Progress on reducing road fatalities in the European Union (EU) has stalled in recent years (at some 25,000 yearly fatalities). While the underlying trend remains downward, progress has slowed down in most countries since 2013 and, at this rate, the EU’s target of halving the number of road deaths by 2030, relative to the 2010 baseline of around 30,000 fatalities, will not be met. Current approaches to the development of safety systems, reactive to evidence on the road, will not be suficient to overcome this situation.
Analysis of road accident statistics suggests that more than 90 per cent of all road collisions are caused by human error, such as distraction, fatigue or infractions1. That is why several leading academic and industrial partners have joined forces to develop a novel proactive approach, enabling the identication of potential safety threats as a basis to develop future safety concepts required in novel emerging mobility scenarios, particularly automation. SAFE-UP, the latest European flagship initiative on road safety, is focused on delivering the promised potential of new automated mobility to keep all road users safe – drivers, vehicle occupants and Vulnerable Road Users (VRUs), including pedestrians, cyclists and users of two-wheelers.
To understand how autonomous vehicles (AVs) will change the situation on our roads, how they will interact with other road users and how safety-critical these interactions will be, it is vital to apply reliable tools to simulate the road transport system in future conditions. SAFE-UP will enhance current state-of-the-art traffic simulation tools, including a collaboration between the transport modelling specialists Aimsun and the automotive firm Applus+ IDIADA, which specialises in automotive engineering services.
The automated vehicle model integrated in the Aimsun Next traffic simulation platform is developed by Applus+ IDIADA. Built from the company’s CAVRide demonstrator, Applus+ IDIADA has aligned its development and validation expertise to engineer a Level 4 autonomous car that meets the necessary functional requirements to safely operate whilst keeping overall car performance. Several so ware components of CAVRide make up the automated vehicle model, meaning that the real vehicle so ware is the same as the one that runs on the simulator. This has great advantages, because the simulation reliably corresponds to reality, while the software that goes on-board the vehicle is being tested in complex scenarios, which would be very expensive to reproduce in reality.
The main components that form the AV model are: prediction, where the vehicle interprets the traffic environment and predicts other vehicles’ actions; planning and decision making, involving behavioural and trajectory planning modules predicting the vehicles’ direction; and high-level control, which is an algorithm that generates effective commands to move the vehicle according to the desired trajectory.
Because the objective of the simulation is to analyse traffic behaviour, there is no need to worry about vehicle dynamics. The sensing devices – such as LIDAR, radar and cameras – are replaced by filters that limit or modify the true characteristics of the road users that are being simulated. For example, rain degrades the automated vehicle’s ability to sense its surroundings, and this can be simulated by decreasing the range in which the simulation objects are reported to the automated vehicle model. In turn, the model, having a restricted view of the environment, could make different decisions than in favourable environmental conditions, where it would have more information about the environment.
The simulation and prediction of future safety-critical scenarios requires the development of a new modelling framework. Until now, driving behaviour of vehicles and riding behaviour of VRUs has been idealised, and simulation of accidents is bounded by the model’s definition. Since road accidents are extremely rare among the high number of trips made on our roads, hard evidence for behavioural mechanisms in accident or near-miss conditions has been difficult to observe.
This is where driving behavioural models are needed to bridge the gap between the fields of traffic flow theory, simulation, traffic safety and psychology. To do this, three types of models are being developed: the first is an advanced human-driven car; the second is automated driving behaviour with 2D-trajectory planning controlling longitudinal and lateral movements; and the third is behavioural models for VRUs, such as cyclists, pedestrians and powered two-wheelers (PTWs).
This new generation of behavioural models together with new safety metrics is systematically integrated in the Aimsun Next traffic simulation platform. It enables harmonised simulation of the next generation of all road users’ behavioural models (driver, AV and VRU), capturing the failure of sensors, risks those drivers and riders might take, their errors of judgement and distracted perception in future mixed traffic conditions.
Aimsun will model and simulate future safety scenarios to set the baseline for safety-critical situations in the defined use cases. The modelling process will build a digital twin of the infrastructure environment, traffic conditions, technology level, penetration rates and behavioural parameters, covering potential crash locations in urban and interurban roads.
Besides reliable simulation tools to assess new safety-critical situations, we believe that an effective introduction of AVs on our roads will only happen with a holistic approach, coupling the technological benefits of the vehicles and simulation with the requirements of infrastructure and policies, together with the definition of new business models. Led by strategy consultancy Bax & Company, another member of the initiative, SAFE-UP will publish before the end of 2021 a new roadmap that presents its view to achieve this.
AVs will change the way that we travel in an unprecedented manner, touching upon road safety, traffic efficiency, reduced emissions, reduced travel costs and more comfort. We expect most of the implications of autonomous driving to be very positive, dramatically reducing the number of road fatalities and incidents. With a potential impact on 64 per cent of all road fatalities (urban, peri-urban and highway settings), SAFE-UP will effectively contribute to reach the EU’s Vision Zero goal for 2050, when road fatalities should be completely avoided.