Civil aviation, be it for passengers or cargo, is a highly competitive market, airlines are therefore strongly driven to increase earnings and reduce costs. The maintenance of the aircraft fleet is one pivotal aspect of this. In the industry two types of events occur, scheduled and unscheduled maintenance. While normal scheduled maintenance is already expensive, unscheduled maintenance events are even more so. The potential for savings is paramount when unscheduled events can be reduced to a minimum. Additionally, the safety of the customers is a huge concern, which is why possible failures ought to be detected as soon as possible. Over the last years, the large amounts of data that became available over the last decade open then door to a new range of applications. It got possible to learn from the past to predict future events, detect abnormal changes or behaviors, based on newly generated data. In this work we describe the application of anomaly detection on aircraft data. The goal is to predict upcoming failures of the turbine's hydraulic pumps, having severe financial implications should they be replaced in a context of unscheduled maintenance. In this context, we describe how we addressed this challenging task, and how crucial expert knowledge is when approaching such difficult undertakings. With our dataset we studied multiple outlier detection methods, ST-DBSCAN has proven to be the best suited method for this use case. We show how we identified the correct data frames to apply the methodology, and evaluate its prediction performance on a real-world dataset from several aircrafts.

Online advertising is increasingly switching to real-time bidding on advertisement inventory, in which the ad slots are sold through real-time auctions upon users visiting websites or using mobile apps. To compete with unknown bidders in such a highly stochastic environment, each bidder is required to estimate the value of each impression and to set a competitive bid price. Previous bidding algorithms have done so without considering the constraint of budget limits, which we address in this paper. We model the bidding process as a Constrained Markov Decision Process based reinforcement learning framework. Our model uses the predicted click-through-rate as the state, bid price as the action, and ad clicks as the reward. We propose a bidding function, which outperforms the state-of-the-art bidding functions in terms of the number of clicks when the budget limit is low. We further simulate different bidding functions competing in the same environment and report the performances of the bidding strategies when required to adapt to a dynamic environment.

To ensure reliable communications in randomly varying and error-prone channels, wireless systems use adaptive modulation and coding (AMC) as well as hybrid ARQ (HARQ). In order to elucidate their compatibility and interaction, we compare the throughput provided by AMC, HARQ, and their combination (AMC-HARQ) under two operational conditions: in slow- and fast block-fading channels. Considering both incremental redundancy HARQ and repetition redundancy HARQ, we optimize the rate-decision regions for AMC/HARQ and compare them in terms of attainable throughput. Under a fairly general model of the channel variation and the decoding functions, we conclude that: 1) adding HARQ on top of AMC may be counterproductive in the high average signal-to-noise ratio regime for fast fading channels and 2) HARQ is useful for slow fading channels, but it provides moderate throughput gains. We provide explanations for these results which allow us to propose paths to improve AMC-HARQ systems.

Modern wireless systems deal with the adverse and unpredictable channel conditions using two main transmission schemes considered as complimentary: adaptive modulation and coding (AMC) and hybrid ARQ (HARQ). In this work we use the effective capacity as the performance measure to evaluate different design options. We thus show how to calculate this performance measure under independent, identically distributed (i.i.d.) block-fading channel model considering AMC, HARQ, and, more importantly, a combination of thereof. Numerical results indicate that AMC alone outperforms the other schemes in the region of high average SNR; we provide an explanation for this observation and discuss possible paths for improvement.

In this work we address the issues of probabilistic modelling of the decoding errors in hybrid ARQ (HARQ) rounds. In particular we i) claim that the assumption of independence of decoding errors, used implicitly in various works on this subject, is an approximation, and ii) propose equally simple but much more accurate method to calculate the probability of the sequence of decoding errors. The model we propose is useful from the point of view of performance evaluation, system-level simulation, and/or link adaptation. Its simplicity leads also to closed form expression for the outage probability and for the average number of transmissions in block-fading channel.

We analyze the problem of transmitting information to multiple users over a shared wireless channel. The problem of resource allocation (RA) for the users with the knowledge of their channel state information has been treated extensively in the literature where various approaches trading off the users' throughput and fairness were proposed. The emphasis was mostly on the time-sharing (TS) approach, where the resource allocated to the user is equivalent to its time share of the channel access. In this work, we propose to take advantage of the broadcast nature of the channel and we adopt superposition coding (SC)-known to outperform TS in multiple users broadcasting scenarios. In SC, users' messages are simultaneously transmitted by superposing their codewords with different power fractions under a total power constraint. The main challenge is to find a simple way to allocate these power fractions to all users taking into account the fairness/throughput tradeoff. We present an algorithm with this purpose and we apply it in the case of popular proportional fairness (PF). The obtained results using SC are illustrated with various numerical examples where, comparing to TS, a rate increase between 20% and 300% is observed.