CN-122021368-A - Asymmetric information multi-boat collaborative defense simulation method based on Bayesian Stackelberg game

CN122021368ACN 122021368 ACN122021368 ACN 122021368ACN-122021368-A

Abstract

The invention discloses an asymmetric information multi-boat cooperative defense simulation method based on a Bayesian Stackelberg game, which is used for constructing a Bayesian Stackelberg game model with an observed boat as a leader and an observed boat as a follower aiming at an asymmetric information scene formed by proprietary type information possessed by the observed boat in the multi-unmanned-boat cooperative defense, defining a type space, a strategy space and a benefit function, dynamically deducing the proprietary type of the observed boat through an online Bayesian updating mechanism to compensate the information disadvantage of the observed boat, and solving the Bayesian Stackelberg equilibrium based on posterior beliefs to generate a distributed cooperative defense strategy of the observed boat. The invention realizes the integrated intelligent flow from intention recognition to collaborative decision, can effectively solve the problem of information asymmetry in dynamic interaction, has good real-time performance, self-adaption and expansibility, and is suitable for the collaborative scene of multiple unmanned boats such as regional patrol, collaborative task execution and the like.

Inventors

LIANG XIAO
GAO PENGCHENG
CHEN CONG
SONG YANKONG
YU CHANGDONG
LIU DIANYONG
LI WEI

Assignees

大连海事大学

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (5)

1. The asymmetric information multi-boat cooperative defense simulation method based on the Bayesian Stackelberg game is characterized by comprising the following steps of: S1, constructing a Bayesian Stackelberg game model, modeling an observed boat as a game leader with private type information, modeling an observed boat as a follower in a game, and simultaneously, constructing a private type space of the observed boat, a strategy space of the observed boat, a profit function of the observed boat, an interest function of the observed boat and an priori belief of the observed boat based on the Bayesian Stackelberg game model; S2, acquiring a posterior concept of the private type of the observed boat according to the Bayesian Stackelberg game model; S3, acquiring a joint strategy for minimizing expected cost of the observed boat according to the posterior concept of the private type of the observed boat and the Bayesian Stackelberg game model; S4, the observation boat executes the combined strategy which minimizes the expected cost of the observation boat, simulation of cooperative defense of the multiple boats is completed when interception of the observed boat is successful or the observed boat leaves the guard range of the protected area, otherwise, S2-S3 is executed again until interception of the observed boat is successful or the observed boat leaves the guard range of the protected area.
2. The asymmetric information multi-boat collaborative defense simulation method based on the Bayesian Stackelberg game according to claim 1, wherein the profit function of the observed boat is expressed as follows: wherein: a gain function for the observed boat; A strategy representing the observed boat; representing the combined strategy of all the observation boats; Is a proprietary type of boat being observed; An index of the type private to the observed boat; is a near benefit; to intercept costs; at the expense of maneuver; Wherein, the Wherein: Is that Weights of (2); the shortest distance expected to reach the mission region for the observed boat; Is a reference distance for normalization; wherein: Is that Risk avoidance weights of (2); Index for viewing boats; to observe the total number of boats; is a risk sensitivity coefficient; As an exponential decay function when The smaller this value, the larger the risk, the easier it is to intercept; is the first A time difference between arrival of the observation boat and the observed boat at the predicted junction; is the first A distance between the positions where the observation boat and the observed boat are expected to meet; is the first Observing the speed of the boat; is the current speed of the observed boat; wherein: the maneuver cost coefficient; at the moment for the observed boat A selected heading angle; at the moment for the observed boat A selected heading angle; is an absolute value.
3. The asymmetric information multi-boat cooperative defense simulation method based on the Bayesian Stackelberg game as claimed in claim 1, wherein the observation boat profit function is as follows: wherein: the representation measures the strategy taken when the boat is observed The observation boat adopts a combined strategy And the private type of the observed boat is When the ship is used, observing the total income paid by the ship; a gain function for the observed boat; is an countermeasure weight coefficient; At the cost of synergy; cost for fuel/energy consumption; Wherein, the Wherein: the weight coefficient of the cooperative cost; Representing standard deviation; is the first Predicting the intersection distance of the observation boat and the observed boat; is the first Observing the speed of the boat; Predicting a collection of interception times for all observation boats; wherein: The fuel cost weight coefficient; is the first The square of the boat speed is observed.
4. The asymmetric information multi-boat cooperative defense simulation method based on the Bayesian Stackelberg game as claimed in claim 1, wherein the posterior concept of the private type of the observed boat is obtained by adopting the following formula: wherein: Posterior concept posterior beliefs, i.e. instant moments, representing the private type of the observed boat The private type of the observed boat is Probability of (2); Is a likelihood function; To at the same time Data observed at the moment; A priori beliefs for the posterior concepts of the proprietary type of observed boat, i.e. at time of day The private type of the observed boat is Probability of (2); Wherein, the Wherein: Is a normalization constant; as a parameter of sensitivity, the sensitivity of the sensor is, ; As a difference metric function; To be at the moment The strategy adopted by the observed boat; a theoretical optimal strategy for the observed boat; Is that A gain function of the observed boat at any time; Is a strategic space of the observed boat; Is the argument parameter when the function takes the maximum value.
5. The asymmetric information multi-boat cooperative defense simulation method based on the Bayesian Stackelberg game as claimed in claim 1, wherein the formula adopted for acquiring the joint strategy for minimizing the expected cost of observing the boat is as follows: wherein: To be at the moment A joint strategy that minimizes the expected cost of viewing the boat; A strategic space for viewing the boat; To pair(s) And Solving expected observation boat cost; the representation measures the strategy taken when the boat is observed The observation boat adopts a combined strategy And the private type of the observed boat is When the ship is used, observing the total income paid by the ship; Is of private type Adopts a combined strategy for the observed boat to the observed boat Theoretical optimum strategy in time; representing maximizing variable values 。

