<p>With the 1999 introduction of IEEE 802.11b, the 2.4 GHz Wireless Local Area Network (WLAN) standard, the WLAN market finally began to experience the growth levels that had been expected for so long. Now, 5 GHz solutions, with the IEEE 802.11a standard leading the way, offer higher throughput and more efficient use of the spectrum. Just as the 2.4 GHz band, the 5 GHz band is unlicensed. A common concern to all unlicensed bands is interference between devices using the spectrum. Furthermore, in the 5 GHz band, WLAN cells can interfere with radar systems operating at the same frequencies. </p><p>This report describes a software model, WOK, suitable for simulations of IEEE 802.11a WLANs operating in various environments and under various ambient conditions. The WOK model can be configured extensively with respect to topology, traffic behavior, channel models, signal attenuation, interference sources and radar systems. </p><p>Further, the concepts of Dynamic Frequency Selection (DFS) and link adaptation are explored in the context of the IEEE 802.11a standard. DFS aims to avoid channels occupied by radar systems and link adaptation aims to maximize the throughput based on current ambient conditions. A DFS algorithm and a link adaptation algorithm are implemented at the Medium Access Control (MAC) layer and evaluated using the WOK model.</p>

With the 1999 introduction of IEEE 802.11b, the 2.4 GHz Wireless Local Area Network (WLAN) standard, the WLAN market finally began to experience the growth levels that had been expected for so long. Now, 5 GHz solutions, with the IEEE 802.11a standard leading the way, offer higher throughput and more efficient use of the spectrum. Just as the 2.4 GHz band, the 5 GHz band is unlicensed. A common concern to all unlicensed bands is interference between devices using the spectrum. Furthermore, in the 5 GHz band, WLAN cells can interfere with radar systems operating at the same frequencies. This report describes a software model, WOK, suitable for simulations of IEEE 802.11a WLANs operating in various environments and under various ambient conditions. The WOK model can be configured extensively with respect to topology, traffic behavior, channel models, signal attenuation, interference sources and radar systems. Further, the concepts of Dynamic Frequency Selection (DFS) and link adaptation are explored in the context of the IEEE 802.11a standard. DFS aims to avoid channels occupied by radar systems and link adaptation aims to maximize the throughput based on current ambient conditions. A DFS algorithm and a link adaptation algorithm are implemented at the Medium Access Control (MAC) layer and evaluated using the WOK model.

The C2X Communication is of high interest to the automotive industry. Ongoing research on this topic mainly bases on the simulation of Vehicular Ad Hoc Networks. In order to estimate the necessary level of simulation details their impact on the results needs to be examined. This thesis focuses on different channel models as the freespace, shadowing, and Ricean model, along with varying parameters. For these simulations the network simulator ns-2 is extended to provide IEEE 802.11p compliance. However, the WAVE mode is not considered since it is still under development and not finally approved. Besides a more sophisticated packet error model than the existing implementation, as well as a link adaptation algorithm, is added. In this thesis several simulations examine specific details of wireless communication systems such as fairness of multiple access, interferences, throughput, and variability. Furthermore, the simulation points out some unexpected phenomena as starving nodes and saturation effects in multi hop networks. Those led to the conclusion that the IEEE 802.11 draft amendment does not solve known problems of the original IEEE 802.11 standard.

