Smart ADAS Verification and Validation Methodology (SAVVY)

人们对新技术的需求日益强烈，以便在安全、可重复、受控和科学严格的环境中实现ADAS功能的稳健设计以及验证与确认（V&V）。这是由一系列挑战驱动的：驾驶员在驾驶任务中的参与度和依赖度降低;用例和测试场景的数量和复杂性非常高;原型车的访问减少;以及有限的测试时间，人力资源和成本限制。因此，该项目将为ADAS技术提供一种新颖，高效和加速的仿真和基于仿真器的V&V过程。该项目将使用深度学习和卷积神经网络（CNN）算法，基于现场可编程门阵列（FPGA）为未来技术的V&V创建构建模块。这些方法将在实时ADAS控制系统的整个产品开发生命周期中进行评估。该项目将促进AVL（财团牵头）、Vertizan、Myrtle Software、沃里克大学和Horiba MIRA之间的合作，并将汇集来自3个当前Innovate UK资助的可行性研究的学习和创新。

There is an emerging and strong demand for new techniques to enable the robust design and verification & validation (V&V) of ADAS features in a safe, repeatable, controlled and scientifically rigorous environment. This is driven by a number of challenges: reduced engagement of, and reliance on, the driver in the driving task; the very high number and complexity of use cases & test scenarios; reduced access to prototype vehicles; and limited test time, human resources and cost constraints. This project will therefore deliver a novel, efficient and accelerated simulation and simulator based V&V process for ADAS technologies. This project will create the building blocks for the V&V of future technologies based on Field Programmable Gate Array (FPGA) using deep learning and Convolutional Neural Network (CNN) algorithms. These methodologies will be evaluated throughout a product development lifecycle of a real-time ADAS control system. This project will facilitate collaboration between AVL (consortium lead), Vertizan, Myrtle Software, Warwick University and Horiba MIRA, and will bring together the learning and innovations from 3 current Innovate UK funded feasibility studies.