Simulation will remain pivotal for self-driving programmes

An autonomous vehicle (AV) cannot be left to its own devices without learning from real-world scenarios. However, modern simulation is becoming so accurate that much of the heavy lifting can now be carried out in a virtual environment. Tests that might otherwise be challenging to perform can be completed countless times over and in a fraction of the time.

Simulation forms part of a wider AV development programme which will also include the use of private test tracks and carefully managed driving on public roads. Any test that involves a real vehicle with humans in the loop takes time and money to plan and execute. It also carries the risk of something going wrong, perhaps with damaging consequences. These programmes are unavoidable in the pursuit of a ‘safe’ AV, but also to understand the practicalities of running an AV business.

Simulation does not carry any of those same risks, and while the necessary computing power and software is not cheap, it is still far more cost-efficient. Millions of miles of driving can be racked up in a short space of time. This is all