EQuaTE: Efficient Quantum Train Engine for Runtime Dynamic Analysis and Visual Feedback in Autonomous Driving

Abstract

This article proposes an efficient quantum train engine (EQuaTE), a novel development tool for quantum neural network (QNN) autonomous driving software, which plots gradient variances to confirm whether the QNN falls into local minima situations (called barren plateaus). Based on this runtime visualization, the stability and feasibility of QNN-based software can be tested during runtime operations of autonomous driving functionalities. This software testing of a QNN via dynamic analysis is essentially required due to undetermined probabilistic qubit states during runtime operations. Furthermore, an EQuaTE is capable of visual feedback because the barren plateaus can be identified at local autonomous driving platforms, and the corresponding information will be visualized at remotely connected cloud. Based on this visualized information at the cloud, the QNN, which is also stored at cloud, should be automatically reorganized and retrained for eliminating barren plateaus. Then, the trained parameters can be downloaded into the QNN of autonomous driving platforms.