張毅恒 劉以安 宋海凌
摘 要:
針對(duì)跳頻序列設(shè)計(jì)中存在的規(guī)模小和難以兼顧多指標(biāo)的問(wèn)題,提出一種基于大規(guī)模多目標(biāo)優(yōu)化的跳頻序列設(shè)計(jì)方法。首先,綜合考慮跳頻序列的多項(xiàng)性能指標(biāo),建立跳頻序列多目標(biāo)優(yōu)化模型;然后,引入大規(guī)模多目標(biāo)優(yōu)化方法,并提出決策變量洗牌策略和反向差分進(jìn)化,通過(guò)重新分配決策變量位置以形成具有多樣性的非支配集,并通過(guò)使反向個(gè)體參與差分進(jìn)化來(lái)為后續(xù)進(jìn)化持續(xù)提供有效的方向;最后,通過(guò)提出算法對(duì)模型進(jìn)行優(yōu)化得到跳頻序列集。實(shí)驗(yàn)結(jié)果表明,所提方法相較于其他多目標(biāo)優(yōu)化方法具有更強(qiáng)的尋優(yōu)能力,得到跳頻序列集的性能指標(biāo)具有明顯優(yōu)勢(shì);所提方法在不同干擾環(huán)境中相較于其他方法具有更低的誤碼率,驗(yàn)證了提出方法的有效性和優(yōu)越性。
關(guān)鍵詞:抗干擾;跳頻序列;大規(guī)模多目標(biāo)優(yōu)化;洗牌策略;反向?qū)W習(xí)
中圖分類(lèi)號(hào):TP391?? 文獻(xiàn)標(biāo)志碼:A??? 文章編號(hào):1001-3695(2024)03-036-0887-07doi: 10.19734/j.issn.1001-3695.2023.08.0336
Frequency-hopping sequence design method based on large-scale multi-objective optimization
Zhang Yiheng1, Liu Yian1, Song Hailing2
(1.School of Artificial Intelligence & Computer Science, Jiangnan University, Wuxi Jiangsu 214122, China; 2. Naval Research Institute, Beijing 100161, China)
Abstract:
Aiming at the problem of small scale and difficulty in taking into account multiple indexes in the design of frequency-hopping sequences, this paper proposed a frequency-hopping sequence design method based on large-scale multi-objective optimization. Firstly, considering the multiple performance indexes of frequency-hopping sequence, this paper established a multi-objective optimization model of frequency-hopping sequence. Then, this paper introduced a large-scale multi-objective optimization algorithm, and proposed the decision variable shuffling strategy and the opposition-based differential evolution, which could form a diverse non-dominated set by redistributing the position of decision variables and provide an effective direction for subsequent evolution by enabling reverse individuals to participate in differential evolution and provides an effective direction for subsequent