CN114326725B - Man-machine interaction-oriented intelligent ship collision prevention method and system - Google Patents

Abstract

An intelligent ship collision avoidance method and system for human-computer interaction is disclosed. The method includes: obtaining the basic scale information of the target ship, and realizing the exchange of future trajectory points or way points between the own ship and the target ship; according to the target obtained by the own ship The future trajectory of the ship eliminates the course and speed that can cause the ship to collide, and provides the optimal collision avoidance decision; the decision-making process is visualized, and the crew is allowed to set priority decision-making areas and arbitrarily modify the collision avoidance decision in the feasible decision-making area; When there are few feasible solutions for collision avoidance decisions, a control warning is issued to remind the crew to take over the ship in time; the ship's safe course and speed are implemented, and the crew is allowed to directly take over control of the ship's propeller speed and rudder angle at any time. The invention can present the collision avoidance decision-making process to the controller in an ergonomic manner, help the controller understand the operation intention of the intelligent system, and support the controller to directly or indirectly intervene in the avoidance action of the intelligent ship.

Description

Intelligent ship collision avoidance method and system for human-computer interaction

Technical field

The invention relates to a method and system for intelligent ship collision avoidance oriented to human-computer interaction.

Background technique

A major symbol of ship intelligence is the realization of autonomous ship navigation, and realizing autonomous ship navigation requires the key technology of autonomous collision avoidance.

From the perspective of the development of ship collision avoidance technology, the current ship autonomous collision avoidance technology is not enough to support fully autonomous navigation. It can be seen that ship collision avoidance will transition from the current human-driven mode to the human-machine integration mode (i.e. semi-autonomous navigation). mode), and finally achieve a fully machine-controlled mode (i.e., fully autonomous mode). Judging from the current development of artificial intelligence technology, human-machine integration will become the norm in future social development.

However, the current ship autonomous collision avoidance technology does not take into account the characteristics of humans and machines. This is reflected in the following: the ship's autonomous collision avoidance algorithm lacks support for human-machine interaction, the collision risk model is not yet suitable for ship control switching tasks, and there is a lack of ship control. Design of human-computer interaction in power switching. Therefore, the existing technology is difficult to meet the task of autonomous ship collision avoidance in the human-machine integration mode.

Simply improving the intelligence level of the machine without supporting human-computer interaction will not only fail to make full use of the crew's driving experience and understanding of collision avoidance rules, but also make it more difficult to gain the trust of the crew; on the other hand, simply enhancing the crew's perception ability will For example, various collision warning systems cannot provide collision avoidance decision-making assistance, and cannot improve the ship's autonomy and adapt to the human-machine integration model.

To sum up, the existing autonomous collision avoidance technology and collision avoidance auxiliary technology are not enough to support the autonomous collision avoidance of ships in the human-machine fusion mode, and the human-machine fusion mode is an inevitable stage in the process of ship intelligence. Therefore, it is necessary to propose an intelligent ship collision avoidance technology oriented to human-computer interaction.

Contents of the invention

Ship intelligence has been a hot topic and focus of shipping development in recent years. Judging from the current trend of intelligent development, using machines to completely replace the role of crew members to achieve autonomous collision avoidance still faces many difficulties. First, the machine’s understanding of rules and good boatmanship is insufficient; second, the machine’s ability to handle unexpected situations is poor; and finally, the machine is still unable to gain the full trust of humans. Therefore, various types of smart ships still require driver supervision and control. In addition, machines and humans have different strengths and weaknesses and are highly complementary. Reflected in ship collision avoidance decision-making, humans are good at in-depth understanding of rules and good boatmanship and handling various maritime emergencies on the spot, while machines have the ability to perceive beyond visual range, fast computing power, and operate in various working environments. The ability to operate stably. Based on the above considerations, the present invention proposes an intelligent ship collision avoidance system and method for human-computer interaction, which can present the collision avoidance decision-making process of the intelligent system to the controller in an ergonomic manner and help the controller understand the intelligence The operating intention of the system supports the controller to directly or indirectly intervene in the avoidance action of the smart ship.

