MANASWINI is Associate Vice President, ADAS and Autonomous Vehicle Group at KPIT
As the industry moves towards deployment of autonomous driving, challenges are emerging in all areas. The development of features to achieve Level 4 and Level 5 autonomy is the most complex area under exploration today. Artificial intelligence and data-driven modelling are among the key areas considered as solutions in feature development. As the feature development progresses, the next big challenge would be the validation of the same. Readers would recall our previous article on verification validation (V&V) in the August 2017 edition of Auto Tech Review.
The feature needs to be validated against the following:
:: Requirement and specification as a software element;
:: Sensor response for different environments; and
:: And, most importantly, the variation in sensor performance and variation in environments.
The most complex variable that autonomous driving undergoes is the variability of the environment and its impact on features. Therefore, it is critical that the autonomous features and systems, which includes sensors, domain controller and actuators, are validated against maximum variability.
One of the possibilities in validation is virtual environment simulation. The virtual environment simulation can be used for validation against requirements, sensor fidelity, platform performance, and environmental conditions.
The other important consideration in feature development is to be on the road and be compliant against regulatory requirements. The challenge lies in understanding the requirement from regulation, and planning techniques to assure coverage of validation against mileage and safety requirements. Virtual environment simulation enables in meeting the regulatory aspects.
In this article, we discuss the key concepts involved in virtual environment simulation and its development.
KEY CONCEPTS IN VIRTUAL ENVIRONMENT SIMULATION
The major concepts involved in virtual environment simulation are still evolving. As of today, many research and exploration are taking place to put in the complete end-to-end solution. Some of the concepts are being explored at present and developments are going on, as represented in (1).
Every concept described above is exclusive, and plays an imperative role in developing the solution:
a. Scenarios are the first step towards creating real road environment in virtual conditions. This means building catalogues that have millions of scenarios. A library of scenarios is developed considering variations such as the road type, radiant, curvature, junctions, weather conditions, and other parameters. Other modelling systems such as 3D modelling and creating digital twins are getting into reality for creating a virtual environment in digital forms. While vehicle dynamic models have matured, the maturity of sensor models is evolving.
b. The fidelity of the sensor model is the key challenge. Data-driven modelling is one of the areas that might bring maturity to sensor models.
c. Artificial intelligence and machine learning are applied to deal with data collected from real sensors.
d. Automation is going to be the next big area of investment and exploration. Similarly, management of the complete end-to-end simulation with automation seems to be the brain behind virtual environment simulation.
This area is just being conceptualised, and any innovation is going to be the next big thing in autonomous driving.
AREAS INVOLVED IN VIRTUAL ENVIRONMENT SIMULATION
The most important areas involved in virtual environment simulations are scenario library, modelling, simulation management, and intelligent reporting, as represented in (2). The areas and concepts overlap a little but provide a broader view.
Concepts like digital twins, object modelling, 3D modelling, etc. are being explored by several organisations. They help in creating a virtual environment as close to the real world. The most important aspect of validation is how close one can make the virtual environment compared to the real-world environment. At the same time, virtual simulation involves hardware, software, and interfaces. Management of the whole end-to-end simulation is a new area that needs an effective solution. Intelligent reporting is another area under exploration.
The core architecture and management of simulation will be a major area of focus in the coming days. As simulation involves humongous data and infrastructure, both at the hardware and application level, the management of its architecture needs innovation and unique solutioning. Machine learning, artificial intelligence, Internet of Things (IoT), and many such technologies need to be integrated to provide a robust and efficient solution for simulation management.
Other than the areas in the software domain and hardware infrastructure, specifying tools and architecture of the concepts involved in design are pertinent to virtual environment simulation. There are individual elements like tools and infrastructure available but creating a complete solution would be the most interesting to focus. Once we have a holistic solution, the virtual environment simulation would provide real evidence of coverage of validation for both requirements and regulatory assessments.
Virtual simulation environment is the way forward and we are excited to work towards building software for fully autonomous vehicles.