evolution. Finally, this method used the proposed algorithm to optimize the model to obtain a frequency-hopping sequence set. The experimental results show that the proposed algorithm has stronger optimization ability than other multi-objective optimization algorithms, and the performance indexes of the frequency-hopping sequence set has obvious advantages. In different interference environments, the proposed design method has lower bit error rate than other methods, which verifies the effectiveness and superiority of the proposed method. Key words:anti-jamming; frequency-hopping sequence(FHS); large-scale multi-objective optimization; shuffle strategy; opposition-based learning
0 引言
跳頻通信技術(shù)通過(guò)跳頻序列控制載波頻率跳變以實(shí)現(xiàn)頻譜的擴(kuò)展,具有抗干擾能力強(qiáng)、隱蔽性好、可實(shí)現(xiàn)碼分多址等優(yōu)點(diǎn),已成為各領(lǐng)域通信中應(yīng)用最為廣泛的一種通信技術(shù)[1~3]。
跳頻序列(FHS)的設(shè)計(jì)問(wèn)題是跳頻通信技術(shù)的核心。優(yōu)秀的跳頻序列能夠滿(mǎn)足多項(xiàng)性能指標(biāo),通常具有較低的頻率碰撞、盡可能長(zhǎng)的周期、在工作帶寬內(nèi)均勻分布以及良好的復(fù)雜度。針對(duì)頻率碰撞,研究人員基于采樣序列[4]、交錯(cuò)技術(shù)[5]、分圓定理[6]、循環(huán)碼[7]等方法,設(shè)計(jì)了達(dá)到最低碰撞界的跳頻序列或跳頻序列族,目前該類(lèi)設(shè)計(jì)方法已相對(duì)成熟,具有完備的理論體系。此外,研究人員還基于混沌理論如四維混沌系統(tǒng)[8]、混沌映射[9]等方法設(shè)計(jì)具有高隨機(jī)性的跳頻序列,以提高跳頻序列的復(fù)雜度,增強(qiáng)通信的保密性。隨著優(yōu)化算法的發(fā)展,研究人員在設(shè)計(jì)跳頻序列時(shí)開(kāi)始對(duì)混沌系統(tǒng)和優(yōu)化模型進(jìn)行優(yōu)化。文獻(xiàn)[10]利用粒子群優(yōu)化算法對(duì)三維混沌系統(tǒng)的參數(shù)進(jìn)行優(yōu)化,優(yōu)化目標(biāo)僅考慮復(fù)雜度,本質(zhì)是單目標(biāo)優(yōu)化。文獻(xiàn)[11,12]都對(duì)多個(gè)目標(biāo)進(jìn)行優(yōu)化,并利用加權(quán)法將多目標(biāo)函數(shù)轉(zhuǎn)換為單目標(biāo)函數(shù),再使用優(yōu)化算法對(duì)目標(biāo)函數(shù)進(jìn)行優(yōu)化,得到跳頻序列。
通過(guò)分析現(xiàn)有設(shè)計(jì)方法,目前對(duì)于跳頻序列的設(shè)計(jì)研究主要存在以下幾點(diǎn)問(wèn)題。首先,以最低頻率碰撞為目標(biāo)的設(shè)計(jì)方法考慮的跳頻序列指標(biāo)單一,這使得跳頻序列不能適應(yīng)各類(lèi)復(fù)雜的干擾環(huán)境,較易被第三方截獲利用;其次,基于混沌序列設(shè)計(jì)的跳頻序列,受制于計(jì)算平臺(tái)的運(yùn)算精度難以無(wú)限提升,在有限精度條件下,系統(tǒng)混沌特性退化,導(dǎo)致序列出現(xiàn)短周期現(xiàn)象,限制了混沌序列的不確定性。最后,在考慮優(yōu)化算法的跳頻序列設(shè)計(jì)中,僅利用加權(quán)法將多目標(biāo)函數(shù)轉(zhuǎn)換為單目標(biāo)函數(shù),并未考慮指標(biāo)之間的影響,且未考慮跳頻序列長(zhǎng)度增加時(shí),優(yōu)化算法易陷入局部最優(yōu)和難以收斂的問(wèn)題。
