Sometimes drivers are driving too close to another car driving ahead. Many drivers have a false belief that in case of the car in front starts to brake they can react, brake and come to a stop, still leaving enough distance between the two vehicles. The distance between cars, called as a safe following distance, should be as long as the vehicle is possible to stop without collision in case of accident or other sudden braking. Suitable safe following distance improves the fluency and safety on the roads but the local weather and road condition should be taken into account, too.
The total stopping distance of a vehicle depends on four things:
This braking distance application has been done with several D2I project partners. In the first phase the braking distance warning information can be calculated for one pre-specified spot where the Noptel’s instruments are installed. The place is called as Tuiran sillat, which locates on a bridge leading north from the city center of Oulu. The needed input data is velocity of the cars, distance between cars and the road surface friction. The estimated braking distance can be calculated if parameters mentioned above are available. The application is running on Centria’s vehicle where the dashboard is extended by an Android tablet which visualizes the braking distance warning. Centria has developed the needed code for the application. The Android program gets the needed data from the database, does the calculation and shows the warning sign if needed.
M. Hippi, J. Miettinen, J. Jämsä & J. Pahkala (2014). Braking distance application developed on Finnish D2I project. 17th International Road Weather Conference SIRWEC 2014. 30 January - 1 February 2014, la Massana, Andorra.
J. Pahkala, T. Karjalainen & M. Hippi (2014). Winter maintenance quality monitoring and stopping distance evaluation. 10th ITS European Congress, Helsinki, Finland 16-19 June 2014.
J. Jämsä, S. Pieskä, M. Luimula (2013). Situation-Awareness In Cognitive Transportation Systems. Infocommunications Journal, December 2013. Volume V, Number 4 pp. 10-16.
The driver decisions which affect fuel consumption can be divided into strategic (vehicle selection and maintenance), tactical (route selection and vehicle road), and operational (driver behaviour). Also, vehicle selection has a dominating effect on fuel economy, but the remaining factors can contribute, in total, to about 45% reduction in fuel consumption. It has been studied that the most influential factors include stops during run, extreme acceleration,and late change from 2nd to 3rd gear. In turn, driving behaviour depends on many factors, among others are street and traffic environment. For instance, the density of junctions controlled by traffic lights seems to have a high effect on driving behaviour and hence on fuel consumption and car emissions.
Driving coach aims to improve driving of individual, specifically to avoid aggressive driving, concerning trip planning, and driving in a fuel-efficient manner by drive analysis and comments provisioning. To achieve this, Driving coach fuses different kinds of information. That is, driving behavior retrieved from OBDII device is combined with the information retrieved from national spatial database. This information is aggregated with traffic fluency situation and road weather information retrieved from third party services. Integrating all this information together allows to understand better the actual situation of a driver. Therefore, personal driving habits affecting the fuel consumption in certain situations can be identified by Driving coach.
Distinguished feature of Driving coach is capability of the system to adapt its decision-making with respect to situation, driver's progress and responses to recommendations. This is required, as driving behaviour develops, therefore, to provide adequate feedback, Driving coach deals with changes and evolves with a user progress. Moreover, different situations may have their effect on driving behaviour, e.g. slippery roads may cause drivers to drive with slower speeds. Driving coach provides situation-aware feedback to the driver.
The architecture of the system is presented on the right. Driving coach is a distributed system, operating with diverse third-party services. System core utilizes machine-learning and rule-based reasoning to develop fuel prediction models and provide appropriate feedback to the driver. Driving coach provides great opportunities to develop diverse client solutions to deliver fuel-efficiency information to a driver.
Development of interfaces to deliver such information to the drivers requires more thorough research. We have started this study and designed first solution to represent such information from mobile phone GUI. Our application demonstrates the spatial information about the route driven, as well as statistical information regarding driven route characteristics and certain driving behaviour factors, like percent of the trip driven on motorway, number of crossings, or percent of trip driven with low speed. These characteristics are demonstrated either with map layout or with graphs. Example of GUI demonstrating the trip analysis is shown below, where user has a map view of the trip, where route characteristics, as well as driving behaviour markers are illustrated to the driver. Also, weather during the trip is shown (red rectangle). Moreover, driver is able to comment what caused bad driving occurence.
