Controlled time series generation for automotive software-in-the-loop testing using GANs

Parthasarathy, Dhasarathy; Bäckström, Karl; Henriksson, Jens; Einarsdóttir, Sólrún Halla

doi:10.1109/aitest49225.2020.00013

Cited by 19 publications

(4 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Shen and Zhou [145] discover latent directions that control semantically meaningful properties of data via decomposing the generator parameters without the need to access the semantic labels as in previous methods. Dhasarathy et al [165] leverages a GANs-based model and linearly interpolates in a metrics space which correlates with the properties instead of directly in the latent space.…”

Section: Post-generation Controlmentioning

confidence: 99%

“…Datasets for controllable deep time series generation include M1 dataset that contains 1,001 time series data, M3 dataset that contains 3,003 time series data and M4 dataset that contains 100,000 data [324], Tourism that includes 366 monthly series, 427 quarterly series and 518 annual series [325], NN5 that has 111 time series representing around two years of daily cash money withdrawal amounts at ATM machines at one of the various cities in the UK [326]. Vehicle and engine speed dataset contains a set of signals, recorded in a fleet of 19 Volvo buses over a 3-5 year period [165]. Philips eICU dataset contains around 200,000 patients from 208 care units across the US, with a total of 224,026,866 entries divided into 33 tables [327].…”

Section: Time Seriesmentioning

confidence: 99%

See 1 more Smart Citation

Controllable Data Generation by Deep Learning: A Review

Wang¹,

Du²,

Guo³

et al. 2022

Preprint

View full text Add to dashboard Cite

Designing and generating new data under targeted properties has been attracting various critical applications such as molecule design, image editing and speech synthesis. Traditional hand-crafted approaches heavily rely on expertise experience and intensive human efforts, yet still suffer from the insufficiency of scientific knowledge and low throughput to support effective and efficient data generation. Recently, the advancement of deep learning induces expressive methods that can learn the underlying representation and properties of data. Such capability provides new opportunities in figuring out the mutual relationship between the structural patterns and functional properties of the data and leveraging such relationship to generate structural data given the desired properties. This article provides a systematic review of this promising research area, commonly known as controllable deep data generation. Firstly, the potential challenges are raised and preliminaries are provided. Then the controllable deep data generation is formally defined, a taxonomy on various techniques is proposed and the evaluation metrics in this specific domain are summarized. After that, exciting applications of controllable deep data generation are introduced and existing works are experimentally analyzed and compared. Finally, the promising future directions of controllable deep data generation are highlighted and five potential challenges are identified.

show abstract

Section: Post-generation Controlmentioning

confidence: 99%

Section: Time Seriesmentioning

confidence: 99%

Controllable Data Generation by Deep Learning: A Review

Wang¹,

Du²,

Guo³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Some areas focused in the last five years (as identified using Google Scholar search) include automotive software architecture [26-28, 31, 44], AI-based solutions [9,42], model-based solutions [37,43,46] and blockchain [7]. These studies touch on aspects such as complexity [32], safety [32], security [38,46], privacy [7], and testing [38,40] that are relevant for automotive software.…”

Section: Automotive Softwarementioning

confidence: 99%

Painting the Landscape of Automotive Software in GitHub

Kochanthara¹,

Dajsuren²,

Cleophas³

et al. 2022

Preprint

View full text Add to dashboard Cite

The automotive industry has transitioned from being an electromechanical to a software-intensive industry. A current high-end production vehicle contains 100 million+ lines of code surpassing modern airplanes, the Large Hadron Collider, the Android OS, and Facebook's front-end software, in code size by a huge margin. Today, software companies worldwide, including Apple, Google, Huawei, Baidu, and Sony are reportedly working to bring their vehicles to the road. This paper ventures into the automotive software landscape in open source, providing a first glimpse into this multi-disciplinary industry with a long history of closed source development. We paint the landscape of automotive software on GitHub by describing its characteristics and development styles.The landscape is defined by 15,000+ users contributing to ≈600 actively-developed automotive software projects created in a span of 12 years from 2010 until 2021. These projects range from vehicle dynamics-related software; firmware and drivers for sensors like LiDAR and camera; algorithms for perception and motion control; to complete operating systems integrating the above. Developments in the field are spearheaded by industry and academia alike, with one in three actively developed automotive software repositories owned by an organization. We observe shifts along multiple dimensions, including preferred language from MATLAB to Python and prevalence of perception and decision-related software over traditional automotive software. This study witnesses open source automotive software boom in its infancy with many implications for future research and practice.

show abstract

“…Structural Similarity Index (SSIM) is a measure of similarity between two images. However, [83] use this with time series data as SSIM does not exclude itself from comparing aligned sequences of fixed length. Of course, some of these metrics are measures of similarities/dissimilarities between two probability distributions, suitable for many types of data, particularly the maximum mean discrepancy (MMD) [84] is very suitable to this task across domains.…”

Section: Evaluation Metricsmentioning

confidence: 99%

Generative adversarial networks in time series: A survey and taxonomy

Brophy,

Wang,

She

et al. 2021

Preprint

View full text Add to dashboard Cite

Generative adversarial networks (GANs) studies have grown exponentially in the past few years. Their impact has been seen mainly in the computer vision field with realistic image and video manipulation, especially generation, making significant advancements. While these computer vision advances have garnered much attention, GAN applications have diversified across disciplines such as time series and sequence generation. As a relatively new niche for GANs, fieldwork is ongoing to develop high quality, diverse and private time series data. In this paper, we review GAN variants designed for time series related applications. We propose a taxonomy of discrete-variant GANs and continuous-variant GANs, in which GANs deal with discrete time series and continuous time series data. Here we showcase the latest and most popular literature in this field; their architectures, results, and applications. We also provide a list of the most popular evaluation metrics and their suitability across applications. Also presented is a discussion of privacy measures for these GANs and further protections and directions for dealing with sensitive data. We aim to frame clearly and concisely the latest and state-of-the-art research in this area and their applications to real-world technologies.

show abstract

Controlled time series generation for automotive software-in-the-loop testing using GANs

Cited by 19 publications

References 8 publications

Controllable Data Generation by Deep Learning: A Review

Controllable Data Generation by Deep Learning: A Review

Painting the Landscape of Automotive Software in GitHub

Generative adversarial networks in time series: A survey and taxonomy

Contact Info

Product

Resources

About