.png){kind=icon}
.png){kind=icon}
Return
Enhancing storage efficiency for cloud-based vehicle data based on advanced lossless compression algorithm
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2025.129917.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Cloud-based vehicle data is being produced in various forms at an unparalleled speed, which requires efficient compression mechanisms to better store, transmit and process such data. In this article, we devise a novel, hybrid, adaptation-enabled framework, RTLAE, built on a neural network (NN) predictor and an adaptive arithmetic coding (AC) encoder, which enables outstanding lossless compression performance on non-stationary (unanticipated contextual changes) or randomly mixed vehicle datasets without requiring any domain-specific prior knowledge (lacking optimal hyperparameter settings). This framework integrates a tailor-made redundancy reduction filter, a temporal convolutional network, a long short-term memory network, and a multi-order adaptive arithmetic encoder to preprocess data sequences, generate accurate estimations, and learn optimal character encoding. Validation results on real-life datasets indicate that the proposed framework has achieved quite superior results on various datasets, with an average performance improvement of about 26%, 46%, 27%, and 12% over AC, Gzip, block-sorting compression algorithm, and DeepZip, respectively, and with a runtime cost that is slightly better than other NN-based compressors. This work emphasizes the potential of NNs and combinatorial modeling to improve general-purpose compressors, and highlights the research prospect of resource utilization of big data platforms based on vehicle-cloud collaborative interconnection.
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W
View more organizations