We are interested in studying other opportunities to deliver fuel-efficiency information to the driver. Therefore, we are planing a user study, where different modalities will be tested to identify bad and good practices.
E. Gilman, Y. Zuo, M. Pyykkönen, S. Pirttikangas & J. Riekki (2016). Delivering eco-driving information to drivers. 11th ITS European Congress, Glasgow, Scotland, 6-9 June 2016.
E. Gilman, A. Keskinarkaus, S. Tamminen, S. Pirttikangas, J. Röning & J. Riekki (2015). Personalised assistance for fuel-efficient driving. Journal of Transportation Research Part C: Emerging Technologies, Volume 58, Part D, September 2015, pp. 681–705.
As the current data sources flowing in the City of Oulu include, for example, magnetic loop data, traffic signalling system data, Wi-Fi/BT access data, traffic cameras, public transport data, parking data, crowdsourced data, spatial databases, road/weather data & predictions, laser range measurements, taxi GPS/OBDII datain, we need robust, scalable systems to handle data management. We have designed different platforms from data ingestion and data provisioning.
One key challenge in Big Data is performing low-latency analysis with real-time data. In vehicle traffic, continuous high speed data streams generate large data volumes. Harnessing new technologies is required to benefit from all the potential this data withholds. We identify seven functional phases from data ingestion to data provisioning and develop analytics platform based on these phases: 1) Ingestion and aggregation, 2) Data pre-processing and fusion, 3) Real-time analysis, 4) Storage and metadata creation, 5) Periodic batch analysis, 6) Archiving, querying and post analysis and 7) Data provisioning. For our system, Cloudera CDH 5 distribution of Hadoop framework is chosen as it provides comprehensive set of components, management interface and API to access management functions. We developed our analytics platform as inspired by the Lambda architecture, considering flexible composition of data ingestion, storage and analytics components.
One solution we tested is depicted in the above figure, Apache Spark Streaming was chosen as the core component for real-time analysis because it is efficient in iterative computing tasks, supports a variety of data sources and programming languages and can be run on Hadoop. Data aggregation is done with Apache Flume in data ingestion pre-stage together with Apache Active MQ broker. This solution provides message routing, aggregation, queuing, load balancing and flexible configuration and integration to other systems. Flume agents ingest high velocity GPS data from taxi cabs and road weather data from weather stations, and forward data to Spark for pre-processing. Data storage is split into two separate tasks: storing the raw data and storing the analysis results. Each data set is stored in a separate HDFS subdirectory. Periodic analysis tasks read the preprocessed and ingested data from the HDFS subdirectories and write results to separate locations. Impala database is chosen since it has good query performance with HDFS. Impala daemons are run on each datanode in Hadoop cluster and Impala query engine retrieves data straight from HDFS directory structure using Parquet storage method. This makes both the raw data and the result data instantly available. We developed Python API to support developers in creating custom Spark analysis scripts, to manage Impala database and HDFS storage, as well as to configure Flume agents and the REST based data provisioning interface. Python Flask based REST interface is implemented for data provisioning which transforms requests to Impala queries and transfers responses to JSON format.
Our future work will cover scalable real-time analysis and decision making for the Internet of Things applications. We will continue our platform evaluation by integrating alternative state-of-the-art technologies to our platform such as Apache Kafka, MQTT and CoAP. Moreover, we will study how semantic technologies can be integrated with our Big Data platform to facilitate interoperability, scalable knowledge discovery and data mining tasks. This work has been already started in our previous research, where we studied distributed reasoning of real-time traffic events.
A.I. Maarala, M. Rautiainen, M. Salmi, S. Pirttikangas & J. Riekki (2015). Low Latency Analytics for Streaming Traffic Data with Apache Spark. Proceedings of the 2015 IEEE International Conference on Big Data, pp. 2855-2858, Santa Clara, CA, USA, October 2015.
A. Maarala, X. Su & J. Riekki (2014). Semantic data provisioning and reasoning for the internet of things. International Conference on the Internet of Things, Oct 2014, pp. 13–18.
S. Pirttikangas, E. Gilman, X. Su, T. Leppänen, A. Keskinarkaus, M. Rautiainen, M. Pyykkönen & J. Riekki (2016). Experiences with Smart City Traffic Pilot. 2016 IEEE International Conference on Big Data (IEEE BigData 2016), Washington DC, 5-8 December.