綜上所述,跳頻序列具有多項(xiàng)性能指標(biāo),為提高跳頻通信系統(tǒng)在不同干擾環(huán)境中的抗干擾能力,需要兼顧不同指標(biāo),因此可通過(guò)多目標(biāo)優(yōu)化方法來(lái)對(duì)跳頻序列的多項(xiàng)指標(biāo)進(jìn)行同時(shí)優(yōu)化。此外,在實(shí)際的跳頻通信系統(tǒng)中,尤其是軍用跳頻系統(tǒng)中,跳頻序列的長(zhǎng)度必須盡可能長(zhǎng),以避免第三方通過(guò)序列預(yù)測(cè)進(jìn)行破譯。這使得多目標(biāo)優(yōu)化中個(gè)體的決策變量維度激增,因此跳頻序列的設(shè)計(jì)問(wèn)題必須作為大規(guī)模多目標(biāo)優(yōu)化問(wèn)題(large-scale multi-objective optimization problem, LSMOP)來(lái)處理。對(duì)此,本文提出基于大規(guī)模多目標(biāo)優(yōu)化的跳頻序列設(shè)計(jì)方法。首先,以跳頻序列的最大漢明自相關(guān)、復(fù)雜度、均勻性為目標(biāo)函數(shù),建立跳頻序列多目標(biāo)優(yōu)化模型。之后,針對(duì)跳頻序列性能指標(biāo)受決策變量順序影響的特點(diǎn),引入基于增強(qiáng)搜索的大規(guī)模多目標(biāo)優(yōu)化算法,并提出決策變量洗牌策略與反向差分進(jìn)化,以提高優(yōu)化過(guò)程中非支配集的多樣性和算法的尋優(yōu)能力。實(shí)驗(yàn)證明,通過(guò)本文方法設(shè)計(jì)的跳頻序列具有更好的性能指標(biāo),并在各類(lèi)干擾環(huán)境中具有更低的誤碼率。
1 跳頻通信系統(tǒng)模型
跳頻通信技術(shù)的原理是將窄帶信號(hào)的載波頻率在跳頻序列控制下進(jìn)行離散跳變,從而達(dá)到擴(kuò)頻的效果,降低通信信號(hào)被偵察和干擾的概率。
設(shè)數(shù)據(jù)流為雙極性信號(hào)a(t),取值為±1,跳頻序列控制下的瞬時(shí)頻率為f(t),隨時(shí)間變化。
2.4 基于大規(guī)模多目標(biāo)優(yōu)化的跳頻序列多目標(biāo)優(yōu)化算法
為了延長(zhǎng)序列周期,減少跳頻序列被預(yù)測(cè)的可能,序列的長(zhǎng)度應(yīng)盡可能長(zhǎng),這使得跳頻序列多目標(biāo)優(yōu)化模型中決策空間呈現(xiàn)大規(guī)模的特點(diǎn),因此該優(yōu)化問(wèn)題必須作為L(zhǎng)SMOP來(lái)處理。
LSMOP的主要難題表現(xiàn)在兩個(gè)方面[18,19]。對(duì)目標(biāo)空間而言,由于目標(biāo)函數(shù)之間的沖突,導(dǎo)致難以存在單個(gè)最優(yōu)解,只能期望獲得一組收斂性好且分布均勻的Pareto最優(yōu)解。對(duì)決策空間而言,隨著決策變量的線性增加,決策空間規(guī)模呈指數(shù)級(jí)擴(kuò)張,容易產(chǎn)生“維度災(zāi)難”問(wèn)題。因此,常規(guī)多目標(biāo)優(yōu)化算法的優(yōu)化性能在求解LSMOP時(shí)會(huì)快速下降。目前為止,針對(duì)LSMOP,研究者提出了一些大規(guī)模多目標(biāo)進(jìn)化算法(large-scale multi-objectuve evolutianary algorithm, LSMOEA)。這些算法大致可分為三類(lèi):a)將決策變量分組,進(jìn)而將LSMOP轉(zhuǎn)換為較小規(guī)模的MOP[20,21];b)采用問(wèn)題轉(zhuǎn)換的方法將LSMOP轉(zhuǎn)換為低維問(wèn)題[22];c)基于增強(qiáng)搜索[23,24],通過(guò)設(shè)計(jì)新的進(jìn)化算子和概率模型,來(lái)對(duì)決策變量進(jìn)行高效的整體優(yōu)化。
