Optimization of artificial intelligence algorithms for deployment in embedded systems with application to autonomous driving

基于大型神经网络的人工智能算法目前在几个计算机视觉任务中提供了最先进的性能，包括实时目标识别和语义分割。深度学习领域的这种最新进展推动了深度学习在自动驾驶等传感应用中的使用。然而，过高的计算和能耗要求仍然是在受限的嵌入式设备上部署的重要障碍。最近探索的解决方案空间在于在设备上使用之前以某种方式压缩-近似或简化-深度神经网络。在这个项目的第一部分，我们将探索压缩任何类型的非常大和深的神经网络的通用技术-包括完全连接的、卷积的、递归的神经网络及其组合，并获得参数冗余的全局视图。将探索一些方法，包括强化学习。在该项目的第二部分，我们将与捷豹路虎(JLR)合作，针对自动驾驶中的一系列应用进一步优化算法。该项目属于EPSRC的职权范围，该委员会强烈认识到人工智能研究对数据科学、机器人和自主系统的重要性。这是蒙大拿州数据科学小组和捷豹路虎新成立的自动驾驶小组之间的新合资企业，该小组由约尔格·施林克海德领导。

AI algorithms based on large neural networks currently provide state-of-the-art performance in several computer vision tasks, including real-time object recognition and semantic segmentation. Such recent advances in deep learning motivate the use of deep learning in sensing applications such as autonomous driving. However, the excessive computational and energy consumption requirements remain an important impediment for the deployment on constrained embedded devices. A recently explored solution space lies in compressing - approximating or simplifying - deep neural networks in some manner before use on the device. In the first part of this project, we will explore general-purpose techniques for compressing any type of very large and deep neural network - including fully-connected, convolutional, recurrent neural networks, as well as their combinations, and obtaining a global view of parameter redundancies. A number of methodologies will be explored, including reinforcement learning. In the second part of the project, in collaboration with Jaguar Land-Rover (JLR), we will further optimise the algorithms for a range ofapplications in autonomous driving. The project falls within the remit of the EPSRC, which strongly recognises AI research's importance to data science, robotics and autonomous systems. This is a new venture between Montana's Data Science group and JLR's newly established group in Automated Driving headed by Joerg Schlinkheider.