Implementazione su fpga di un pre-distorsore digitale a banda larga per sistemi a microonde di backhaul

The development of next generation microwave technology for backhauling systems is driven by an increasing capacity demand. In order to provide higher data rates and throughputs over a point-to-point link, a cost-effective performance improvement is enabled by an enhanced energy-efficiency of the transmit power amplification stage, whereas a combination of spectrally efficient modulation formats and wider bandwidths is supported by amplifiers that fulfil strict constraints in terms of linearity. An optimal trade-off between these conflicting requirements can be achieved by resorting to flexible digital signal processing techniques at baseband. In such a scenario, the adaptive digital pre-distortion is a well-known linearization method, that comes up to be a potentially widely-used solution since it can be easily integrated into base stations. Its operation can effectively compensate for the inter-modulation distortion introduced by the power amplifier, keeping up with the frequency-dependent time-varying behaviour of the relative nonlinear characteristic. In particular, the impact of the memory effects become more relevant and their equalisation become more challenging as the input discrete signal feature a wider bandwidth and a faster envelope to pre-distort. This thesis project involves the research, design and simulation a pre-distorter implementation at RTL based on a novel polyphase architecture, which makes it capable of operating over very wideband signals at a sampling rate that complies with the actual available clock speed of current digital devices. The motivation behind this structure is to carry out a feasible pre-distortion for the multi-band spectrally efficient complex signals carrying multiple channels that are going to be transmitted in near future high capacity and reliability microwave backhaul links.

A study on tackling visual odometry by a transformer architecture

This dissertation describes a deepening study about Visual Odometry problem tackled with transformer architectures. The existing VO algorithms are based on heavily hand-crafted features and are not able to generalize well to new environments. To train them, we need carefully fine-tune the hyper-parameters and the network architecture. We propose to tackle the VO problem with transformer because it is a general-purpose architecture and because it was designed to transformer sequences of data from a domain to another one, which is the case of the VO problem. Our first goal is to create synthetic dataset using BlenderProc2 framework to mitigate the problem of the dataset scarcity. The second goal is to tackle the VO problem by using different versions of the transformer architecture, which will be pre-trained on the synthetic dataset and fine-tuned on the real dataset, KITTI dataset. Our approach is defined as follows: we use a feature-extractor to extract features embeddings from a sequence of images, then we feed this sequence of embeddings to the transformer architecture, finally, an MLP is used to predict the sequence of camera poses.