This invention proposes autonomous collision avoidance technology based on shared decision-making. On the basis of realizing autonomous collision avoidance, by marking the safety control set and the dangerous decision set in the decision-making space, it helps the crew understand the collision avoidance decision-making of the autonomous collision avoidance system and supports the crew. Find new collision avoidance operations in the safety control center, promote the crew and the system to reach a decision-making consensus, and ultimately achieve communication, collaboration and integration between the autonomous collision avoidance system and the crew.

This invention proposes a control right switching method based on the encounter situation. By evaluating the reliability and remaining operating margin (i.e., safety control set) of the decision-making space of the ship's autonomous collision avoidance system during the encounter, it helps the autonomous system proactively evaluate system failure. possibilities and identify the scenarios and timing of control transfer, ultimately achieving safe and timely switching of control.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present invention more clearly, the drawings of the embodiments will be briefly introduced below.

Figure 1 is a flow chart of a ship intelligent collision avoidance method for human-computer interaction provided by an embodiment of the present invention.

Figures 2a and 2b are schematic diagrams of the dangerous location area when two ships meet according to an embodiment of the present invention.

Figures 3a and 3b are schematic diagrams of obtaining the minimum safety zone and dangerous state set at each moment from the trajectory of other ships according to an embodiment of the present invention.

Figures 4a and 4b show the dangerous decision domain (Figure 4b) in the decision space obtained from the dangerous state set (Figure 4a) according to an embodiment of the present invention.

Figures 5a and 5b are schematic diagrams of the convex safety decision domain (Figure 5a) and the optimal collision avoidance solution (Figure 5b) provided by an embodiment of the present invention.

Detailed ways

The invention provides a ship intelligent collision avoidance system oriented to human-computer interaction, which includes a ship automatic identification and trajectory exchange module, a ship autonomous decision-making module, a ship decision-making space visualization module, a ship control right takeover early warning module and the execution of ship collision avoidance decisions. module.

The automatic ship identification and trajectory exchange module is used to obtain the basic scale information of the target ship and realize the exchange of future trajectory points or way points between the own ship and the target ship, thereby helping the own ship fully understand the status and future movement of the target ship. The basic scale information of the target ship includes captain, course, position and speed information. The ship's autonomous decision-making module is used to eliminate the course and speed that may cause the ship to collide based on the future trajectory of the target ship obtained by the own ship, and provide the optimal collision avoidance decision. The ship decision space visualization module is used to visualize the ship decision-making process, and allows the crew to set priority decision-making areas and arbitrarily modify the ship's collision avoidance decision in the feasible decision-making area. The ship control right takeover early warning module is used to issue a control early warning when there are few feasible solutions for collision avoidance decisions to remind the crew to take over the ship in time. The execution module of the ship's collision avoidance decision is used to execute the ship's safe course and speed, and allows the crew to directly take over the control of the ship's propeller speed and rudder angle at any time.

The ship autonomous decision-making module includes a first marking module, an identification module, a second marking module and an optimal collision avoidance decision-making module. The first marking module is used to mark the geographical coordinates, geometric boundaries and time of the area where the own ship and the target ship will arrive in the future, that is, the dangerous location area, according to the scale information and future trajectory information of the target ship. The identification module is used to identify the speed that will cause the own ship and the target ship to occupy the same position at the same time in the future, and calls it a speed obstacle. The second marking module is used to mark the speed obstacles corresponding to the future position of each target ship, which is called a speed obstacle set. The optimal collision avoidance decision-making module is used to remove the outer envelope of the speed obstacle set, and take a convex set of the envelope to obtain a convex speed obstacle area, and then use a half plane to represent the convex speed obstacle area, Obtain a feasible half-plane, and find the point closest to the current heading and speed within the feasible half-plane to obtain the optimal collision avoidance decision.

