Partnership aims to make autonomous driving more realistic
Simulation software provider dSPACE is working together cogniBIT to integrate its AI-based driver model, to integrate the unpredictable behaviour of human drivers into simulation. This will help to make traffic scenarios for the development and testing of autonomous vehicles even more realistic.
In the first stage, dSPACE is pairing its simulation environment Automotive Simulation Models (ASM) with driveBOT, cogniBIT’s AI-based driver model. In the future, behaviour models from further road users, such as pedestrians or motorcyclists, will be successively integrated into simulations.
Dr Lukas Brostek, co-founder and CEO of cogniBIT comments: ‘With its ASM simulation environment, dSPACE provides a powerful and widely-used tool in the industry for developing functions for automated and autonomous driving. We at cogniBIT are pleased to be able to complement ASM's already-high fidelity in areas such as vehicle dynamics, sensor and environment modelling with a valid driver and road user model.’
Traffic situations are influenced by the behaviour of individual road users. This includes emotions such as fear, surprise, and happiness but also limitations such as the road users' impaired view. Highly automated and autonomous driving will become safer in practice only once all of these aspects are realistically integrated into the simulation for the development and testing of functions for autonomous driving. driveBOT, the AI-based driver model from cogniBIT, lets users replicate human-like behaviour in the simulation and reproduce realistic traffic scenarios.
Realistic driver models play an important part in ADAS/AD development because, in the first stage, the functions for autonomous driving are designed for the operational design domains (ODD). The automated-driving systems have to function in these defined fields of operation. The system will request the driver to take control if the vehicle leaves this field.
To make the assistance systems more realistic in the next stage, further factors can be included using cogniBIT’s AI-based driver models, for example, a nervous new driver or a stressed driver behind the wheel.
The AI-based driver models from cogniBIT can be used in the ASM simulation environment to simulate more realistic surrounding traffic, not only for the ego-vehicle but also in multi-agent simulation for the surrounding traffic participants (fellows). The realistic movement of the fellows allows simulated traffic scenarios to be varied efficiently, to identify corner cases, and therefore, to define the limits of a driving function.
Dr of Engineering Christopher Wiegand, strategic product manager of the dSPACE Automated Driving and Software Solutions business unit added: ‘Driving function for SAE Levels 2 - 5 must be carefully validated with regard to safe interaction with human road users. The involvement of neuroscientists and sensorimotor findings into simulation leads to valid and meaningful simulation results and will make driving functions safer.’