CN-122002293-A - Cooperative defense and attack behavior cognition method for multi-robot system

CN122002293ACN 122002293 ACN122002293 ACN 122002293ACN-122002293-A

Abstract

The invention discloses a collaborative defense and attack behavior cognition method for a multi-robot system, which comprises the steps of obtaining suspicious traffic intensity of defense nodes of each robot, constructing a collaborative filtering architecture, defining edge/cloud shunt proportion for each defense node, establishing an edge filtering time delay model and an individual filtering cost function, modeling a collaborative filtering process under multi-node resource sharing as a selfish filtering game, utilizing load balancing property of balanced lower edge filtering time delay consistency, designing a distributed iterative algorithm to solve a unique Nash balanced shunt strategy in a limited iteration, and constructing an optimized model of attacker budget constraint under the defense balanced strategy to solve an optimal attack strategy for cognizing, predicting and early warning attack intensity, target preference and attack type. By using the invention, the autonomous decision making capability of the multi-robot system in a complex countermeasure environment can be improved. The invention can be widely applied to the technical field of multi-robot cooperative control.

Inventors

MA QIAN
CHEN XU
WU YIFEI

Assignees

中山大学·深圳
中山大学

Dates

Publication Date: 20260508
Application Date: 20260129

Claims (8)

1. A cooperative defense and attack behavior cognition method for a multi-robot system is characterized by comprising the following steps: Acquiring resource information of a multi-robot system and defining a defending node, an edge filtering node and a cloud filtering node; estimating suspicious traffic intensity of the defending node and constructing collaborative filtering strategy variables; Calculating the filtering time delay of the edge filtering node based on the collaborative filtering strategy variable, and constructing an individual filtering cost function of the defending node; The defense node aims at minimizing an individual filtering cost function, common filtering delay parameters are introduced by utilizing the load balancing property of the filtering delay consistency of the edge filtering node under the balanced condition, and interactive calculation is carried out in a distributed iteration mode to obtain a Nash balanced collaborative filtering strategy; According to the Nash equilibrium collaborative filtering strategy, the edge filtering node and the cloud filtering node cooperatively execute suspicious flow diversion and filtering; Based on the Nash equilibrium collaborative filtering strategy, an optimized model for maximizing the total filtering cost of the system under the constraint of attack budget is constructed for an attacker, and the optimized attack strategy is obtained by solving; And outputting a behavior cognition result and early warning information for attack intensity, attack target preference and attack type based on the optimal attack strategy.
2. The cooperative defense and attack behavior recognition method for a multi-robot system according to claim 1, wherein the calculation formula of the suspicious traffic intensity is as follows: Wherein, the Representing defensive nodes Is a function of the suspicious traffic intensity of the (c), Representing defensive nodes Is used for controlling the normal flow rate of the air conditioner, Representing defensive nodes Is a malicious traffic of (1).
3. The multi-robot-system-oriented collaborative defense and attack behavior awareness method of claim 2, wherein collaborative filtering policy variables are expressed as follows: Wherein, the Representing the collaborative filtering policy variables, Distribution of representations to edge filter nodes Is used in the ratio of (a), The ratio of the allocation to the cloud is represented, Representing a set of edge filter nodes.
4. A multi-robot system oriented cooperative defense and attack recognition method according to claim 3, wherein the individual cost function is formulated as follows: Wherein, the Indicating defender Filtering strategies of other defenders outside the system, Representing the attack strategy of the attacker, Representing edge filtering nodes Is used for the filtering time delay of the (a), , And representing the cloud transmission delay coefficient.
5. The method for cognizing cooperative defense and attack behaviors of a multi-robot system according to claim 2, wherein the interactive computation process by adopting a distributed iterative manner comprises the following steps: Each edge filtering node broadcasts filtering capability parameters to the defending nodes; the defending node calculates initial delay influence parameters according to the suspicious traffic intensity and sends the initial delay influence parameters to a network management entity; The network management entity sequences the defending nodes according to the initial delay influence parameters of the defending nodes and calculates the initial common filtering delay of the system; In the first place In the round iteration: The network management entity broadcasts the current common filtering time delay to the current defending node to be updated; the defense node calculates an intermediate variable according to the current common filtering time delay, updates time delay influence parameters of the intermediate variable according to a piecewise function based on the intermediate variable, and transmits the time delay influence parameters back to a network management entity; After receiving the updated information of the defending node, the network management entity corrects the common filtering time delay by combining the time delay influence parameters of the updated defending node and the un-updated defending node to obtain the next round of common filtering time delay; And (3) carrying out loop iteration until the common filtering time delay of the next round meets the convergence condition, stopping iteration and outputting a Nash equilibrium collaborative filtering strategy.
6. The cooperative defense and attack behavior recognition method for a multi-robot system according to claim 5, wherein the update formula of the delay influencing parameter is expressed as follows: Wherein, the Representing defensive nodes To at The time delay of the round of iteration affects the parameters, Represents an intermediate variable which is referred to as, Representing defensive nodes The normal flow rate of the air is controlled, Representing defensive nodes Is used for the control of the traffic flow, Representing edge filtering nodes Is used for the filtering capacity of the filter, Representing cloud transmission delay coefficients Represent the first The system of round-robin filters the time delay, Represent the first The time delay of the round iteration affects the parameters.
7. A multi-robot system-oriented cooperative defense and attack behavior recognition system, comprising: the modeling module is used for acquiring the resource information of the multi-robot system and defining a defending node, an edge filtering node and a cloud filtering node; The flow sensing module is used for estimating the suspicious flow intensity of the defending node and constructing a collaborative filtering strategy variable; the parameter acquisition module is used for calculating the filtering time delay of the edge filtering node based on the collaborative filtering strategy variable; the game modeling module is used for constructing an individual filtering cost function of the defending node; the equalization solving module is used for enabling the defending node to aim at minimizing an individual filtering cost function, introducing common filtering delay parameters by utilizing the load balancing property of the filtering delay consistency of the edge filtering nodes under the equalization condition, and performing interactive calculation in a distributed iterative mode to obtain a Nash equalization collaborative filtering strategy; the rule issuing and filtering execution module is used for executing suspicious flow diversion and filtering cooperatively by the edge filtering node and the cloud filtering node according to the Nash equilibrium collaborative filtering strategy; the behavior cognition module is used for constructing an optimization model for maximizing the total filtering cost of the system under the constraint of attack budget by an attacker based on the Nash equilibrium collaborative filtering strategy, solving the optimization model to obtain an optimal attack strategy, and outputting a behavior cognition result and early warning information for attack intensity, attack target preference and attack type based on the optimal attack strategy.
8. A multi-robot system oriented cooperative defense and attack behavior recognition device, comprising: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor is caused to implement a multi-robot system oriented cooperative defense and attack behavior awareness method according to any of claims 1-6.