在跳頻序列多目標(biāo)優(yōu)化模型中,優(yōu)化目標(biāo)不止取決于決策變量的取值,還與決策變量的順序直接相關(guān),因此決策變量之間的關(guān)系相較于普通問(wèn)題更為復(fù)雜。若采用決策變量分組的方法,交互變量易被分至不同組,且分組效果易不穩(wěn)定,子問(wèn)題仍可能是LSMOP。若采用問(wèn)題轉(zhuǎn)換的方法,盡管可以縮小搜索范圍,但容易丟失全局最優(yōu)解,且較難選取合適的轉(zhuǎn)換函數(shù)。
本文引入Zhang等人[25]提出的LSMaODE算法,該算法基于增強(qiáng)搜索的思想,能夠?qū)Υ笠?guī)模決策變量進(jìn)行整體優(yōu)化,并且不會(huì)破壞決策變量之間的順序關(guān)系,更適用于跳頻序列多目標(biāo)優(yōu)化。首先,將種群分為兩個(gè)子種群,對(duì)其中10%的個(gè)體采用隨機(jī)坐標(biāo)下降(randomized coordinate descent,RCD),以獨(dú)立開(kāi)發(fā)和探索決策變量,并且保證決策空間的多樣性以避免過(guò)早收斂到局部最優(yōu)。其次,剩余90%的個(gè)體根據(jù)非支配引導(dǎo)隨機(jī)插值(nondominated guided random interpolation,NGRI)進(jìn)行變異,隨機(jī)選擇三個(gè)非支配解,并在其中插值生成新個(gè)體,從而在引導(dǎo)子種群快速收斂到非支配解的同時(shí),保持個(gè)體良好的分布。
此外,針對(duì)跳頻序列在多目標(biāo)優(yōu)化過(guò)程中的特點(diǎn),提出決策變量洗牌策略與反向差分進(jìn)化,并與LSMaODE相結(jié)合,提出針對(duì)跳頻序列多目標(biāo)優(yōu)化模型的LSMaODE-FHS算法,以提高種群中個(gè)體跳出局部最優(yōu)的能力和Pareto最優(yōu)解集的多樣性。
2.4.1 RCD
設(shè)P(i)t是第t代群體Pt中的第i個(gè)個(gè)體,x(i)k是個(gè)體P(i)t的第k個(gè)決策變量。對(duì)于每個(gè)個(gè)體,對(duì)其決策變量獨(dú)立進(jìn)行變異和選擇。令NewP(i)t=P(i)t,如果隨機(jī)數(shù)大于0.1,則通過(guò)一個(gè)標(biāo)準(zhǔn)差為(UBk-LBk)/10的高斯隨機(jī)數(shù)來(lái)生成突變算子,對(duì)決策變量進(jìn)行變異,得到新個(gè)體的決策變量NewP(i)t.xk。否則,從種群中隨機(jī)選擇三個(gè)候選個(gè)體P(r1)t、P(r2)t和P(r3)t,通過(guò)差分進(jìn)化算子進(jìn)行變異,描述如下:
2.6 時(shí)間復(fù)雜度分析
RCD的時(shí)間復(fù)雜度為O(N2P2), NGRI的時(shí)間復(fù)雜度為O(MP2)。由于N>>M,則O(N2P2)>>O(MP2),所以LSMaODE的時(shí)間復(fù)雜度可記為O(N2P2)。
決策變量洗牌策略的時(shí)間復(fù)雜度為O(NP),在改進(jìn)RCD后,時(shí)間復(fù)雜度保持不變;反向差分進(jìn)化無(wú)須計(jì)算目標(biāo)函數(shù),因此時(shí)間復(fù)雜度為O(NP),在改進(jìn)NGRI后,時(shí)間復(fù)雜度為O(NP+MP2)。由于O(N2P2)>>O(NP+MP2),所以改進(jìn)后算法的時(shí)間復(fù)雜度仍可記為O(N2P2)。
3 實(shí)驗(yàn)分析
3.1 優(yōu)化性能分析
本實(shí)驗(yàn)中,采用超體積指標(biāo)HV(hypervolume)指標(biāo)[29]來(lái)評(píng)估多目標(biāo)優(yōu)化算法的性能。HV指標(biāo)能夠根據(jù)非支配集個(gè)體與參考點(diǎn)在目標(biāo)空間中所圍成的超立方體體積,對(duì)解集的收斂性和多樣性進(jìn)行評(píng)價(jià)。HV越大,解集的收斂性和多樣性越好。
