A defining feature of MOVE_UK Phase 2 was the addition of front radar sensors to all trial vehicles, which extended the vehicles’ sensor perception and allowed experimentation with methods for validating different sensor modalities. The integration of the radar sensors in the vehicles required both hardware and software adaptions. The project’s data recording and analysis tools were also updated and extended to allow easy access to statistics and visualisations of the radar data collected.

The primary focus of this report is set on the four new use cases, which make use of the Phase 2 radar capabilities: Lead Vehicle Statistics (LVS), Radar-based Autonomous Emergency Braking (ARB), Cut-in Scenarios (CIN) and further telematics (Telematics 2). View the latest report >

LVS started from the requirement to understand better how drivers normally follow other vehicles. Conditions of particular interest are the gaps maintained between vehicles, including at what speeds and their duration, as well as what causes the ego driver to stop following the lead vehicle. The intention is that this information can help to understand how to make autonomous driving more comfortable. Emerging clusters in distance-speed distribution can be used to deduce driving behaviour regarding the gaps left between cars and what is common for most drivers. On the other hand, outliers in the point plot can inform about extreme cases that have to be anticipated.

The LVS use case serves as a proof of concept for using existing continuous 1 Hz collected CAN data for statistical evaluation rather unlike the CIN use case, where a trigger was designed and the collection of a new stream of dense CAN data was collected for a limited time at higher frequency (10 Hz).

The Cut-In scenario use case implemented in the MOVE_UK vehicles is used as an example to show the benefits of real-world event data. The use case is used to help understand driver behaviour (of both driver and surrounding drivers) in the particular situation of a Cut-In during a traffic jam on the motorway. A number of behaviours can be associated with a Cut-In situation, including but not limited to, drivers trying to cut across lanes in order to exit a motorway at a junction, changing lanes to avoid a long queue, or to get past a slow-moving vehicle. Understanding driver behaviour in these situations can benefit other Consortium projects by helping them define parameters for safe and human-like automated driving functions.

Due to time and performance considerations, the trigger conditions in the vehicles were kept simple. A first step in refinement is to consider which of the Cut-In sequences collected can be classified as real Cut-Ins and which represent false positives. At the moment, drawing any firm conclusion regarding the parameters for a safe Cut-In, or even to say which event is a true Cut-In or a cut-out, would be premature. Collection of such events and using big data analysis techniques to process, analyse and eliminate any false positives to ascertain true Cut-In events within the dataset is underway and results will be available in Phase 3 of the project.

As with the video AEB use case, the purpose of the Subcritical radar-based AEB use case is to develop, trial and demonstrate the capabilities required to perform silent connected validation for ADAS or ADS systems. In order to use synergies from Phase 1, the video AEB use case was modified appropriately to be applied to radar.

The parameters for the ARB function of the radar sensors are set to a more sensitive level than used in production in order to detect more situations and collect more data. Within the Subcritical ARB use case, 58 events were captured during the course of a nine-month period within Phase 2 of MOVE_UK.

With this ARB use case collecting subcritical ARB situations the following capabilities were successfully demonstrated: identifying real-world false positive situations; automated capturing of events which are activated by the system within the vehicle (on board); transmitting high volumes of data over the air; and the process of using real word data to re-simulate in order to optimise parameters that then can be re-flashed to the control unit in the field for the next iteration. In reaching our goal for these capabilities, methodologies have now been generated and an example infrastructure to validate actual and future ADS systems has been developed.

The analysis of the situations showed that the driver reaction time parameter for the trigger was set to a very subcritical level, resulting in many non-dynamic sequences with less value for the analysis. Therefore, a new subcritical parameter set closer to production setting was found through resimulation of these sequences. In order to still gather a satisfactory amount of data, the parameter set is kept at a level of sub-criticality which will permit some sequences with low dynamics. A further, final adjustment of the parameter settings is planned after a sufficient amount of sequences is gathered. The finalsetting will be very close to production levels and only near critical situations will be captured. The collection of radar based subcritical ARB sequences will continue during Phase 3 of MOVE_UK.

Telematics data continues to be gathered alongside the core CAN Bus data from the test vehicles in Phase 2. This data gathering is geared to still support a number of key aims:

1. Comparison to existing risk understanding to help support development of improved risk estimation.

2. Comparison to existing event understanding to help support development of improved event data recorder (EDR) functionality that works alongside advanced vehicle technology.

3. To act as an independent data gathering to support statistical analysis for captured vs missed data.

In Phase 2, data gathering is checked to investigate geospatial bias, so as to ensure that meaningful analysis can be performed. This analysis helps to ensure geographically representative data, thereby enabling a meaningful statistical analysis of behaviour by locations. Study of the Greenwich test region (apart from the fleet base of operation which has disproportionately high test vehicle traffic) highlights a smooth and fairly even coverage of test vehicle traffic volume over the borough. The gathering region has been extended to enable capture of data in differing areas, thereby widening it to differing road types, out of the main urban testbed.

The ultimate aim of the telematics use case is to construct a method through which a comprehensive and synoptic description of risk can be computed. To achieve this, the MOVE_UK team is seeking to combine knowledge of telemetry, driver behaviour, the conduct of surrounding vehicles and geographical factors into a single parameterisation. However, to obtain such an understanding will require continued studies to obtain more data using a greater number of vehicles, ideally with the ability to distinguish between specific drivers. However, data collected during Phase 2 can be used to reach a good understanding of vehicles-in-front for incorporating into models of risk. Knowledge of the traffic context in which the vehicle is driving is clearly important, letting us observe not just how the driver is driving, but whether the observed behaviour is normal. 

View the latest report >

Register for our newsletter

Learn more about TRL with regular updates

Media membership

Our latest press releases and news

Get in touch

Have a question? Speak to one of our experts today.