Functional and extra functional simulation of a RISC-V based unmanned aerial vehicle through SystemC-AMS

Nowadays, the rise of the RISC-V instruction set architecture (ISA) has led to an increasing interest towards the development of royalty-free and open-source products. Thanks to RISC-V, hardware architectures can be customised to suit specific product needs, offering a flexible and scalable approach that benefits both Original Equipment Manufacturer (OEM) producers and final consumers. To design products, companies strongly rely on simulations to explore the various design choices and the possible flaws of a design, reducing both the development times and costs. Being a rather recent technology, however, there is still a lack of comprehensive simulation frameworks for RISC-V cores, especially for ones targeting both functional and extra-functional aspects such as power consumption. In the landscape of functional simulation, GVSoC represents one of the most noteworthy options, allowing for a highly configurable and timing-accurate simulation for GAP8, a powerful RISC-V based IoT-oriented processor. GVSoC enables practical functional and performance analysis at the full-platform level, but lacks the support for any other external components and extra functional properties. To overcome these shortcomings, the MESSY framework has recently been developed: by combining GVSoC and the expressive power of SystemC-AMS, this tool provides a scalable yet easily customizable solution to satisfy the need of a system-level simulation framework for embedded to industrial applications. Although the software has already reached a quite mature state, very few tests have been conducted to evaluate the effectiveness of the system in simulating properties of real, complex products. This thesis thus focuses on modelling a complex robotic system within MESSY. To model the complex mechanics of a robotic system without impacting the complexity of simulation setup, MESSY has been connected to Webots, an open source application which provides a complete development environment to model, program and simulate robots and that can provide realistic sensor data to further enhance the realism of the simulation environment. The target platform for this study is the GAP8-based Crazyflie 2.1 nano drone. Nano drones have recently received a lot of interest in the academic world due to their versatility and the widespread of on-edge Artificial Intelligence (AI). As battery-powered devices, the interest in extra functional properties is extremely valuable, as simulations allow for estimating the impact on battery life of various control algorithms and the exploration of several hardware choices. Additionally, nano drones present unique challenges due to their complex control and the need for multiple sensors, making them ideal candidates for the purpose of this work. All of these aspects have been integrated in a virtual platform that combines MESSY and Webots, in order to achieve a "digital twin" version of the Crazyflie drone that closely represents its architecture behaviour and power consumption. The thesis successfully demonstrates the effectiveness of the system-level simulation capabilities offered by MESSY, along with its minimal impact on the Webots simulation times. Furthermore, the analyses of the impact on flight time of different batteries, control algorithms and environment conditions serve as an example to highlight some of the meaningful insights that can be gathered from such simulation framework, showcasing its potential for comprehensive performance evaluation and optimization.