Recent advancements in Transformers have unlocked a new relational analysis technique for Reinforcement Learning (RL). This thesis researches the models for DownLink Link Adaptation (DLLA). Radio resource management methods such as DLLA form a critical facet for radio-access networks, where intricate optimization problems are continuously resolved under strict latency constraints in the order of milliseconds. Although previous work has showcased improved downlink throughput in an online RL approach, time dependence of DLLA obstructs its wider adoption. Consequently, this thesis ventures into uncharted territory by extending the DLLA framework with sequence modelling to fit the Transformer architecture. The objective of this thesis is to assess the efficacy of an autoregressive sequence modelling based offline RL Transformer model for DLLA using a Decision Transformer. Experimentally, the thesis demonstrates that the attention mechanism models environment dynamics effectively. However, the Decision Transformer framework lacks in performance compared to the baseline, calling for a different Transformer model. / De senaste framstegen inom Transformers har möjliggjort ny teknik för Reinforcement Learning (RL). I denna uppsats undersöks modeller för länkanpassning, närmare bestämt DownLink Link Adaptation (DLLA). Metoder för hantering av radioresurser som DLLA utgör en kritisk aspekt för radioåtkomstnätverk, där invecklade optimeringsproblem löses kontinuerligt under strikta villkor kring latens och annat, i storleksordningen millisekunder. Även om tidigare arbeten har påvisat förbättrad länkgenomströmning med en online-RL-metod, så gäller att tidsberoenden i DLLA hindrar dess bredare användning. Följaktligen utökas här DLLA-ramverket med sekvensmodellering för att passa Transformer-arkitekturer. Syftet är att bedöma effekten av en autoregressiv sekvensmodelleringsbaserad offline-RL-modell för DLLA med en Transformer för beslutsstöd. Experimentellt visas att uppmärksamhetsmekanismen modellerar miljöns dynamik effektivt. Men ramverket saknar prestanda jämfört med tidigare forsknings- och utvecklingprojekt, vilket antyder att en annan Transformer-modell krävs.

Dans cette thèse, les techniques de coopération ont été étudiées pour un canal multi-accès multi-relais composé d'au moins deux sources qui communiquent avec une seule destination à l'aide d'au moins deux nœuds de relayage en mode semi-duplex. Le multiplexage par répartition dans le temps est supposé. Tout d'abord, l’algorithme d’adaptation de lien est exécuté par l'ordonnanceur centralisé. Durant la première phase de transmission, les sources transmettent chacune à leur tour leur message respectif pendant des intervalles de temps consécutifs. Dans chaque intervalle de temps dans la deuxième phase, la destination planifie un nœud pour transmettre les redondances, mettant en œuvre un protocole coopératif d'Hybrid Automatic Repeat reQuest (HARQ), où les canaux de contrôle limités bidirectionnels sont disponibles depuis les sources et les relais vers la destination. Dans la première partie de la thèse, les stratégies de sélection des nœuds centralisé sont proposées pour la deuxième phase de transmission. Les décisions d’ordonnancement sont prises en fonction de la connaissance des ensembles de sources correctement décodées par chaque noeud et ayant comme objectif de maximiser l’efficacité spectrale moyenne. L'analyse de la probabilité de coupure de l'information ainsi que les simulations Monte-Carlo (MC) sont effectués afin de valider ces stratégies. Dans la seconde partie, un algorithme d’adaptation de lien lent est proposé afin de maximiser l’efficacité spectrale moyenne sous contrainte de vérification d'une qualité de service individuelle cible pour une famille donnée de schémas de modulation et de codage, réposant sur l'information sur la distribution des canaux signalée. Les débits des sources discrets sont déterminés en utilisant l’approche "Genie-Aided" suivie d’un algorithme itératif de correction de débit. Les simulations MC montrent que l’algorithme d’adaptation de lien proposé offre des performances proches de celles de la recherche exhaustive. Dans la troisième partie, les performances de protocole HARQ à redondance incrémentale (IR) avec codage mono et multi-utilisateur, ainsi que l'HARQ de type Chase Combining avec codage mono-utilisateur sont comparées. Les simulations MC montrent que l'IR-HARQ avec codage mono-utilisateur offre le meilleur compromis entre performance et complexité pour le scénario de petit nombre de sources. Un schéma de codage pratique est proposé et validé à l'aide de simulations MC. / In this thesis, cooperation techniques have been studied for Multiple Access Multiple Relay Channel, consisted of at least two sources which communicate with a single destination with the help of at least two half-duplex relaying nodes. Time Division Multiplexing is assumed. First, the link adaptation algorithm is performed at the centralised scheduler. Sources transmit in turns in consecutive time slots during the first transmission phase. In each time slot of the second phase, the destination schedules a node to transmit redundancies, implementing a cooperative Hybrid Automatic Repeat reQuest (HARQ) protocol, where bidirectional limited control channels are available from sources and relays towards the destination. In the first part of the thesis, centralized node selection strategies are proposed for the second phase. The scheduling decisions are made based on the knowledge of the correctly decoded source sets of each node, with the goal to maximize the average spectral efficiency. An information outage analysis is conducted and Monte-Carlo (MC) simulations are performed to evaluate their performance. In the second part, a slow-link adaptation algorithm is proposed which aims at maximizing the average spectral efficiency under individual QoS targets for a given modulation and coding scheme family relying on the reported Channel Distribution Information of all channels. Discrete source rates are first determined using the "Genie-Aided" assumption, which is followed by an iterative rate correction algorithm. The resulting link adaptation algorithm yields performance close to the exhaustive search approach as demonstrated by MC simulations. In the third part, performances of Incremental Redundancy (IR) HARQ with Single and Multi User encoding, as well as the Chase Combining HARQ with Single User encoding are compared. MC simulations demonstrate that IR-HARQ with Single User encoding offers the best trade-off between performance and complexity for a small number of sources in our setting. Practical coding scheme is proposed and validated using MC simulations.

