| The satellite communication system has many advantages,such as broad coverage,long transmission distance,flexible networking,and without restrictions of topography.It has become a vital means and component of wireless communication and an important direction of future communication development.However,with increasing congestion of the spectrum space,and the continuous development and maturity of various technologies and means for reconnaissance,interception,jamming,and destruction of satellite commu-nication systems,the satellite communication system is encountering more and more seri-ous threats.On the other hand,the rapid development of artificial intelligence provides a new technical approach for satellite communication and its countermeasures.As the nerve and link of the future intelligent war,the satellite communication systems are suffering fiercer competitions,including reconnaissance and anti-reconnaissance,interception and anti-interception,jamming and anti-jamming,destroy and anti-destroy.Therefore it is difficult for the current satellite communication system,which uses the high-speed broad-band frequency-hopping(FH)as their mainstream anti-jamming measures,to adapt to the intelligent confrontation.At the same time,artificial intelligence also promotes the over-all intellectualization of the anti-jamming cognition loop,including composed of jamming cognition,anti-jamming decision,anti-jamming waveform reconstruction,anti-jamming effectiveness evaluation,and so on.Under this background,combining the characteristics of satellite communication,such as large delay,multi-user,on-board processing,etc.,the intelligent anti-jamming decision of the anti-jamming cognition loop is researched.The main results are as follows.1)Considering the large transmission delay of satellite communication systems,a multi-step prediction of Markov decision process(MPMDP)is proposed to defend the intelligent sweep jamming,which is the greatest threat to satellite communication sys-tems when the tracking-jamming condition is not met.A multi-step prediction Bellman iterative equation(MPBIE)is set up according to the MPMDP and multi-step state tran-sition probabilities are derived to solve the MPBIE.The complete strategy is proposed and a simplified algorithm is designed to maximize the utility of the FH communication system.In addition,with the assumption that the jammer can learn the MPMDP strategy of the FH communication system,we furtherly study its optional sweep jamming attack strategy.By analyzing probabilities of the communication system being jammed,the MPMDP strategy and the optional intelligent sweep jamming attack strategy as a Nash equilibrium of the communication system and the jammer are proved.2)Considering synchronous orbit satellite communication system with much more delay,in which the state feedback has not been able to provide any additional gains for the FH strategy based on MPMDP,we established an expected utility reverse recursive optimization model,which bases on the jamming cognition and makes a decision of the optimal FH speed to maximize the gain of the communication system.A corresponding optimal FH speed recursive algorithm is put forward and a quadratic explicit expression and the optimal analytical solution of the inverse recursive optimization model are ob-tained by solving the high-order matrix furtherly.It is proved that the fastest intelligent frequency sweep jamming and the best FH speed anti-jamming based on jamming cog-nition are the only Nash equilibrium strategies for communication antagonism when the tracking interference condition is not satisfied.In addition,considering the unknown jam-ming parameters,an intelligent jamming parameter learning method based on maximum likelihood is proposed.3)Considering both the user and jammer can use multiple channels and the jammer can track or block jam all the channels of the user,which leads to the failure of the intelli-gent anti-jamming decision in the frequency domain,an asymmetric Colonel Blotto game model is established for the multi-channel power allocation confrontation under the total power constraints of both sides of communication confrontation.Under the condition of complete information,the equivalent optimal single-channel power allocation strategies of the communicator and the jammer under various power budget constraints are deduced.A unique mixed Nash equilibrium strategy between the communicator and the jammer is proved,and the Nash equilibrium benefits are obtained.Based on the equivalent optimum power distribution of a single channel,a multi-scan direct column element exchange al-gorithm is proposed to construct a multi-channel mixed power distribution matrix rapidly,which can adapt to more channels and a wider power range than the linear programming method.4)Considering the satellite communication system with on-board regeneration pro-cessing,which usually requires all users to send constant power in its best range of de-modulation,leading the intelligent anti-jamming decision in the power domain invalid,a cross-domain multilinear game model of the communication antagonist is setup,in which communication party optimizes its waveforms through all available channels to defense the intelligent power allocation of the jammer.By transforming it into a minimax opti-mization problem on the equivalent single channel,the unique mixed Nash equilibrium strategy and its corresponding Nash equilibrium benefits of the game are obtained by using Karush-Kuhn-Tucker conditions.Then,to support the implementation of anti-jamming decision result of optimal waveform configuration,a modem based on the deep autoen-coder network(DAN)is designed to train a variety of waveforms by a unified network architecture,which not only support the waveform with different modulation type con-figuration hop by hop but also improves the anti-interception and anti-jamming ability of the waveforms.Finally,because of the channel synchronization problems introduced by DAN,a channel synchronization method is proposed.Through the frequency domain circular convolution,the reverse transmission blocked problem caused by the triangle activation function of the time-domain carrier synchronization network is solved.By the time domain circular convolution,the mismatch problem of the training data and the label is solved.Employing this method,carrier synchronization and bit timing synchronization can be completed quickly with little training data,the gain of the received signal can be adjusted,and the jamming can be suppressed,which make it possible to complete channel synchronization training online in the jamming environment,and provide a feasible way for the intelligent communication system based on deep learning. |
PDF Full Text Request