The ship decision space visualization module includes a mapping module, an information acquisition module and a processing module. The mapping module is used to map the obtained set of speed obstacles in the visualization space based on the speed of the own ship. The information acquisition module is used to obtain the priority solution area input by the crew in the speed visualization space. The processing module is used to find the decision point closest to the optimal collision avoidance decision in the priority solution area.

In the ship control takeover early warning module, when the crew does not enter the priority solution area, the default priority solution area is [-90,90], [0, v _max ], where v _max is the ship's Maximum speed, the ship control takeover early warning module will evaluate the ratio of feasible areas and infeasible areas in the priority solution area in real time. When the ratio of feasible areas is less than the set value, an early warning will be issued to remind the driver to take over the ship.

Figure 1 shows an intelligent ship collision avoidance method oriented to human-computer interaction. The method includes:

Step 1: Obtain the basic scale information of the target ship, and realize the exchange of future trajectory points or way points between the own ship and the target ship, thereby helping the own ship fully understand the status and future movement of the target ship. The basic scale information of the target ship includes the length of the target ship. , heading, position and speed information;

Step 2: Based on the future trajectory of the target ship obtained by the own ship, eliminate the course and speed that may cause the ship to collide, and provide the optimal collision avoidance decision;

Step 3: Visualize the decision-making process and allow crew members to set priority decision-making areas and arbitrarily modify collision avoidance decisions in feasible decision-making areas;

Step 4: When there are few feasible solutions for collision avoidance decision-making, a control warning is issued to remind the crew to take over the ship in time;

Step 5: Carry out the ship's safe course and speed, and allow the crew to directly take over control of the ship's propeller speed and rudder angle at any time.

The specific method for obtaining the optimal collision avoidance decision is:

Step 20: According to the scale information and future trajectory information of the target ship, mark the geographical coordinates, geometric boundaries and time of the area where the future ship will arrive, that is, the dangerous location area;

Step 21: Identify the speed that causes the own ship and the target ship to occupy the same position at the same time, and call it a speed obstacle;

Step 22: Mark the speed obstacles corresponding to the future positions of each other ship, which is called the speed obstacle set;

Step 23: Remove the outer envelope of the speed obstacle set and take a convex set of the envelope to obtain the convex speed obstacle area. Then use a half plane to represent the convex speed obstacle area to obtain a feasible half plane, and Find the point closest to the current heading and speed in the feasible half-plane to obtain the optimal collision avoidance decision.

The process of visualizing the decision-making process includes: step 30, mapping the obtained speed obstacle set in the own ship's speed visualization space; step 31, obtaining the priority solution area input by the crew in the speed visualization space; step 32, in Find the decision point closest to the optimal collision avoidance decision in the priority solution area.

In step 5, when the crew does not enter the priority solution area, the default priority solution area is [-90, 90], [0, v _max ], where v _max is the maximum speed of the ship, and real-time evaluation is in the priority solution area When the ratio of the feasible area is less than the set value, an early warning is issued to remind the driver to take over the ship.

(1) Define state space and decision space

First, the meaning of state space and decision space is given. The state space refers to the space composed of the ship's state vector x, denoted as C. The state space can be divided into a dangerous state set C _coll and a safe state set C _free , that is, C=C _coll ∪C _free . When the ship's state x belongs to the dangerous state set (x∈C _coll ), an accident occurs. The decision space is a space composed of the ship's control vector u, denoted as U, and can be divided into two types of sets: the dangerous decision domain U _coll and the safety control set U _free . When the ship's control input u belongs to the dangerous decision set (i.e. u∈U _coll ), then So that x(t)∈C _coll , that is, the accident will occur at some time in the future.

