node, and y new is the ordinate of the new node;

[0040] Traverse all the original nodes to obtain the path node set S 1 = [(x start , y start )…(x i-1 , y i-1 )…(x m , y m )] after the first node optimization, where, x m is the abscissa of the m-th node in the path node set after the first node optimization, y m is the ordinate of the m-th node in the path node set after the first node optimization, and m is the total number of path nodes after the first node optimization.

[0041] Furthermore, the second node optimization includes:

[0042] For the set S after the first node optimization 1= [(x start , y start )…(x i-1 , y i-1 )…(x m , y m )] is optimized. Set an empty set S 2 of path nodes after optimizing the secondary nodes. Add the starting point (x start , y start ) to the set S 2 and use it as the judgment starting point; Judge whether there are obstacles on the connection line between the nodes in the path node set S 1 and the starting point (x start , y start ). If there are no obstacles, judge the next node until the connection line between the node (x j , y j ) and the starting point (x start , y start ) has obstacles. Add the previous node (x j , y j ) of the node (x j-1 , y j-1 ) to the new path node set S 2 , and use the node (x j-1 , y j-1 ) as the new starting point to repeat the optimal selection judgment process; Among them, x j represents the abscissa of the jth node, and y j represents the ordinate of the jth node; x j-1 represents the abscissa of the (j - 1)th node, and y j-1 represents the ordinate of the (j - 1)th node;

[0043] Traverse all the original nodes to obtain the set S 2 of path nodes after optimizing the primary nodes = [(x start , y start )…(x j-1 , y j-1 )…(x k , y k )], where k is the total number of path nodes after optimizing the secondary nodes.

[0044] In step S3, the arc optimizes the smooth path, including:

[0045] For each corner in the path, use the minimum turning radius R of the vehicle to generate an arc tangent to the line segment to obtain the tangent points p 1 , p 2 ; In the path node set S 2 = [(x start , y start )…(xj-1 , y j-1 )…(x k , y j )] in the path inflection point (x j , y j ) is deleted, and the tangent point p 1 to p 2 of the arc is used instead. After replacing all the inflection points in the path with arcs, a smooth path is obtained.

[0046] Another object of the present invention is to provide a vehicle path planning system in the narrow and long deck space of a roll-on / roll-off ship, which is used to control the vehicle path planning method in the narrow and long deck space of the roll-on / roll-off ship. The system includes:

[0047] A parameter setting module, after setting algorithm parameters, inputs vehicle information and starting and target positions;

[0048] A path node search module, based on the set parameters, conducts path node search, and alternately expands the start tree and the target tree using the improved Bi-RRT algorithm; combined with the Bi-RRT algorithm search, a heuristic sampling strategy based on the local expansion method of expanding the obstacle vertices is proposed, which solves the problems of blind node expansion and poor obstacle avoidance ability; a variable step size strategy is adopted to solve the problem of slow expansion speed of nodes in different regions;

[0049] A node optimization and path smoothing processing module is used to perform primary node optimization, secondary node optimization, and arc optimization to smooth the path and generate a vehicle path.

[0050] Combining all the above technical solutions, the advantages and positive effects of the present invention are as follows:

[0051] First, for the vehicle path planning problem in the narrow and long deck of a roll-on / roll-off ship, specifically by improving the Bi-RRT algorithm, aiming at the problems of low search efficiency, poor obstacle avoidance ability and strong randomness in the current sampling method, a target-biased sampling strategy based on expanding the obstacle vertices is proposed, which can enhance the obstacle avoidance ability during the search process. Based on this node expansion strategy, the present invention proposes a method of secondary node optimization. By screening redundant nodes twice, the quality of nodes is effectively guaranteed, and the rapid generation of vehicle paths can be ensured. At the same time, aiming at the problem of low expansion ability of nodes with a fixed expansion step size in different regions, the present invention proposes a variable step size (large step size, medium step size and small step size) method to improve the expansion search ability of nodes; finally, the arc optimization method is used to smooth the path to ensure that the path conforms to the vehicle driving characteristics. Using this method can improve the algorithm search speed and obstacle avoidance ability, and enable the vehicle to quickly generate a smooth path from the starting point to the target point.

[0052] Second, for the problem of vehicle path planning within the roll-on / roll-off ship deck, the present invention conducts path planning for vehicles on the premise of considering the rapidity and accuracy of planning, and uses an improved Bi-RRT algorithm for spatial search to achieve path planning. The traditional Bi-RRT algorithm has a random sampling method in vehicle path planning, a fixed expansion step size, and a non-smooth enough path, resulting in low sampling efficiency and generating a large number of invalid nodes in complex environments. Its fixed step size is difficult to adapt to different obstacle environments, lacking node optimization and curve optimization, making the path have more turns and not meeting the actual driving requirements of vehicles. In contrast, the target bias sampling strategy based on obstacle vertex outward expansion, variable step size, secondary node optimization, and arc optimization methods proposed in this method significantly improve the planning effect. The target bias sampling is biased towards the target and the edge area of obstacles, improving the sampling efficiency and reducing the generation of invalid nodes; the variable step size is adjusted according to the environmental density, expanding rapidly in open areas and gradually refining in obstacle areas, enhancing the adaptability of the algorithm; the secondary node optimization removes redundant nodes and reduces turns; the arc optimization makes the path conform to the kinematic constraints of the vehicle and improves the driving smoothness. At the same time, this method reduces the number of search nodes, shortens the path length, and significantly reduces the number of iterations, effectively improving the path planning speed and the quality of the generated path, and can effectively improve the safety and efficiency of vehicle driving within the roll-on / roll-off ship deck.

