We investigate a class of finite-dimensional time decoding algorithms that: 1) is insensitive with respect to the time-encoding parameters; 2) is highly efficient and stable; and 3) can be implemented in real time. These algorithms are based on the observation that the recovery of time encoded signals given a finite number of observations has the property that the quality of signal recovery is very high in a reduced time range. We show how to obtain a local representation of the time encoded signal in an efficient and stable manner using a Vandermonde formulation of the recovery algorithm. Once the signal values are obtained from a finite number of possibly overlapping observations, the reduced-range segments are stitched together. The signal obtained by segment stitching is subsequently filtered for improved performance in recovery. Finally, we evaluate the complexity of the algorithms and their computational requirements for real-time implementation.

No abstract

Abstract-The race among manufacturers to build convenient, safe, and autonomous Connected Cars by applying the latest digital technologies, and ultimately a completely self-driving vehicle, is already underway. One of the cornerstones of such vehicles is the continuous ingestion of massive amount of data from wide variety of hardware components, including sensors, onboard cameras, and further external sources. Cloud computing and big data processing are ideal candidates and already proven technologies in order to store and process the heterogeneous, rapidly growing, and large-scale data sets. The cloud may act as a kind of central hub or as an Internet of Things (IoT) back-end where the sensor and the other available data can be gathered while also offering an elastic platform where the vast amount of data can be processed, analyzed and distributed real-time. In our paper we detail the evolution of a cloud-based, scalable IoT back-end framework and services built on top for handling and processing vehicular data in various use case scenarios: CAN data collection, remote device flashing, Eco-driving, weather report and forecast. The first version is an Infrastructure-as-a-Service (IaaS) solution with a reference implementation deployed on an OpenNebula based cloud. The second iteration runs on a private Platform-as-a-Service (PaaS) cloud built on the Cloud Foundry platform within the premises of an automotive supplier company. Both variants have been successfully evaluated and validated with benchmarks.

No abstract