Algorithms and Architectures: A Case Study in When, Where and How to Connect Vehicles

Connected vehicle technologies and applications must coevolve. Existing applications are often suboptimally implemented, stemming from a dearth of design approaches considering resource use, sensor selection, and communication method. Such design tools are necessary to optimize connected vehicle implementations, as applications may have varied impact on individual vehicle and fleet performance, efficiency, and comfort depending on their technical and algorithmic implementation. This paper first introduces several key considerations in connected vehicle design, then explores the implications of varying input richness, connectivity methods, and data availability on an example application predicting automotive idle times. We illustrate common design tradeoffs by evaluating the predictor's accuracy for different implementations, providing developers a look at a design approach that will serve as a useful example framework for future application development. We close with a simplified cost/benefit analysis for our demonstration application to illustrate how feasible, cost-effective application implementations might be identified.