[0053] Third, during the transformation process of the technical solution of the invention, the improved Bi-RRT algorithm can more accurately plan the optimal driving path of each vehicle in the ship, ensuring that the vehicle can reach the target position quickly and smoothly, and avoiding excessive waiting time caused by problems such as path crossing and congestion. Especially in a complex multi-layer deck environment, the high efficiency of the algorithm can make the vehicle scheduling process smoother and reduce the frequency of misoperations and adjustments. From a commercial value perspective, the improvement in vehicle scheduling efficiency brought by the algorithm improvement will directly shorten the ship's berthing and departure times, improve the cargo turnover efficiency, and thus increase the ship's operating benefits. Brief Description of the Drawings

[0054] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure;

[0055] Figure 1 is a flowchart of the vehicle path planning method in the long and narrow deck space of the roll-on / roll-off ship provided by the embodiment of the present invention;

[0056] Figure 2 is a diagram of the obstacle vertex outward expansion strategy provided by the embodiment of the present invention;

[0057] Figure 3 is a schematic diagram of arc smoothing processing provided by the embodiment of the present invention, where (a) is a schematic diagram before path smoothing processing, and (b) is a schematic diagram after path smoothing processing, (p1 , p 2 ) is the tangent point, (x j , u j ) is the j-th path node;

[0058] Figure 4 is the path planning effect diagram of the improved algorithm in the simple obstacle scenario provided by the embodiment of the present invention;

[0059] Figure 5 is the comparison diagram of 100 simulation times in the simple obstacle environment provided by the embodiment of the present invention;

[0060] Figure 6 is the comparison diagram of 100 simulation path lengths in the simple obstacle environment provided by the embodiment of the present invention;

[0061] Figure 7 is the path planning effect diagram of the improved algorithm in the narrow channel scenario provided by the embodiment of the present invention;

[0062] Figure 8 is the comparison diagram of 100 simulation times in the narrow channel scenario provided by the embodiment of the present invention;

[0063] Figure 9 is the comparison diagram of 100 simulation path lengths in the narrow channel scenario provided by the embodiment of the present invention;

[0064] Figure 10 is the path planning effect diagram of the improved algorithm in the long and narrow deck space scenario provided by the embodiment of the present invention;

[0065] Figure 11 is the comparison diagram of 100 simulation times in the long and narrow deck space environment provided by the embodiment of the present invention;

[0066] Figure 12 is the comparison diagram of 100 simulation path lengths in the long and narrow deck space environment provided by the embodiment of the present invention. Detailed implementation manners

[0067] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific implementation manners of the present invention will be given with reference to the accompanying drawings. Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific implementations disclosed below.

[0068] Technical term introduction:

[0069] (1) Vehicle information: In the present invention, the vehicle information is the vehicle expansion circle radius (the minimum circumscribed circle of the vehicle contour) and the minimum turning circle radius of the vehicle;

[0070] (2) Node information: In the present invention, the node information is node coordinates and a set of path nodes.

[0071] The innovation of the present invention lies in: The present invention improves the Bi - RRT algorithm for vehicle path planning problems. First, the start tree and the target tree are alternately expanded. During the expansion process, the traditional target - bias sampling strategy is used for nodes. When the expanded node collides with an obstacle, the target - bias sampling strategy based on the outward expansion of the obstacle vertex is used to directly generate a new node, which can effectively solve the blindness of node expansion and enhance the obstacle - avoidance ability of node expansion. At the same time, during the expansion process, a suitable step size is selected from three types of set step sizes (large step size, medium step size, and small step size) to expand the nodes, so as to enhance the node expansion ability and search speed in different regions. When the two trees are connected after reaching a certain distance, a set of path nodes is obtained. At this time, there are redundant nodes in the set of path nodes. The redundant nodes are filtered out by using quadratic node optimization, and the path is smoothed by using arc optimization to ensure that the driving route is smooth and meets the vehicle kinematic constraints.

[0072] The present invention improves the Bi - RRT algorithm for vehicle path planning problems. First, the start tree and the target tree are alternately expanded. During the expansion process, the traditional target - bias sampling strategy is used for nodes. When the expanded node collides with an obstacle, the target - bias sampling strategy based on the outward expansion of the obstacle vertex is used to directly generate a new node to solve the blindness of node expansion. At the same time, during the expansion process, a suitable step size is selected from three types of set step sizes (large step size, medium step size, and small step size) to expand the nodes, so as to enhance the node expansion ability in different regions. When the two trees are connected after reaching a certain distance, a set of path nodes is obtained. At this time, there are redundant nodes in the set of path nodes. The redundant nodes are filtered out by using quadratic node optimization, and the path is smoothed. The main focuses of this algorithm include the following:

[0073] (1) Aiming at the problems of low search efficiency and poor obstacle - avoidance ability in the path planning process, a target - bias sampling strategy based on the outward expansion of the obstacle vertex is designed. By obtaining the nearest vertex of the obstacle that conflicts with the node, and using this vertex as the center, a rectangular range is outwardly expanded with a minimum turning circle radius of R - 1.5R. A random point within the range is randomly selected as the new node. This method makes the node generation position specific to the vicinity of the obstacle vertex, effectively improving the search efficiency and obstacle - avoidance ability.