Description

Cooperative defense and attack behavior cognition method for multi-robot system Technical Field The invention relates to the technical field of multi-robot cooperative control, in particular to a cooperative defense and attack behavior cognition method for a multi-robot system. Background Multi-robot systems (e.g., unmanned aerial vehicle clusters, ground mobile robot formations, warehouse logistics robot populations) typically rely on 5G networks to achieve low latency, high bandwidth collaborative awareness, collaborative decision-making, and collaborative behavior control, and often combine with edge computing platforms to carry computationally intensive functions such as task offloading, collaborative positioning, map construction, model reasoning, and the like. However, the open access and high concurrency characteristics of 5G multi-robot systems also make them more vulnerable to denial of service attacks. An attacker can continuously send a large amount of invalid requests/flows to a 5G access network, an edge node or a cooperative control service by utilizing an external zombie terminal or an invaded robot node, so that key control links are congested, edge computing resources are exhausted, cooperative task delay is increased or even interrupted, and formation instability, task failure or safety accidents are caused. The existing DDoS defense scheme is dependent on centralized cleaning or static threshold current limiting, is difficult to adapt to the characteristics of multi-source flow, heterogeneous calculation force, dynamic topology and task load time variation in a multi-robot system, and meanwhile, the existing scheme often lacks an interpretable cognition and prediction mechanism for attack behaviors, and can limit the autonomous decision making capability of the multi-robot system in a complex countermeasure environment. Disclosure of Invention In view of this, in order to solve the technical problem that the existing multi-robot system lacks the cognition and prediction process of attack behaviors, which is further unfavorable for autonomous collaborative optimization decision-making of the multi-robot system, the invention provides a collaborative defense and attack behavior cognition method for the multi-robot system, which comprises the following steps: And system modeling, namely defining a defending node set, an edge filtering node set and cloud filtering nodes in the multi-robot system, acquiring filtering capacity and cloud transmission delay coefficients of each edge filtering node, and counting suspicious traffic intensity of each defending node. The collaborative defense game modeling comprises the steps of defining a collaborative filtering strategy for each defense node, defining filtering time delay of edge filtering nodes, constructing individual cost functions of the defense nodes, forming a selfish filtering game by each defense node with the aim of minimizing self cost, and obtaining Nash equilibrium filtering strategies. The distributed equalization solution comprises the steps of introducing common filtering delay parameters and designing a distributed iterative algorithm by utilizing the property of the consistency of the filtering delay of edge filtering nodes under the equalization condition, wherein each defending node can obtain Nash equalization in limited iteration only by updating local variables according to the self flow intensity and the broadcasted information, thereby realizing collaborative filtering of load equalization; attacker policy solution, obtaining defense equilibrium policy Then, an optimized model for maximizing the total filtering cost of the system under the budget constraint by an attacker is constructedAnd in the attack feasible domainBy means ofCognizing and predicting attack strength, attack target preference and attack type, and providing prospective configuration of early warning and defending resources for the multi-robot system. Based on the scheme, the collaborative defense and attack behavior cognition method for the multi-robot system is provided, stable collaborative filtering distribution is realized under the conditions of heterogeneous edge computing power and resource sharing, filtering congestion is remarkably reduced, the time delay robustness of collaborative tasks of the multi-robot is improved, distributed equilibrium solution is adopted, dependence on the global network state is reduced, communication and calculation expenditure is reduced, the method is suitable for a 5G multi-robot dynamic topology scene, interpretable cognition and prediction of attack behavior is realized through explicit attacker optimization modeling, and behavior cognition research under autonomous collaborative optimization decision and countermeasure environment is facilitated. Drawings FIG. 1 is a flow chart of steps of a cooperative defense and attack behavior cognition method for a multi-robot system of the present invention; FIG. 2 is a schem