令種群規(guī)模為100,迭代次數(shù)為200,LB=3,UB=35,跳頻序列長(zhǎng)度為N,頻點(diǎn)數(shù)q=32。指標(biāo)計(jì)算中,散布熵參數(shù)取m=2、c=4、d=1。
對(duì)比算法選用LMEA[20]、LMOCSO[24]、IM-MOEA/D[30]、NAS-MOEA[31]、LSMaODE和LSMaODE-FHS,在跳頻序列多目標(biāo)優(yōu)化模型上的針對(duì)不同跳頻序列長(zhǎng)度各自重復(fù)實(shí)驗(yàn)30次,計(jì)算HV指標(biāo),并將所得解集轉(zhuǎn)換為跳頻序列集,計(jì)算最大漢明自相關(guān)、散布熵和均勻性的平均值,如表1~4所示。
LMEA、LMOCSO在各測(cè)試中的性能表現(xiàn)明顯差于其他算法,說(shuō)明聚類(lèi)分組、競(jìng)爭(zhēng)粒子群優(yōu)化的進(jìn)化方式較難處理跳頻序列多目標(biāo)優(yōu)化問(wèn)題。IM-MOEA/D與NAS-MOEA的優(yōu)化性能雖然優(yōu)于LMEA與LMOCSO,但差于LSMaODE與LSMaODE-FHS,這說(shuō)明更為精細(xì)的分組策略和進(jìn)化方法對(duì)優(yōu)化性能有所提升,但優(yōu)化程度仍舊不足,得到跳頻序列集的各項(xiàng)指標(biāo)還是較差。LSMaODE采用增強(qiáng)搜索的方法,對(duì)決策變量進(jìn)行整體優(yōu)化,在N=128、N=256時(shí),取得了質(zhì)量較高的解集,但隨著決策變量的增加,得到跳頻序列集的最大漢明自相關(guān)明顯變差,但均勻性變好,這是由于 LSMaODE在RCD中對(duì)每個(gè)決策變量獨(dú)立變異,更易找到均勻性更好的個(gè)體,而最大漢明自相關(guān)需要考慮到變量的順序,所以在高維時(shí)難以?xún)?yōu)化。LSMaODE-FHS在所有測(cè)試中,取得了最好的HV指標(biāo)和最好的最大漢明自相關(guān),復(fù)雜度在高維時(shí)也能取得最好,均勻性雖差于LSMaODE,但差距不大,說(shuō)明加入決策變量洗牌策略后,個(gè)體在RCD過(guò)程中具備更多變異可能,因此最大漢明自相關(guān)和復(fù)雜度的優(yōu)化更為充分,再配合反向差分進(jìn)化,使各項(xiàng)指標(biāo)得到均衡優(yōu)化,從而得到收斂性和多樣性更好的解集。
為進(jìn)一步對(duì)算法性能進(jìn)行分析,給出不同決策變量維度下各算法的解集在目標(biāo)空間的分布,如圖3~6所示。
LMEA解集分布位置較差,說(shuō)明聚類(lèi)分組的方法在跳頻序列多目標(biāo)問(wèn)題中的優(yōu)化程度不足,得到的跳頻序列指標(biāo)較差。LMOCSO雖然分布位置與LSMaODE和LSMaODE-FHS近似,但多出了很多優(yōu)化不均衡的個(gè)體,說(shuō)明競(jìng)爭(zhēng)粒子群優(yōu)化的進(jìn)化方式對(duì)跳頻序列的各項(xiàng)指標(biāo)存在顧此失彼的情況。IM-MOEA/D與NAS-MOEA的解集分布都較為分散,且各項(xiàng)性能指標(biāo)明顯不夠優(yōu)秀,說(shuō)明IM-MOEA/D與NAS-MOEA的優(yōu)化不夠充分,解集在目標(biāo)空間上沒(méi)有充分逼近Pareto前沿。在N=128時(shí),LSMaODE與LSMaODE-FHS解集分布近似;在N=256、N=512、N=1 024時(shí),LSMaODE解集的分布在最大漢明自相關(guān)上更大,而LSMaODE-FHS加入決策變量洗牌策略,因此分布位置更好,同時(shí)反向差分進(jìn)化增強(qiáng)了算法尋優(yōu)能力,避免了優(yōu)化不均衡個(gè)體的出現(xiàn)。
綜合來(lái)看,相較于其他大規(guī)模多目標(biāo)優(yōu)化算法,LSMaODE-FHS在跳頻序列多目標(biāo)優(yōu)化模型中具有更好的尋優(yōu)能力,得到的解集具有更好的收斂性和多樣性,對(duì)應(yīng)的跳頻序列集性能指標(biāo)綜合更好。
3.2 抗干擾性能分析