[0074] (2) Aiming at the problem of redundant points in path nodes, based on an improved bias sampling strategy, a quadratic node optimization method is proposed. This method takes the starting point as the judgment starting point, connects the nodes in the path node set with the starting point respectively, judges whether there is a conflict with the obstacle, and continues to screen until a conflicting node is found. The previous node of the conflicting node is added to the new path node set and continues to be screened until all nodes are traversed. When optimizing nodes for the first time, the new nodes extended from the obstacle vertices are retained, while in the quadratic node optimization, they are not retained and are screened like other nodes. This method can filter out redundant nodes and effectively improve the path quality.

[0075] (3) Aiming at the problem of low node expansion ability in path planning, a variable step size method is proposed. By calculating the distance relationship between the vehicle expansion center and the obstacle, a suitable step size is selected from the three types of set step sizes (large step size, medium step size, and small step size) to expand the nodes, effectively improving the node expansion ability of the algorithm in different regions.

[0076] Example 1, as Figure 1 shown, the vehicle path planning method in the narrow and long deck space of a ro-ro ship provided by the embodiment of the present invention includes the following steps:

[0077] S1, after setting the algorithm parameters, input vehicle information and the starting position and the target position;

[0078] The vehicle information V has a total of 4 parameters, which are successively represented as:

[0079] Starting point (x start , y start ), where x start is the abscissa of the starting point, and y start is the ordinate of the starting point;

[0080] End point (x goal , y goal ), where x goal is the abscissa of the end point, and y goal is the ordinate of the end point;

[0081] Vehicle expansion circle radius r;

[0082] Vehicle minimum turning circle radius R;

[0083] The above parameters are all scalars, all length units are m, and all coordinates are (x, y).

[0084] S2. Based on the set parameters, perform path node search, and use the improved Bi-RRT algorithm to alternately expand the start tree and the target tree. First, in combination with the Bi-RRT algorithm search, a heuristic sampling strategy based on local expansion method outside the obstacle vertices is proposed to solve the problems of blind node expansion and poor obstacle avoidance ability. Secondly, a variable step size strategy is adopted to solve the problem of slow expansion speed of nodes in different regions. Finally, combined with the characteristics of the nodes obtained by heuristic sampling, a secondary node optimization strategy is proposed to screen out redundant nodes, and an arc optimization method is used to achieve path length optimization and smoothness requirements.

[0085] (a) Select the step size: When the number of iterations is even, expand the start tree T start , and when it is odd, expand the target tree T goal , and select the step size H for expansion;

[0086] The step size H is divided into three categories: large, medium, and small. When the distance d between the vehicle center and the obstacle is between r - 2r, use the small step size h; when the distance d between the vehicle center and the obstacle is between 2r - 3r, use the medium step size 2h; when the distance d between the vehicle center and the obstacle is greater than 3r, use the large step size 3h. The expression is:

[0087]

[0088] In the formula, h is the set step size, 2h is twice the set step size, 3h is three times the set step size, r is the vehicle expansion circle radius, and d is the distance between the vehicle center and the obstacle;

[0089] When d < r, it indicates that the expanded node collides with the obstacle, and the node expansion is restarted. The specific process is carried out according to the node expansion in step (b).

[0090] (b) Node expansion: Expand a new node N node from the node N according to the selected step size H new , and adopt the target bias sampling strategy to expand the new node. The target bias sampling strategy is defined as:

[0091]

[0092] In the formula, q rand is a random point, p rand is a random number, q is the expansion direction point, and p 0 is the bias sampling probability;

[0093] Generate a random point q rand in each iteration cycle, generate a random number p rand , and perform probability size discrimination. When the random number p rand is less than or equal to the set bias sampling probability p 0 , use the random point qrand Expand nodes in the direction of; when the random number p rand is greater than the set bias sampling probability p 0 , if the starting tree T start performs node expansion, the node N node expands in the direction with the end point as the direction, q = (x goal , y goal ); if the target tree T goal performs node expansion at this time, then the node N node expands in the direction with the starting point as the direction, q = (x start , y start );

[0094] Then, perform collision detection on the new node N new = (x new , y new ) according to the size of the vehicle expansion circle with a set radius of r, where x new is the abscissa of the new node, and y new is the ordinate of the new node.

[0095] If the new node N new = (x new , y new ) has no collision with the obstacle set O = [(x 1l , y 1l , x 1r , y 1r )…(x il , y il , x ir , y ir )…(x ol , y ol , x or , y or )], where x new is the abscissa of the new node, and y new is the ordinate of the new node; i = 1, 2…o, x il is the abscissa of the lower left corner of the i-th obstacle rectangle, y il is the ordinate of the lower left corner of the i-th obstacle rectangle, x ir is the abscissa of the upper right corner of the i-th obstacle rectangle, y ir is the ordinate of the upper right corner of the i-th obstacle rectangle, and o is the total number of obstacles; judge whether the starting tree T start and the target tree T goal are connected, and judge the latest node of the starting tree and the latest node of the target tree Whether the distance dis is less than or equal to the radius r of the vehicle's expansion circle. If it is less than or equal to r, it means connection is possible; otherwise, it means connection is not possible.