This thesis studies offline reinforcement learning as an optimization technique for downlink link adaptation, which is one of many control loops in Radio access networks. The work studies the impact of the quality of pre-collected datasets, in terms of how much the data covers the state-action space and whether it is collected by an expert policy or not. The data quality is evaluated by training three different algorithms: Deep Q-networks, Critic regularized regression, and Monotonic advantage re-weighted imitation learning. The performance is measured for each combination of algorithm and dataset, and their need for hyperparameter tuning and sample efficiency is studied. The results showed Critic regularized regression to be the most robust because it could learn well from any of the datasets that were used in the study and did not require extensive hyperparameter tuning. Deep Q-networks required careful hyperparameter tuning, but paired with the expert data it managed to reach rewards equally as high as the agents trained with Critic Regularized Regression. Monotonic advantage re-weighted imitation learning needed data from an expert policy to reach a high reward. In summary, offline reinforcement learning can perform with success in a telecommunication use case such as downlink link adaptation. Critic regularized regression was the preferred algorithm because it could perform great with all the three different datasets presented in the thesis. / Denna avhandling studerar offline reinforcement learning som en optimeringsteknik för nedlänks länkanpassning, vilket är en av många kontrollcyklar i radio access networks. Arbetet undersöker inverkan av kvaliteten på förinsamlade dataset, i form av hur mycket datan täcker state-action rymden och om den samlats in av en expertpolicy eller inte. Datakvaliteten utvärderas genom att träna tre olika algoritmer: Deep Q-nätverk, Critic regularized regression och Monotonic advantage re-weighted imitation learning. Prestanda mäts för varje kombination av algoritm och dataset, och deras behov av hyperparameterinställning och effektiv användning av data studeras. Resultaten visade att Critic regularized regression var mest robust, eftersom att den lyckades lära sig mycket från alla dataseten som användes i studien och inte krävde omfattande hyperparameterinställning. Deep Q-nätverk krävde noggrann hyperparameterinställning och tillsammans med expertdata lyckades den nå högst prestanda av alla agenter i studien. Monotonic advantage re-weighted imitation learning behövde data från en expertpolicy för att lyckas lära sig problemet. Det datasetet som var mest framgångsrikt var expertdatan. Sammanfattningsvis kan offline reinforcement learning vara framgångsrik inom telekommunikation, specifikt nedlänks länkanpassning. Critic regularized regression var den föredragna algoritmen för att den var stabil och kunde prestera bra med alla tre olika dataseten som presenterades i avhandlingen.