SP 1Sensing the environment ­­and situational understanding

Mostly, previous developments used for semi-automated and highly automated driving functions directed their focus on motorways and other dual carriageways with lanes in both directions divided by barriers. The complexity and number of typical vehicle manoeuvres remain low in these scenarios. The explicit interaction between road users can often be represented in the form of rule-based behavioural strategies. When it comes to sensor-based detection of the infrastructure on the road and of other road users, for example by camera, radar and LiDAR systems, a technologically advanced level has now been reached for motorway scenarios. In contrast, sensor-based detection of traffic situations still needs to overcome major challenges if it is to be used for automated driving in an urban environment.

The figure shows an example of an urban intersection and highlights the diversity of, for example, road users, traffic light infrastructure, arrows indicating directions, and parking cars. In comparison to the well structured environment found on motorways, the road rules that must be observed and the interaction with other road users is considerably more complex.

The variability of what are appearing forms of traffic guidance that have evolved over time is very distinct in urban areas, and the conditions of traffic infrastructures are not always good. Moreover, the diversity of road users is much higher in comparison to motorways, and the objectives and intentions the road users are pursuing are manifold. This is the reason why their behaviour exhibits a very broad spectrum. Predicting the future behaviour of road users therefore requires a higher level of detail in the way the situational context, along with the corresponding constraints, are sensed.

In order to be able to generate anticipatory and cooperative driving behaviour by the automated vehicle in interaction with other road users – such as non-automated vehicles, pedestrians and cyclists – the automated vehicle must be able to predict what other road users will do within the 

next few seconds and evaluate the reciprocal effects that the vehicle’s own actions will have on other road users, and vice versa. This prediction requires that more details of the situational context, along with the corresponding constraints, be sensed than is possible using current technology.

The automated vehicle must therefore identify the way traffic is guided, which in scenarios occurring at intersections, in particular, cannot be resolved using current sensor-based technology – which makes supplementing it with precise digital map information necessary. The system must interact with other road users on the basis of the traffic rules and, in particular, the applicable rules governing the right of way regulations. The level of detail provided by the current object models and the modelling of static environment are still not sufficient. Inhomogeneous road surfaces and road boundaries, in particular, make determining

the navigable route using data provided exclusively by sensors much more complicated. Again sensing the environment must also be supplemented using digital map information validated by sensors.