[0096] If the new node N new =(x new , y new ) collides with the obstacle set O = [(x 1l , y 1l , x 1r , y 1r )…(x il , y il , x ir , y ir )…(x ol , y ol , x or , y or )], a target biasing sampling strategy based on obstacle vertex outward expansion is performed to recalculate the new node N new =(x new , y new ).

[0097] (c) Target biasing sampling strategy based on obstacle vertex outward expansion: Obtain the coordinates (x new , y new , x new ) of the obstacles that collide with the new node N il =(x il , y ir , x ir ) in the obstacle set when obtaining the expanded new node N il , y il , x ir , y ir ) in the obstacle set, and calculate the coordinates (x node , y f , x f ) of the vertex closest to the node N f among the coordinates (x f , y new ) of the obstacles that collide. Here, x new is the abscissa of the vertex, and y new is the ordinate of the vertex; Using this vertex as the center, randomly select a point within the rectangle range of the vehicle's minimum turning radius R - 1.5R and directly use it as the new node N start =(x goal , y Figure 2 ) shown in the figure. In the figure, r is the radius of the vehicle's expansion circle, and R is the minimum turning circle radius of the vehicle.

[0098] S3. Perform primary node optimization, secondary node optimization, and arc optimization to smooth the path and generate the vehicle path.

[0099] Expand the initial tree T through the above steps start and the target tree T goal until the path node set S = [(x 1 , y 1 )...(x n , y n )] (n is the total number of path nodes) is obtained after connection. The searched path may have invalid branch paths, path fluctuations in a local range, resulting in route redundancy and detours. Referencing the idea of the greedy algorithm, make an optimal decision on the nodes of the original path.

[0100] Primary node optimization includes:

[0101] Set an empty path node set S after primary node optimization 1 , add the starting point (x start , y start ) to the set S 1 and use it as the decision starting point; judge whether there are obstacles on the connection line between the nodes in the path node set S = [(x 1 , y 1 )...(x n , y n )] and the starting point (x start , y start ). Among them, x n is the abscissa of the nth node in the path node set, and y n is the ordinate of the nth node in the path node set, and n is the total number of path nodes; if there is no obstacle, proceed to the determination of the next node until one of the following two situations occurs:

[0102] The first: when there is an obstacle on the connection line between the node (x i , y i ) and the starting point (x start , y start ), add the previous node (x i , y i ) of this node (x i-1 , y i-1 ) to the new path node set S 1 , and use the node (x i-1 , y i-1 ) as the new starting point to repeat the optimal decision process; among them, x i represents the abscissa of the i-th node, and y i represents the ordinate of the i-th node; x i-1 represents the abscissa of the (i - 1)-th node, and y i-1Represents the ordinate of the (i - 1)-th node;

[0103] The second type: When the node is a new node N extended from the vertex of the obstacle new =(x new , y new ), directly add N new =(x new , y new ) to the new path node set S 1 , and use N new =(x new , y new ) as the new starting point to repeat the optimal selection determination process; where x new is the abscissa of the new node, and y new is the ordinate of the new node;

[0104] Traverse all the original nodes to obtain the path node set S 1 =[(x start , y start )…(x i-1 , y i-1 )…(x m , y m )] after the first - stage node optimization, where x m is the abscissa of the m - th node in the path node set after the first - stage node optimization, y m is the ordinate of the m - th node in the path node set after the first - stage node optimization, and m is the total number of path nodes after the first - stage node optimization.

[0105] Second - stage node optimization, including:

[0106] Optimize the set S 1 =[(x start , y start )…(x i-1 , y i-1 )…(x m , y m )] after the first - stage node optimization, set an empty path node set S 2 after the second - stage node optimization, add the starting point (x start , y start ) to the set S 2 and use it as the determination starting point; Determine whether there is an obstacle on the line connecting the nodes in the path node set S 1 and the starting point (x start , y start ). If there is no obstacle, proceed to the determination of the next node until a node (x j , y j ) and the starting point (x start , y start)There is an obstacle in the connection. Add the previous node (x j ,y j ) of the node (x j-1 ,y j-1 ) to the new path node set S 2 , and use the node (x j-1 ,y j-1 ) as the new starting point to repeat the optimal selection determination process; where x j represents the abscissa of the j-th node, and y j represents the ordinate of the j-th node; x j-1 represents the abscissa of the (j - 1)-th node, and y j-1 represents the ordinate of the (j - 1)-th node;

[0107] Traverse all the original nodes to obtain the path node set S 2 = [(x start ,y start )…(x j-1 ,y j-1 )…(x k ,y k )] after the first node optimization, where k is the total number of path nodes after the second node optimization.

[0108] Arc optimization for a smooth path, including: for each corner in the path, use the minimum turning radius R of the vehicle to generate an arc tangent to the line segment to obtain the tangent points p 1 ,p 2 ; Delete the path inflection point (x 2 = [(x start ,y start )…(x j-1 ,y j-1 )…(x k ,y k )] in the path node set S j ,y j ), and replace it with the arc from the tangent point p 1 to p 2 . After replacing all the inflection points in the path with arcs, a smooth path is obtained; the specific method is as Figure 3 shown.

