The objective of the European SARWS Celtic-plus consortium (over 30 partners from 7 countries) is to provide real-time services that ensure scalable, robust, secure, efficient, safe and energetically sustainable smart mobility. To improve road safety, the Flemish partners will research the use of crowd-sourced vehicle data to enable real-time warning services for local weather phenomena and dangerous road conditions that surpass the accuracy and timeliness of current warning systems. Local weather data will be gathered from the CAN/OBD-bus and external sensors using In-Car Smart Sensor Nodes (ICSSN), (VPS, IMEC) using a secure data distribution framework (Inuits). Initial data is obtained from IMEC and VPS test vehicles. Once prototypes are completed, 30 bpost vehicles will be equipped with ICSSNs. Local weather conditions will be extracted from the collected data using distributed machine learning algorithms (IMEC, Be-Mobile, VPS, RMI) for application in the following use cases: (i) time-series data analytics for weather-related vehicle behaviour, (ii) validating and improving the accuracy of weather and road weather models, (iii) real-time weather services that warn drivers and other stakeholders (e.g. AWV) about dangerous road conditions and an In Car Driver App (ICDA) will allow the driver to interact with the system (notifications, event tagging).
Concrete objectives and criteria
Primary targeted weather conditions are visibility (e.g. fog) and road condition (slipperiness, aquaplaning, snow, black ice). Secondary targets are precipitation (intensity, type), local temperature and wind gusts (crosswind in particular) and will be considered if research on the primary targets is successfully completed.
Smart sensing: Define a methodology for selection, calibration and fusion of sensors, CAN signals and user feedback for each of the primary (and by extension secondary) targets.
Data Distribution: Design a scalable hardware and software platform that allows data collection from a large vehicle fleet (30 bpost vehicles in SARWS, potentially the full fleet of 6500 after the project)
- for multiple weather conditions (see primary and secondary sensing objectives)
- using limited bandwidth (3G, 144kbit to 2Mbit depending on vehicle speed) to transmit vehicle data (up to 25GB/h per vehicle) in real-time without significant information loss, through data compression, reduction, collection and code distribution methodologies.
- using limited in-vehicle hardware resources (i.e. a smartphone-grade embedded CPU in the ICSSN)
- that is automatically optimized for specific data collection tasks depending on the required information using adaptive code distribution.
- that is expandable with future applications by defining software interfaces and a methodology on KPI analysis and code distribution so that future software components can be made compatible.
- Define a methodology for classifying weather conditions from mixed data streams (CAN, sensors and user feedback) given driver actions, vehicle behavior and low-resolution regional weather data.
- Verification and real-time adjustment of NWP output using this new source of highly localized data
- Extend the state-of-the-art in road weather models by blending classical inputs (NWP , radar, road weather stations) with crowd-sourced car sensor data.
- Demonstrate real-time road weather warning services for 250m road sections, for road managers such as AWV and drivers based on this new generation of road weather models.
Privacy: Identify and research technical and organizational privacy measures to (1) comply with GDPR and (2) that allow large-scale, real-time data collection without loss of road weather information, validated using KPI-analysis and regression testing.
Security: Define a secure architecture, including end-to-end encrypted V2I/I2V and soft/hardware measures ensuring read-only CAN-bus access to prevent the ICSSN from becoming an intrusion point for the vehicle