Data Compression in a Vehicular Environment
Johan Lovén · KTH Publication Database DiVA (KTH Royal Institute of Technology) · 2014
Vehicles with sensors produce large amounts of data. Transmission of data is connectedwith costs. By using data compression, less data has to be transmitted, reducing priceof operation. A number of algorithms are evaluated through tests with GPS and accelerometerdata. Performance is measured based on time usage, compression ratio,and euclidean errors (in the case of lossy compression). No algorithm or combinationof algorithms is shown to outperform the others on all data and all metrics. Threedierent recommendations are made. For lossless data compression focused on speed, acombination of Move-To-Front, Zero-Run-Length-Encoding, and Human coding is recommended.For lossless data compression focusing on compression ratio, a combinationof delta coding, Lempel-Ziv coding, and Human coding is recommended. Finally, fordata that can aord losses and that has a low amount of unique symbols, a model basedapproach is recommended.