[0109] Example 2, The vehicle path planning system in the long and narrow deck space of the ro-ro ship provided by the embodiment of the present invention specifically includes:

[0110] Parameter setting module, after setting the algorithm parameters, input the vehicle information and the starting position and the target position;

[0111] The path node search module performs path node search based on set parameters, and uses the improved Bi-RRT algorithm to alternately expand the start tree and the target tree; in combination with the Bi-RRT algorithm search, a heuristic sampling strategy based on the local expansion method of expanding the obstacle vertices is proposed, which solves the problems of blind node expansion and poor obstacle avoidance ability; a variable step size strategy is adopted to solve the problem of slow expansion speed of nodes in different regions.

[0112] The node optimization and path smoothing processing module is used to perform primary node optimization, secondary node optimization, and arc optimization to smooth the path and generate a vehicle path.

[0113] To further prove the positive effects of the above embodiments, the present invention conducts the following experiments based on the above technical solutions.

[0114] Three examples are given, and simulation analysis is carried out in three different scenarios: a simple obstacle scenario, a narrow passage scenario, and a long and narrow deck space scenario. The parameter settings are shown in Table 1.

[0115] Table 1 Parameter Setting Table

[0116] Parameter Name Value Vehicle Expansion Circle Radius r (m) 3 Vehicle Minimum Turning Circle Radius R (m) 4 Set Expansion Step h (m) 1 <![CDATA[Bias sampling probability p 0 (%)]]> 60 <![CDATA[Starting point coordinates (x start , y start )]]> (175，7) <![CDATA[Target point coordinates (x goal , y goal )]]> (7，42)

[0117] The obstacle settings in the three scenarios are as follows:

[0118] Simple obstacle scenario: [(75,25,40,25),(16,35,40,15),(40,0,20,20), (120,0,45,15)]

[0119] Narrow passage scenario: [(20,0,5,10),(20,19,5,31),(155,0,5,35),(155,44,5,6)]

[0120] Long and narrow space scenario: [(15,35,15,15),(20,5,5,5),(60,25,5,25),(40,0,10,6),(70,40,15,5),

[0121] (60,0,15,15),(80,2,10,9),(100,30,5,20),(95,5,5,10),(120,10,5,40),

[0122] (160,40,5,5),(160,0,5,15),(140,40,5,5),(135,0,5,15),(180,40,5,5),

[0123] (180,5,5,5),(25,17,20,10),(70,25,15,0),(150,25,5,15)]

[0124] The simulation was run 100 times respectively using the RRT* algorithm, Bi-RRT* algorithm, Informed-RRT* algorithm and the improved algorithm of this patent under three scenarios, and the simulation results were statistically analyzed. Among them, the results of each index are all average values.

[0125] (1) Simple obstacle scenario;

[0126] Table 2 is a statistical table of simulation data (average values) under the simple obstacle scenario, Figure 4 is the path planning effect diagram of the improved algorithm. The black dotted line is the node and path before optimization, and the blue is the final path after optimization.

[0127] Table 2 Comparison table of simulation results under the simple obstacle scenario

[0128] Algorithm Number of Sampling Points Number of Path Nodes Number of Iterations Path Length (m) RRT* Algorithm 447 207 1182 205 Bi-RRT* Algorithm 1617 225 5209 401 Informed-RRT* Algorithm 412 32 1034 185 Improved Algorithm 61 6 50 188

[0129] It can be seen from Table 2 that the improved algorithm reduces by 86.35%, 96.22%, and 85.19% respectively compared with the other three algorithms in terms of the average number of sampling points, with a significant improvement effect and effectively improving the search efficiency; in terms of the average number of path nodes, it reduces by 97.1%, 97.33%, and 81.25% respectively, effectively removing redundant nodes and facilitating the rapid smoothing of the path; in terms of the average number of iterations, it reduces by 95.77%, 99.04%, and 95.16% respectively, effectively improving the path planning efficiency; in terms of the average path length, it reduces by 8.29% and 53.12% respectively compared with the RRT* algorithm and the Informed-RRT* algorithm, with an obvious improvement, but it increases by 1.6% compared with the Informed-RRT* algorithm because the path generated by the Informed-RRT* algorithm is not smooth and does not meet the vehicle driving requirements. After smoothing, the path will increase to a certain extent, while the path generated by the improved algorithm is after smoothing and meets the vehicle kinematics. It can be concluded that the improved algorithm has a good improvement effect under the simple obstacle scenario.

[0130] As Figure 5 shown, in 100 simulations in the simple obstacle environment, the time range of the improved algorithm is within 10-1 to 10-2 s, the time range of the RRT* algorithm and the Informed-RRT* algorithm is roughly within 10-2 to 10 s, and the time range of the Bi-RRT* algorithm is within 10 to 102 s. The improved algorithm is much smaller than the other three algorithms. The standard deviations of the running times of the improved algorithm, RRT* algorithm, Bi-RRT* algorithm and Informed-RRT* algorithm are: 0.03, 0.15, 7.05, 0.21 respectively. It can be seen that the standard deviation of the Bi-RRT* algorithm is larger than the other three algorithms, and the standard deviation of the running time of the improved algorithm is the smallest, being more stable than other algorithms.

