Flexible Computing Systems For AI Acceleration At The Extreme Edge Of The IoT

Embedding intelligence in extreme edge devices allows distilling raw data acquired from sensors into actionable information, directly on IoT end-nodes. This computing paradigm, in which end-nodes no longer depend entirely on the Cloud, offers undeniable benefits, driving a large research area (TinyML) to deploy leading Machine Learning (ML) algorithms on micro-controller class of devices. To fit the limited memory storage capability of these tiny platforms, full-precision Deep Neural Networks (DNNs) are compressed by representing their data down to byte and sub-byte formats, in the integer domain. However, the current generation of micro-controller systems can barely cope with the computing requirements of QNNs. This thesis tackles the challenge from many perspectives, presenting solutions both at software and hardware levels, exploiting parallelism, heterogeneity and software programmability to guarantee high flexibility and high energy-performance proportionality. The first contribution, PULP-NN, is an optimized software computing library for QNN inference on parallel ultra-low-power (PULP) clusters of RISC-V processors, showing one order of magnitude improvements in performance and energy efficiency, compared to current State-of-the-Art (SoA) STM32 micro-controller systems (MCUs) based on ARM Cortex-M cores. The second contribution is XpulpNN, a set of RISC-V domain specific instruction set architecture (ISA) extensions to deal with sub-byte integer arithmetic computation. The solution, including the ISA extensions and the micro-architecture to support them, achieves energy efficiency comparable with dedicated DNN accelerators and surpasses the efficiency of SoA ARM Cortex-M based MCUs, such as the low-end STM32M4 and the high-end STM32H7 devices, by up to three orders of magnitude. To overcome the Von Neumann bottleneck while guaranteeing the highest flexibility, the final contribution integrates an Analog In-Memory Computing accelerator into the PULP cluster, creating a fully programmable heterogeneous fabric that demonstrates end-to-end inference capabilities of SoA MobileNetV2 models, showing two orders of magnitude performance improvements over current SoA analog/digital solutions.

Contributions to the performance analysis of continuum parallel robots

Continuum parallel robots (CPRs) are manipulators employing multiple flexible beams arranged in parallel and connected to a rigid end-effector. CPRs promise higher payload and accuracy than serial CRs while keeping great flexibility. As the risk of injury during accidental contacts between a human and a CPR should be reduced, CPRs may be used in large-scale collaborative tasks or assisted robotic surgery. There exist various CPR designs, but the prototype conception is rarely based on performance considerations, and the CPRs realization in mainly based on intuitions or rigid-link parallel manipulators architectures. This thesis focuses on the performance analysis of CPRs, and the tools needed for such evaluation, such as workspace computation algorithms. In particular, workspace computation strategies for CPRs are essential for the performance assessment, since the CPRs workspace may be used as a performance index or it can serve for optimal-design tools. Two new workspace computation algorithms are proposed in this manuscript, the former focusing on the workspace volume computation and the certification of its numerical results, while the latter aims at computing the workspace boundary only. Due to the elastic nature of CPRs, a key performance indicator for these robots is the stability of their equilibrium configurations. This thesis proposes the experimental validation of the equilibrium stability assessment on a real prototype, demonstrating limitations of some commonly used assumptions. Additionally, a performance index measuring the distance to instability is originally proposed in this manuscript. Differently from the majority of the existing approaches, the clear advantage of the proposed index is a sound physical meaning; accordingly, the index can be used for a more straightforward performance quantification, and to derive robot specifications.