Description

Asymmetric information multi-boat collaborative defense simulation method based on Bayesian Stackelberg game Technical Field The invention relates to the technical field of dynamic game decision and unmanned system cooperative control, in particular to an asymmetric information multi-boat cooperative defense simulation method based on Bayesian Stackelberg game. Background With the increasing wide application of Unmanned Surface Vessels (USV), many unmanned surface vessel cooperative systems are widely applied to various dynamic tasks due to flexibility, robustness and high efficiency. The unmanned ship system with the capabilities of real-time perception, intention understanding and collaborative decision making is constructed, and is very important for completing complex dynamic tasks. Unlike research under complete information conditions, in an actual multi-boat cooperative defense scenario, the observed boat generally holds private information (e.g., true intent, action preference, etc.) that is not known to the observed boat, creating an information asymmetric advantage, making cooperative defense of the observed boat more challenging. The method based on the game theory models the observed boat and the observed boat as rational decision makers, can formally describe the strategy interaction process in the cooperative defense scene, and is a theoretical framework for constructing the self-adaptive interaction system. In many game models, the Stackelberg game is highly compatible with the logic height of an observation boat according to which decisions are made after observation, due to the sequential decision structure of the leader-follower. However, the traditional Stackelberg game research has two disadvantages, namely firstly, presetting an observation boat (defender) as a game leader, which is not in accordance with the actual situation that the observed boat usually occupies a decision-making precursor in an actual scene, and secondly, not describing uncertainty caused by private type of the opposite party based on complete information assumption. The defects cause that the prior method is difficult to learn the hidden type of the observed boat on line, and cannot support the collaborative decision of the observed boat under the condition of incomplete information. Disclosure of Invention The invention discloses an asymmetric information multi-boat collaborative defense simulation method based on a Bayesian Stackelberg game, which aims to overcome the technical problems. In order to achieve the above object, the technical scheme of the present invention is as follows: an asymmetric information multi-boat cooperative defense simulation method based on Bayesian Stackelberg game comprises the following steps: S1, constructing a Bayesian Stackelberg game model, modeling an observed boat as a game leader with private type information, modeling an observed boat as a follower in a game, and simultaneously, constructing a private type space of the observed boat, a strategy space of the observed boat, a profit function of the observed boat, an interest function of the observed boat and an priori belief of the observed boat based on the Bayesian Stackelberg game model; S2, acquiring a posterior concept of the private type of the observed boat according to the Bayesian Stackelberg game model; S3, acquiring a joint strategy for minimizing expected cost of the observed boat according to the posterior concept of the private type of the observed boat and the Bayesian Stackelberg game model; S4, the observation boat executes the combined strategy which minimizes the expected cost of the observation boat, simulation of cooperative defense of the multiple boats is completed when interception of the observed boat is successful or the observed boat leaves the guard range of the protected area, otherwise, S2-S3 is executed again until interception of the observed boat is successful or the observed boat leaves the guard range of the protected area. Further, the gain function of the observed boat is expressed as follows: ; wherein: a gain function for the observed boat; A strategy representing the observed boat; representing the combined strategy of all the observation boats; Is a proprietary type of boat being observed; An index of the type private to the observed boat; is a near benefit; to intercept costs; at the expense of maneuver; Wherein, the ; Wherein: Is that Weights of (2); the shortest distance expected to reach the mission region for the observed boat; Is a reference distance for normalization; ; ; wherein: Is that Risk avoidance weights of (2); Index for viewing boats; to observe the total number of boats; is a risk sensitivity coefficient; As an exponential decay function when The smaller this value, the larger the risk, the easier it is to intercept; is the first A time difference between arrival of the observation boat and the observed boat at the predicted junction; is the first A distance