[0131] As Figure 6 shown, in 100 simulations in a simple obstacle environment, the path length range of the improved algorithm is roughly between 175m and 195m, which is much smaller than the path length range of 360m - 460m of the Bi - RRT* algorithm. At the same time, the standard deviations of the path lengths of the improved algorithm, RRT* algorithm, Bi - RRT* algorithm, and Informed - RRT* algorithm are 1.85, 6.98, 16.56, and 4.12 respectively. It can be seen that the standard deviation of the path length of the improved algorithm is smaller and more stable than other algorithms.

[0132] (2) Narrow channel scenario

[0133] Table 3 is a statistical table of simulation data (average values) in the narrow channel scenario, Figure 7 which is the path planning effect diagram of the improved algorithm.

[0134] Table 3 Comparison table of simulation results in the narrow channel scenario

[0135] Algorithm Number of Sampling Points Number of Path Nodes Number of Iterations Path Length (m) RRT* Algorithm 1648 263 5249 261 Bi-RRT* Algorithm 3283 274 11135 451 Informed-RRT* Algorithm 2089 62 7444 227 Improved Algorithm 82 6 56 214

[0136] It can be seen from Table 3 that the improved algorithm reduces by 95.02%, 99.93%, and 96.07% respectively compared with the other three algorithms in terms of the average number of sampling points; reduces by 97.72%, 97.81%, and 90.32% respectively in terms of the average number of path nodes; reduces by 98.93%, 99.5%, and 99.25% respectively in terms of the average number of iterations; and reduces by 18%, 52.55%, and 5.73% respectively in terms of the average path length. The performance of each parameter of the improved algorithm has been improved. Thus, it can be concluded that the improved algorithm has a good improvement effect in the narrow channel scenario.

[0137] As Figure 8 shown, in 100 simulations in the narrow channel scenario, the time range of the improved algorithm is roughly within 10^(-2) - 10^(-1) s, the time ranges of the RRT* algorithm and Informed - RRT* algorithm are roughly within 10^(-1) - 10 s, and the time range of the Bi - RRT* algorithm is within 10 - 10^3 s. It can be seen that the search time of the improved algorithm is significantly reduced. The standard deviations of the running times of the improved algorithm, RRT* algorithm, Bi - RRT* algorithm, and Informed - RRT* algorithm are 0.03, 2.83, 207.49, and 5.98 respectively. It can be seen that the standard deviation of the running time of the improved algorithm is significantly smaller than other algorithms, and the algorithm is more stable.

[0138] As Figure 9As shown, in 100 simulations in the narrow passage scenario, the path length range of the improved algorithm is 210m - 220m, which is much smaller than the other three algorithms. The standard deviations of the path lengths of the improved algorithm, RRT* algorithm, Bi-RRT* algorithm, and Informed-RRT* algorithm are 0.75, 8.53, 11.49, and 3.71 respectively. It can be seen that the standard deviation of the path length of the improved algorithm is significantly smaller than that of other algorithms, and the algorithm is more stable.

[0139] (3) Narrow and long deck space scenario

[0140] Table 4 is a statistical table of simulation data (average values) in the narrow and long deck space scenario. Figure 10 It is the path planning effect diagram of the improved algorithm.

[0141] Table 4 Comparison table of simulation results in the narrow and long deck space scenario

[0142] Algorithm Number of Sampling Points Number of Path Nodes Number of Iterations Path Length (m) RRT* Algorithm 1552 240 7063 239 Bi-RRT* Algorithm 7054 247 17473 423 Informed-RRT* Algorithm 1831 65 8253 206 Improved Algorithm 105 10 85 210

[0143] It can be seen from Table 4 that the improved algorithm reduces the average sampling points by 93.23%, 98.51%, and 94.27% respectively compared with the other three algorithms, greatly reducing the invalid search nodes and improving the search efficiency; in terms of the average number of path nodes, it reduces by 95.83%, 95.95%, and 84.62% respectively, effectively removing redundant nodes and facilitating the rapid smoothing of the path; in terms of the average number of iterations, it reduces by 98.8%, 99.51%, and 98.97% respectively; in terms of the average path length, it reduces by 12.13% and 50.35% respectively compared with the RRT* algorithm and the Informed-RRT* algorithm, showing obvious improvement. However, compared with the Informed-RRT* algorithm, it increases by 1.94% because the path generated by the Informed-RRT* algorithm is not smooth and does not meet the vehicle driving requirements. After smoothing, the path will increase to a certain extent, while the path generated by the improved algorithm is after smoothing and meets the vehicle kinematics. It can be concluded that the improved algorithm has a good improvement effect in the narrow and long deck space scenario.

[0144] As Figure 11 shown, in 100 simulations in the narrow and long deck space environment, the time of the improved algorithm is about 10 - 1s, which is much smaller than the other three algorithms. The standard deviations of the running times of the improved algorithm, RRT* algorithm, Bi-RRT* algorithm, and Informed-RRT* algorithm are 0.11, 2, 324.11, and 13.66 respectively. It can be seen that the standard deviation of the running time of the improved algorithm is significantly smaller than that of other algorithms, and the algorithm is more stable.