Secondly, the decision variables are selected according to crew habits and the decision space is defined. Define several groups of decision variable candidate solutions based on literature review, such as heading and speed, speed and steering speed, rudder angle and gear, etc. Select a set of decision variables through crew surveys or questionnaires, and use these decision variables to define the decision space.

(2) Construct a dangerous state set and derive the dangerous decision domain of the ship

First, obtain the movement trajectory of other ships (target ships). Use communication information to obtain the future trajectory of other ships.

Secondly, according to the predicted/acquired trajectory, the dangerous state set C _coll is marked in the state space. Considering the geometry of the ship, the minimum safe area of the ship is defined, as shown in Figure 2b. Considering the predicted trajectory and error of other ships, the state space occupied by the dangerous position area of the ship at each moment in the future is obtained, as shown in Figure 3b. By connecting the minimum safe area at each moment with an envelope, the dangerous state set can be obtained, as shown in Figure 4a.

Finally, find the projection of the dangerous state set in the control space. According to the system dynamic equation, numerical methods and analytical methods are combined to solve the input-to-input mapping relationships h(·):u→x and h ^-1 (·):x→u. Then, using the reachable set analysis method and the generalized speed barrier method, the reachable set U _coll of the state set C _coll in the decision space is derived, which is the dangerous decision domain, as shown in Figure 4b.

(3) Find the optimal collision avoidance solution with the hazard domain as the constraint

Since the security decision-making domain is non-convex, it needs to be convex. There are many principles for convexity in the security domain. How to choose the convexity strategy is the key point. From the dangerous decision domain U _coll , the expression of the safe domain U _safe can be obtained as:

U _safe ＝R ^m \U _Coll ,

Among them, m is the dimension of the decision variable u; "\" is the complement operation, for example: "A\B" means the set of A that is not B. The convexization of the safety decision domain is to use a half-plane to convex the non-convex safety domain.

The convexization result of a non-convex set is not unique. Figure 5a shows an approximate scheme of a safety control set, and its algebraic expression is:

Among them, H _i is the convex half-plane, which is a subset of U _safe ; n _i is the normal vector of the plane boundary; b _i is a constant related to the vertical axis intercept of the boundary.

The core of convexization is to select an appropriate half-plane, and the key to selecting a half-plane is to select the normal vector n _i of the boundary. The criteria for selecting normal vectors can be summarized into three categories:

1) On the boundary of the U _coll domain, select the point closest to the current input, make a tangent line to the U _coll domain, and determine the half-plane with the normal vector of the tangent line;

2) On the boundary of the U _coll domain, select the point closest to the ideal input, make a tangent line to the U _coll domain, and determine the half-plane with the normal vector of the tangent line;

3) Select the convexization scheme that maximizes the feasible region, that is, on the boundary of the U _coll domain, select the point closest to the center of the decision domain, make a tangent line to the U _coll domain, and determine the half-plane with the normal vector of the tangent line.

By comparing multiple safety control set convexization schemes, the optimal convexity scheme is selected.

The objective function is introduced to find the optimal collision avoidance solution, u*, in the feasible region where the convex safety control set is the feasible region, as shown in Figure 5b, and its algebraic expression is:

Among them, J is the preset optimization goal, and n* and b* represent the optimal convexity results of the safety domain.

Claims (6)

