Abstract
Complete and up-to-date map data plays a critical role in many contemporary and future applications such as autonomous driving level 3+. In terms of crowdsourcing, a data basis can be created that meets these stringent requirements without dedicating additional resources. With ROADR, we present a holistic platform to gather knowledge about a road network and its properties to further enhance ...
Abstract
Complete and up-to-date map data plays a critical role in many contemporary and future applications such as autonomous driving level 3+. In terms of crowdsourcing, a data basis can be created that meets these stringent requirements without dedicating additional resources. With ROADR, we present a holistic platform to gather knowledge about a road network and its properties to further enhance either semantic or syntactic information. The privacy-by-design platform uses a smartphone application to collect crowdsourced data and performs local machine learning. Only less sensitive data is forwarded to a centralized platform that aggregates and processes information from the crowd to provide value-added information found in a vehicle's trajectory. Also, the paper provides a thorough analysis of the respective Floating Phone Data indicating two exemplary events, namely traffic light and traffic circles. Our evaluation shows that the recognition is done in real-time but in a resource-efficient way.
More literature (via CORE)