抗干擾性能分析基于某軍用跳頻通信設(shè)備,信息脈沖共32 bit,脈寬0.5 μs,采樣頻率200 MHz,上采樣倍數(shù)為100倍。脈沖成型采用根升余弦滾降濾波器,滾降參數(shù)R=0.22,碼元速率0.5 MHz,基帶濾波采樣速率2 MHz,工作帶寬為3 MHz~35 MHz,跳頻間隔為1 MHz,頻點(diǎn)數(shù)q=32,跳頻序列長(zhǎng)度N=512。
令發(fā)送時(shí)間為10 s,考慮到實(shí)際情況中的多變干擾環(huán)境,將發(fā)送分為前后半段,構(gòu)建多種干擾環(huán)境,如表5所示。
將文獻(xiàn)[7]得到的跳頻序列集定義為FHS1,文獻(xiàn)[8]得到的跳頻序列為FHS2,文獻(xiàn)[11]得到的跳頻序列為FHS3,本文方法得到的跳頻序列集為FHS4。
設(shè)置干擾信噪比為-10 dB到0 dB,每間隔2 dB進(jìn)行30次測(cè)試。對(duì)FHS1和FHS4,在測(cè)試中對(duì)跳頻序列集中的每個(gè)序列進(jìn)行分別測(cè)試,并將誤碼率取平均值,作為最終的結(jié)果。計(jì)算不同設(shè)計(jì)方法在不同干擾環(huán)境中的誤碼率,如圖7~10所示。
FHS1與FHS2的誤碼率較為接近,這主要是由于兩者均能實(shí)現(xiàn)較長(zhǎng)的跳頻序列,但FHS1僅考慮漢明相關(guān)性,而FHS2基于混沌系統(tǒng),兩者參與的性能指標(biāo)不足,所以誤碼率相較于FHS4略高。FHS3基于優(yōu)化算法對(duì)加權(quán)目標(biāo)函數(shù)進(jìn)行優(yōu)化,在大規(guī)模優(yōu)化中普通優(yōu)化算法難以收斂,且只能針對(duì)一種干擾類(lèi)型,因此誤碼率大幅上升。FHS4在各干擾環(huán)境中均保持了較低誤碼率,說(shuō)明跳頻序列多目標(biāo)優(yōu)化模型能夠設(shè)計(jì)出具有較強(qiáng)適應(yīng)力的跳頻序列集,且LSMaODE-FHS能夠保證長(zhǎng)序列的抗干擾能力。
4 結(jié)束語(yǔ)
本文對(duì)跳頻序列的設(shè)計(jì)問(wèn)題進(jìn)行了研究,通過(guò)分析現(xiàn)有設(shè)計(jì)方法的優(yōu)缺點(diǎn),針對(duì)跳頻序列設(shè)計(jì)中優(yōu)化目標(biāo)少、決策變量規(guī)模小的問(wèn)題,提出基于大規(guī)模多目標(biāo)優(yōu)化的跳頻序列設(shè)計(jì)方法。綜合考慮跳頻序列的最大漢明自相關(guān)、復(fù)雜度和均勻度,結(jié)合大規(guī)模多目標(biāo)優(yōu)化理論,提出LSMaODE-FHS算法,在RCD中加入決策變量洗牌策略以提高非支配集的多樣性,在NGRI中加入反向差分進(jìn)化,將候選個(gè)體的反向個(gè)體參與到差分進(jìn)化中,提高算法尋優(yōu)能力。
該方法從大規(guī)模多目標(biāo)優(yōu)化的角度入手,在高維決策變量情況下,對(duì)各優(yōu)化目標(biāo)進(jìn)行同時(shí)優(yōu)化,以兼顧跳頻序列的各項(xiàng)性能指標(biāo),更貼合跳頻序列的實(shí)際使用需求。此外,該方法一次可以得到多個(gè)跳頻序列,在實(shí)際使用中可以隨時(shí)變更跳頻序列,相較于單目標(biāo)優(yōu)化具有更高的使用價(jià)值和保密性。
參考文獻(xiàn):
[1]劉鑫,陳璐瑤,姚昌華,等. 一種新型的增強(qiáng)型差分跳頻通信抗干擾系統(tǒng) [J]. 計(jì)算機(jī)應(yīng)用研究,2022,39(6): 1820-1824. (Liu Xin,Chen Luyao,Yao Changhua,et al. Novel anti-jamming system for enhanced differential frequency hopping [J]. Application Research of Computers,2022,39(6): 1820-1824
[2]Alexandre F,Miguel L,André Z. A fair channel hopping scheme for lora networks with multiple single-channel gateways [J]. Sensors,2022,14: 5260-5260.
