Deep Learning Based Automotive Requirements Analysis

Author:

D H Sharath,P.C. Karthik,TG Sreekanth,Ansari Asadullah

Abstract

<div class="section abstract"><div class="htmlview paragraph">Automotive system functionalities spread over a wide range of sub-domains ranging from non-driving related components to complex autonomous driving related components. The requirements to design and develop these components span across software, hardware, firmware, etc. elements. The successful development of these components to achieve the needs from the stockholders requires accurate understanding and traceability of the requirements of these component systems. The high-level customer requirements transformation into low level granularity requires an efficient requirement engineer. The manual understanding of the customer requirements from the requirement documents are influenced by the context and the knowledge gap of the requirement engineer in understanding and transforming the requirements. The manual way of understanding the requirements of the automotive systems always involve a certain amount of ambiguity, misunderstanding, bias etc. in analyzing the functionality of the requirements. The complex automotive system, which is solely developed based on human understanding always causes some violations in transforming the actual requirements from the stockholders in a product functionality. Hence, to mitigate this human influence on this aspect of requirement understanding, an intelligent system, which either to assists the manual requirement analysis or completely analyze the requirements alone is required. In this regard, an intelligent system is proposed here to analyze the automotive requirements efficiently by reducing human conflict, manual efforts, and to improve design and execution performance of an automotive component. The proposed system uses deep learning based Natural Language Processing (NLP) based algorithms to analyze and understand the requirement corpus from a set of platform requirements. The training of the deep learning CNN algorithms is performed on a huge set of pre-implemented platform requirements. The inference of the new customer requirements is done using the trained deep learning-based models to classify the requirements into one of the pre-defined platform requirement classes, thereby assisting the manual analysis using its intelligent component by also providing the traceability.</div></div>

Publisher

SAE International

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3