Research On Time-delay Signature Concealment Of Semiconductor Laser Chaotic System And Its Application To Reinforcement

With the rapid development of information technology,peoples requirements for information transmission rate are gradually increasing.Because of the randomness and confidentiality of laser chaotic entropy source,it has great potential in the fields of secure communication and high-speed random number generation.As one of the most widely used light sources,semiconductor laser can achieve high-dimensional,broadband and complex chaotic output.Laser chaotic entropy source with low time delay signature(TDS)and high complexity based on semiconductor laser has attracted much attention at home and abroad.At the same time,with the development of The Times,the field of artificial intelligence has received more and more attention.The photonic technologies are used to supplement the conventional computer algorithms,in order to overcome the electronic bottleneck of the conventional approaches and improve the energy efficiency and speed.This thesis focuses on the research of TDS concealment of semiconductor laser chaotic system and its application in reinforcement learning.The main work is summarized as follows:1.We propose and experimentally demonstrate the generation of dual-channels chaos with TDS concealment by introducing a phase-modulated Sagnac loop in mutually coupled semiconductor lasers(MCSL).For the purpose of comparison,the MCSL with a conventional Sagnac loop is also considered.The influence of tunable laser parameters on TDS concealment performance of chaotic signals is analyzed quantitatively by using the autocorrelation function as the quantifier The results show that with the introduction of phase-modulated Sagnac loop,TDS concealment of dual-channels chaotic signals can be obtained in a larger parameter range of coupling strength,frequency detuning,phase modulation and coupling delay are obtained.Finally,dual-channels chaotic signals without TDS are obtained simultaneously.2.Combining with the international research hotspots,the potential application of laser chaos in the field of reinforcement learning is further explored.The dual-channels chaotic signals are used to solve the multi-arm bandit(MAB)problem in reinforcement learning.The decision process is performed based on tug-of-war(TOW)algorithm.In addition,the key factors that affecting the learning speed are studied by theoritical and experimental investigation.The internal mechanism of reinforcement learning process is also revealed.Compared with the traditional reinforcement learning using single chaotic signal,the proposed dual-channels chaotic system has a faster learning speed,which is of great value to further improve the performance of photonic intelligence.3.In order to adapt to the development of network and improve the scalability of the number of arms in the MAB problem,we further explore the chaotic properties of semiconductor laser network and its application in reinforcement learning.The influence of tunable parameters in the laser network on TDS is analyzed quantitatively by simulation and experiment.The triple-channels TDS-concealed chaotic signals are obtained by experimentally,which is applied to slove the eight-arm bandit problem.Besides,we use triple-channels chaotic signals to solve the problem of eight-arm bandit problem in parallel,which greatly improves the learning speed.In addition,the internal mechanism of reinforcement learning process is revealed and the influencing factors of learning speed are analyzed.Finally,the scalability of the number of arms in the MAB problem is explained.