[0145] As Figure 12As shown, in 100 simulations in the narrow deck space environment, the path length range of the improved algorithm is approximately 200m to 220m, which is much smaller than that of the Bi-RRT* algorithm. The standard deviations of the path lengths of the improved algorithm, RRT* algorithm, Bi-RRT* algorithm, and Informed-RRT* algorithm are 1.96, 9.46, 21.33, and 5.53 respectively. It can be seen that the standard deviation of the path length of the improved algorithm is significantly smaller than that of other algorithms, and the algorithm is more stable.

[0146] As described above, only the preferred specific embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A vehicle path planning method in a narrow and long deck space of a ro-ro ship, characterized in that: The method includes the steps of: S1. After setting the algorithm parameters, input the vehicle information, as well as the starting position and the target position; S2. Based on the set parameters, conduct path node search, and use the improved Bi-RRT algorithm to alternately expand the starting tree and the target tree; Combined with the Bi-RRT algorithm search, a heuristic sampling strategy based on the local expansion method of expanding the obstacle vertices outward is proposed, which solves the problems of blind node expansion and poor obstacle avoidance ability; a variable step size strategy is adopted to solve the problem of slow expansion speed of nodes in different regions; S3. Conduct primary node optimization, secondary node optimization, and arc optimization to smooth the path and generate the vehicle path.

2. The vehicle path planning method in the narrow and long deck space of a ro-ro ship according to claim 1 is characterized in that: In step S1, the vehicle information V has a total of 4 parameters, which are successively represented as: Starting point (x start ,y start ), where x start is the horizontal coordinate of the starting point, y start is the vertical coordinate of the starting point; End point (x goal ,y goal ), where x goal is the abscissa of the end point, y goal is the ordinate of the end point; The vehicle inflation circle radius r; The vehicle minimum turning circle radius R; The above parameters are all scalars, all length units are m, and all coordinates are (x, y).

3. The vehicle path planning method in the narrow and long deck space of a ro-ro ship according to claim 2 is characterized in that: In step S2, a path node search is performed, including: when the number of iterations is an even number, the starting tree T is expanded start , when odd, expand the target tree T goal , select step size H for expansion; The step size H is divided into three categories: large, medium, and small. When the distance d between the vehicle center and the obstacle is between r - 2r, the small step size h is used; when the distance d between the vehicle center and the obstacle is between 2r - 3r, the medium step size 2h is used; when the distance d between the vehicle center and the obstacle is greater than 3r, the large step size 3h is used. The expression is: In the formula, h is the set step size, 2h is twice the set step size, 3h is three times the set step size, r is the vehicle inflation circle radius, and d is the distance between the vehicle center and the obstacle; When d < r, it indicates that the expanded node collides with the obstacle, and the node expansion is restarted.

4. The vehicle path planning method in the narrow and long deck space of a ro-ro ship according to claim 3 is characterized in that: According to the selected step size H, from node N node Expand new node N new , the target bias sampling strategy is used to expand new nodes. The target bias sampling strategy is defined as: In the formula, q rand is a random point, p rand is a random number, q is the expansion direction point, and p0 is the bias sampling probability; Generate a random point q in each iteration cycle rand , generate a random number p rand , to judge the probability, when the random number p rand Less than or equal to the set bias sampling probability p0, with random point q rand Node expansion is performed in the direction; when the random number p rand If the starting tree T is greater than the set bias sampling probability p0, start Expand the nodes, node N node The expansion direction is to expand the node in the direction of the end point, q = (x goal ,y goal ); If the target tree T goal For node expansion, node N node The expansion direction is to expand the node from the starting point, q = (x start ,y start ); The new node N new =(X new ,y new ) performs collision detection according to the vehicle expansion circle size with a set radius of r, where x new is the horizontal coordinate of the new node, y new is the ordinate of the new node.

5. The vehicle path planning method in the narrow and long deck space of a ro-ro ship according to claim 4 is characterized in that: According to the vehicle expansion circle size with a set radius of r, collision detection is performed. If the new node N new =(x new ,y new ) and the obstacle set O = [(x 1l ,y 1l ,x 1r ,y 1r )…(x il ,y il ,x ir ,y ir )…(x ol ,y ol ,x or ,y or )] No collision, where x new is the horizontal coordinate of the new node, y new is the ordinate of the new node; i = 1, 2…o, x il is the horizontal coordinate of the lower left corner of the i-th obstacle rectangle, y il is the ordinate of the lower left corner of the i-th obstacle rectangle, x ir is the horizontal coordinate of the upper right corner of the i-th obstacle rectangle, y ir is the ordinate of the upper right corner of the i-th obstacle rectangle, and o is the total number of obstacles; determine the starting tree T start and the target tree T goal Whether to connect, determine the latest node of the starting tree The latest node of the target tree Is the distance dis less than or equal to the radius r of the vehicle expansion circle? If so, it means the connection is possible; otherwise, it means the connection is not possible. If the new node N new =(x new ,y new ) and the obstacle set O = [(x 1l ,y 1l ,x 1r ,y 1r )…(x il ,y il ,x ir ,y ir )…(x ol ,y ol ,x or ,y or )] If there is a collision, the target bias sampling strategy based on the obstacle vertex extension is performed to recalculate the new node N new =(x new ,y new ).