[3]譚思煒,唐波,張林森,等. 跳頻電磁引信干擾感知技術(shù)方案研究 [J]. 系統(tǒng)工程與電子技術(shù),2022,44(11): 3330-3337. (Tan Siwei,Tang Bo,Zhang Linsen,et al. Interference sensing technical solution for frequency hopping electromagnetic fuze [J]. Systems Engineering and Electronics,2022,44(11): 3330-3337.)
[4]Yin Wenjuan,Xiang Can,F(xiàn)u Fangwei. New frequency-hopping sequence sets with good parameters under aperiodic hamming correlation [J]. Cryptography and Communications,2022,15(1): 159-169.
[5]Zhou Limengnan,Liu Xing. Families of optimal low-hit-zone frequency-hopping sequence sets under the periodic partial hamming correlation properties [J]. IEEE Access,2020,8: 14991-14998.
[6]Fan Mingyue. Frequency-hopping sequences with optimal average Hamming correlation and their applications in energy and spectrum harvesting technologies area [J]. IEEE Access,2021,9: 1388-1393.
[7]Niu Xianhua,Xing Chaoping,Yuan Chen. Asymptotic Gilbert-Varshamov bound on frequency hopping sequences [J]. IEEE Trans on Information Theory,2020,66(2): 1213-1218.
[8]楊宇曉,汪德鑫,黃琪. 四維超混沌射頻隱身跳頻通信設(shè)計(jì)方法 [J]. 宇航學(xué)報(bào),2020,41(10): 1341-1349. (Yang Yuxiao,Wang Dexin,Huang Qi. Design method of radio frequency stealth frequency hopping communications based on four-dimensional hyperchaotic system [J]. Journal of Astronautics,2020,41(10): 1341-1349.)
[9]Jung J,Lim J,Park S,et al. Analysis of low probability of detection capability for chaotic standard map-based FH-OFDMA system [J]. Applied Sciences,2021,11(5): 2198-2198.
[10]劉鵬飛,杜欣軍. 基于加權(quán)反饋三維混沌系統(tǒng)的調(diào)制跳變通信 [J]. 計(jì)算機(jī)工程,2021,47(3): 183-189,195. (Liu Pengfei,Du Xinjun. Modulation hopping communication based on three-dimensional chaotic system with weighted feedback [J]. Computer Engineering,2021,47(3): 183-189,195.)
[11]趙知?jiǎng)?,王安?qiáng),尚俊娜,等. 基于灰狼算法的抗干擾跳頻序列設(shè)計(jì) [J]. 信號(hào)處理,2021,37(6): 1046-1054. (Zhao Zhijin,Wang Anqiang,Shang Junna,et al. Design of anti-jamming frequency-hopping sequence based on grey wolf algorithm [J]. Journal of Signal Processing,2021,37(6): 1046-1054.)
[12]黃琪,楊宇曉,江陳卓. 組合跳變隨機(jī)平移寬間隔混沌跳頻序列設(shè)計(jì) [J]. 電訊技術(shù),2022,62(6): 755-761. (Huang Qi,Yang Yuxiao,Jiang Chenzhuo. Design of combined-hopping-random-shift wide-gap chaotic frequency hopping sequences [J]. Telecommunication Engineering,2022,62(6): 755-761.)
[13]Cao Bin,F(xiàn)an Shanshan,Zhao Jianwei,et al. Quantum-enhanced multiobjective large-scale optimization via parallelism [J]. Swarm and Evolutionary Computation,2020,57: 100697.
[14]Tian Ye,Si Langchun,Zhang Xingyi,et al. Evolutionary large-scale multi-objective optimization: a survey [J]. ACM Computing Surveys,2021,54(8): 1-34.
[15]Lempel A,Greenberger H. Families of sequences with optimal Hamming correlation properties [J]. IEEE Trans on Information Theory,1974,20(1): 90-94.