1. An intelligent ship collision avoidance system oriented to human-computer interaction, which is characterized by including: The automatic ship identification and trajectory exchange module is used to obtain the basic scale information of the target ship and realize the exchange of future trajectory points or waypoints between the own ship and the target ship; The ship's autonomous decision-making module is used to eliminate the course and speed that can cause the ship to collide based on the future trajectory of the target ship obtained by the own ship, and provide the optimal collision avoidance decision. The ship's autonomous decision-making module includes: a first marking module, with Based on the scale information and future trajectory information of the target ship, it marks the geographical coordinates, geometric boundaries and time of the area where the own ship and the target ship will arrive in the future; the identification module is used to identify the areas that will cause the own ship and the target ship to occupy the same position at the same time in the future. speed obstacles; the second marking module is used to mark the speed obstacles corresponding to the future position of each target ship to obtain the speed obstacle set; and the optimal collision avoidance decision-making module is used to remove the outer envelope of the speed obstacle set, And take the convex set of the envelope to get the convex speed obstacle area, then use the half plane to represent the convex speed obstacle area, get the feasible half plane, and find the closest distance to the current heading and speed in the feasible half plane point to obtain the optimal collision avoidance decision; The ship decision space visualization module is used to visualize the decision-making process, and allows crew members to set priority decision-making areas and arbitrarily modify collision avoidance decisions in feasible decision-making areas. The ship decision-making space visualization module includes: a mapping module, which is used to map all The obtained speed obstacle set is mapped in the own ship's speed visualization space; an information acquisition module is used to obtain the priority solution area input by the crew in the speed visualization space; and a processing module is used to find in the priority solution area The decision point closest to the optimal collision avoidance decision; The ship control right takeover early warning module is used to evaluate the ratio of feasible areas and infeasible areas in the priority solution area in real time. When the ratio of feasible areas is less than the set value, an early warning is issued to remind the driver to take over the ship; and The execution module of the ship's collision avoidance decision-making is used to implement the ship's safe course and speed, and allows the crew to directly take over control of the ship's propeller speed and rudder angle at any time. 2. The intelligent ship collision avoidance system for human-computer interaction according to claim 1, characterized in that the basic scale information of the target ship includes captain, course, position and speed information. 3. The intelligent ship collision avoidance system for human-computer interaction according to claim 1, characterized in that when the crew does not input the priority solution area, the default priority solution area is [-90, 90], [ 0, v _max ], where v _max is the maximum speed of the ship. 4. An intelligent ship collision avoidance method oriented to human-computer interaction, which is characterized by including: Obtain the basic scale information of the target ship and realize the exchange of future trajectory points or waypoints between the own ship and the target ship; According to the future trajectory of the target ship obtained by the own ship, the course and speed that can cause the ship to collide are eliminated, and the optimal collision avoidance decision is provided. The optimal collision avoidance decision is obtained by: according to the scale information of the target ship and the future Trajectory information marks the geographical coordinates, geometric boundaries and time of the area where the future ship will arrive; identifies the speed obstacles that will cause the own ship and the target ship to occupy the same position at the same time in the future; marks the speed obstacles corresponding to the future position of each target ship, Obtain the speed obstacle set; remove the outer envelope of the speed obstacle set, and take a convex set of the envelope to obtain the convex speed obstacle area, and then use a half plane to represent the convex speed obstacle area to obtain a feasible half plane , and find the point closest to the current heading and speed in the feasible half-plane to obtain the optimal collision avoidance decision; Visualize the decision-making process and allow crew members to set priority decision-making areas and arbitrarily modify collision avoidance decisions in feasible decision-making areas. The process of visualizing the decision-making process includes: mapping the obtained set of speed obstacles in the speed visualization space of the own ship , obtain the priority solution area input by the crew in the speed visualization space, and find the decision point closest to the optimal collision avoidance decision in the priority solution area; Real-time assessment of the ratio of feasible areas to infeasible areas in the priority solution area. When the ratio of feasible areas is less than the set value, an early warning is issued to remind the driver to take over the ship; Carry out the ship's safe course and speed, and allow the crew to directly take over control of the ship's propeller speed and rudder angle at any time. 5. The intelligent ship collision avoidance method for human-computer interaction according to claim 4, characterized in that the basic scale information of the target ship includes captain, course, position and speed information. 6. The intelligent ship collision avoidance method for human-computer interaction according to claim 4, characterized in that when the crew does not input the priority solution area, the default priority solution area is [-90, 90], [ 0, v _max ], where v _max is the maximum speed of the ship.