6. The vehicle path planning method in the narrow and long deck space of a ro-ro ship according to claim 5 is characterized in that: The target bias sampling strategy based on expanding the obstacle vertices outward includes: Get the new node N new =(x new ,y new ) when the coordinates (x il ,y il ,x ir ,y ir ), calculate the coordinates of the obstacle that collides with (x il ,y il ,x ir ,y ir ) from node N node The coordinates of the nearest vertex (x f ,y f ), where x f is the horizontal coordinate of the vertex, y f is the vertical coordinate of the vertex; taking this vertex as the center of the circle, randomly select a point within the rectangular range of the vehicle's minimum turning radius R-1.5R and directly use it as the new node N new =(x new ,y new ); In the starting tree T start and the target tree T goal During the expansion process, obstacle vertices are used only once.

7. The vehicle path planning method in the narrow and long deck space of a ro-ro ship according to claim 2 is characterized in that: In step S3, the primary node optimization includes: Set an empty node set S1 after the first node optimization, and change the starting point (x start ,y start ) is added to the set S1 and used as the starting point for determination; the set of path nodes S = [(x1, y1)…(x n ,y n )] and the starting point (x start ,y start ) Whether there is an obstacle on the line, where x n is the horizontal coordinate of the nth node in the path node set, y n is the ordinate of the nth node in the path node set, and n is the total number of path nodes; if there is no obstacle, the next node is determined until the following two situations occur: The first one: When a node (x i ,y i ) and the starting point (x start ,y start ) connection is blocked, this node (x i ,y i )'s previous node (x i-1 ,y i-1 ) is added to the new path node set S1, and the node (x i-1 ,y i-1 ) will be used as a new starting point to repeat the optimal decision process; where x i Indicates the horizontal coordinate of the i-th node, y i represents the ordinate of the i-th node; x i-1 Indicates the horizontal coordinate of the i-1th node, y i-1 Represents the ordinate of the i-1th node; The second type: when the node is a new node N that is expanded outside the obstacle vertex new =(x new ,y new ), directly change N new =(x new ,y new ) is added to the new path node set S1, and N new =(x new ,y new ) as a new starting point to repeat the optimal decision process; where x new is the horizontal coordinate of the new node, y new is the ordinate of the new node; Traverse all the original nodes and get the path node set S1 after node optimization = [(x start ,y start )…(x i-1 ,y i-1 )…(x m ,y m )], where x m is the horizontal coordinate of the mth node in the path node set after a node optimization, y m It is the ordinate of the mth node in the path node set after a node optimization, and m is the total number of path nodes after a node optimization.

8. The vehicle path planning method in the narrow and long deck space of a ro-ro ship according to claim 7 is characterized in that: The secondary node optimization includes: After a node optimization, the set S1 = [(x start ,y start )…(x i-1 ,y i-1 )…(x m ,y m )] to optimize, set the empty secondary node after the optimized path node set S2, and set the starting point (x start ,y start ) is added to the set S2 and used as the starting point for determination; the nodes in the path node set S1 are determined to be consistent with the starting point (x start ,y start ) to see if there is an obstacle on the line. If there is no obstacle, the next node is determined until a node (x j ,y j ) and the starting point (x start ,y start ) connection is blocked, and the node (x j ,y j )'s previous node (x j-1 ,y j-1 ) is added to the new path node set S2, and the node (x j-1 ,y j-1 ) as a new starting point to repeat the optimal decision process; where x j Indicates the horizontal coordinate of the jth node, y j represents the ordinate of the jth node; x j-1 Indicates the horizontal coordinate of the j-1th node, y j-1 Represents the ordinate of the j-1th node; Traverse all the original nodes and obtain the path node set S2 after node optimization = [(x start ,y start )…(x j-1 ,y j-1 )…(x k ,y k )], where k is the total number of path nodes after secondary node optimization.

9. The vehicle path planning method in the narrow and long deck space of a ro-ro ship according to claim 1, characterized in that: In step S3, the arc optimization to smooth the path includes: For each corner in the path, use the vehicle's minimum turning radius R to generate an arc tangent to the line segment to obtain the tangent points p1 and p2; in the path node set S2 = [(x start ,y start )…(x j-1 ,y j-1 )…(x k ,y k )] to delete the path inflection point (x j ,y j ), and use the arc from the tangent point p1 to p2 instead, and replace all the inflection points in the path with arcs to obtain a smooth path.

10. A vehicle path planning system in the narrow and long deck space of a ro-ro ship, characterized in that: This system is used to control the vehicle path planning method in the narrow deck space of the ro-ro ship described in any one of claims 1 - 9. This system includes: A parameter setting module, which, after setting the algorithm parameters, inputs the vehicle information, as well as the starting position and the target position; A path node search module, which, based on the set parameters, conducts path node search, and uses the improved Bi-RRT algorithm to alternately expand the starting tree and the target tree; combined with the Bi-RRT algorithm search, a heuristic sampling strategy based on the local expansion method of expanding the obstacle vertices outward is proposed, which solves the problems of blind node expansion and poor obstacle avoidance ability; a variable step size strategy is adopted to solve the problem of slow expansion speed of nodes in different regions; A node optimization and path smoothing processing module, which is used to conduct primary node optimization, secondary node optimization, and arc optimization to smooth the path and generate the vehicle path.

Images