[16]Rostaghi M,Azami H. Dispersion entropy: a measure for time series analysis [J]. IEEE Signal Processing Letter,2016,23(5): 610-614.
[17]甘良才,吳燕翔. 一類(lèi)混沌映射產(chǎn)生跳頻序列的方法 [J]. 電子學(xué)報(bào),2000,28(4): 109-111. (Gan Liangcai,Wu Yanxiang. Genera-ting FH sequences by a class of chaotic maps [J]. Acta Electronica Sinica,2000,28(4): 109-111.)
[18]Hong Wenjing,Yang Peng,Tang Ke. Evolutionary computation for large-scale multi-objective optimization: a decade of progresses [J]. International Journal of Automation and Computing,2021,18(2): 1-15.
[19]Zhang Maoqing,Wang Lei,Guo Weian,et al. Many-objective evolutionary algorithm based on relative non-dominance matrix [J]. Information Sciences,2021,547: 963-983.
[20]Zhang Xingyi,Tian Ye,Cheng Ran,et al. A decision variable clustering-based evolutionary algorithm for large-scale many-objective optimization [J]. IEEE Trans on Evolutionary Computation,2018,22(1): 97-112.
[21]Liu Songbai,Lin Qiuzhen,Tian Ye,et al. A variable importance-based differential evolution for large-scale multiobjective optimization [J]. IEEE Trans on Cybernetics,2022,52(12): 13048-13062.
[22]Tian Ye,Lu Chang,Zhang Xingyi,et al. Solving large-scale multi-objective optimization problems with sparse optimal solutions via unsupervised neural networks [J]. IEEE Trans on Cybernetics,2021,51(6): 3115-3128.
[23]Chen Huangke,Cheng Ran,Wen Jinming,et al. Solving large-scale many-objective optimization problems by covariance matrix adaptation evolution strategy with scalable small subpopulations [J]. Information Sciences,2020,509(C): 457-469.
[24]Tian Ye,Zheng Xiutao,Zhang Xingyi,et al. Efficient large-scale multiobjective optimization based on a competitive swarm optimizer [J]. IEEE Trans on Cybernetics,2019,50(8): 1-13.
[25]Zhang Kai,Shen Chaonan,Yen G G. Multipopulation-based differential evolution for large-scale many-objective optimization [EB/OL]. (2022-06-22) [2023-08-09]. http://doi. org/10. 1109/TCYB. 2022. 3178929.
[26]Zhang Kai,Shen Chaonan,Yen G G,et al. Two-stage double niched evolution strategy for multimodal multiobjective optimization [J]. IEEE Trans on Cybernetics,2022,25(4): 754-768.
[27]Zhang Kai,Xu Zhiwei,Xie Shengli,et al. Evolution strategy-based many-objective evolutionary algorithm through vector equilibrium [J]. IEEE Trans on Cybernetics,2021,51(11): 5455-5467.
[28]Tizhoosh H R. Opposition-based learning: a new scheme for machine intelligence [C]//Proc of International Conference on Computational Intelligence for Modelling,Control and Automation.Piscataway,NJ: IEEE Press,2005: 695-701.
[29]Zitzler E,Thiele L. Multiobjective evolutionary algorithms: a compa-rative case study and the strength Pareto approach [J]. IEEE Trans on Evolutionary Computation,1993,3(4): 257-271.
[30]Lucas R C,Aluizio F R. IM-MOEA/D: an inverse modeling multi-objective evolutionary algorithm based on decomposition [C]// Proc of IEEE International Conference on Systems,Man,and Cybernetics. Piscataway,NJ:IEEE Press,2021: 462-467.
[31]閆世瑛,顏克斐,方偉,等. 基于差分進(jìn)化鄰域自適應(yīng)的大規(guī)模多目標(biāo)算法 [J]. 系統(tǒng)工程與電子技術(shù),2022,44(7): 2112-2124. (Yan Shiying,Yan Kefei,F(xiàn)ang Wei,et al. Large-scale multi-objective algorithm based on neighborhood adaptive of differential evolution [J]. Systems Engineering and Electronics,2022,44(7): 2112-2124.)