發(fā)布時(shí)間：2024-06-04 00:22

　　在流量密集的無人機(jī)ad-hoc網(wǎng)絡(luò)中,流量擁塞會(huì)增加網(wǎng)絡(luò)時(shí)延和丟包,大大限制網(wǎng)絡(luò)性能。因此,需要一個(gè)流量平衡策略控制流量。本文提出TQNGPSR,一個(gè)基于GPSR和Q網(wǎng)絡(luò)的流量感知無人機(jī)ad-hoc網(wǎng)絡(luò)路由協(xié)議。該協(xié)議利用鄰居節(jié)點(diǎn)的擁塞信息實(shí)現(xiàn)流量平衡,并用一種強(qiáng)化學(xué)習(xí)算法—Q網(wǎng)絡(luò)算法—評(píng)價(jià)當(dāng)前節(jié)點(diǎn)每條無線鏈接的質(zhì)量�；�于對(duì)這些鏈接的評(píng)估,該協(xié)議可在多個(gè)選擇中做出合理決定,降低網(wǎng)絡(luò)時(shí)延和丟包率。在仿真環(huán)境中測(cè)試TQNGPSR、AODV、OLSR、GPSR和QNGPSR。結(jié)果表明,相比于GPSR和QNGPSR, TQNGPSR有更高包到達(dá)率和更低端到端時(shí)延。在高節(jié)點(diǎn)密度場(chǎng)景中,TQNGPSR在包到達(dá)率、端到端時(shí)延和吞吐量上優(yōu)于AODV和OLSR。

【文章頁數(shù)】：14 頁

【文章目錄】：
1 Introduction
2 Background and related works
    2.1 Traffic balancing
    2.2 Reinforcement learning
    2.3 Reinforcement learning based routing protocols
3 Traffic-aware Q-network enhanced geographic routing
    3.1 Traffic balancing
    3.2 Q-network based route selection
4 Simulation results
    4.1 Simulation results and comparison with other protocols
    4.2 Simulation results under different penalty factors
    4.3 Complexity analysis and comparison with other protocols
5 Conclusions



本文編號(hào)：3988600