Simulated driving can reduce development costs and improve time to market for AVs.
Simulated driving can reduce development costs and improve time to market for AVs.
( Source: Public Domain / Pexels)

SIMULATION The role of simulation in autonomous driving

Author / Editor: Jamie Thomson / Florian Richert

One of the biggest benefits promised by autonomous driving is improved road safety. However, in order to ensure that self-driving cars become safer than their human-driven counterparts, extensive testing is required.

According to some reports, autonomous vehicles will need to drive 8 billion miles to reach a human-level error rate, whereas others estimate the figure being closer to 11 billion. By comparison, Google’s AV project Waymo, had only completed 20 million miles of testing as of the beginning of 2020.
In any case, at this level, it would take automotive manufacturers hundreds of years of physical testing to arrive at the level of safety currently achievable by human drivers. This challenge alone, is significant and if autonomous vehicles are to replace human-driven cars, the physical testing process needs to be much quicker.

This is where simulation can help. Driving simulators are already used to train drivers across the world. In many countries, car driving assessments include a simulated test to ensure that road users have the cognitive skills to drive safely on public roads. In the development of autonomous vehicles, simulation can be used in a similar way, to help vehicles reach the required number of testing miles.

Let’s take a closer look at the role of simulation in autonomous driving:

The benefits of simulated road testing for AVs

Simulation is a vital component in the design, testing and validation of autonomous vehicles. Because simulation happens in a virtual environment, it makes the development of AVs quicker, less expensive and it provides more insights into underlying mechanics of a self-driving vehicle.
Given the huge number of miles AVs need to drive to achieve a human-level error rate, simulation is the only practical way of monitoring and analysing an AVs performance within a realistic timescale. With simulation, automotive makers can improve the overall safety and quality of their vehicles.
Self-driving cars need to be compatible with many difference technologies, such as LIDAR systems and lithium-ion batteries and simulation provides manufacturers with opportunities to prototype designs that will work. Simulation can also overcome other challenges associated with autonomous driving, such as testing sensors, cameras and actuators and driving control software. As such, simulation can reduce manufacturing costs in the long-terms and accelerate time to market.

How simulated driving technology works

Historically, simulators were based on the architecture used in the aircraft industry. Although they were suitable for recreating slower aircraft dynamics, their response was too limited for road vehicles. This led to the development of driver-in-the-loop simulators that offered increased dynamic performance.
Motion cueing algorithms are used in a driving simulator to provide cues that make the simulation feel as close to real-time driving as possible. The model is used to predict the response of the vehicle in various driving situations and the cueing algorithm is used to create the driver sensation.

One of the leading simulation technologies on the market is NVIDIA. Their DRIVE Sim software and DRIVE Constellation simulator delivers a scalable testing environment. Its platforms are currently used by over 400 companies in the autonomous driving space and the platforms are capable of generating billions of miles of vehicle testing. Using a hardware-in-the-Loop platform, environment data is processed using cameras, radars, lidars and ultrasonics.

How simulation is meeting the AV challenge

Because simulation can speed up the AV testing process and thus increase time to market, a number of companies are using the approach to become leaders in the field. For example, Audi’s (now-abandoned) Traffic Jam Pilot was initially designed for use in it’s new A8 model.
When stuck in traffic, drivers could go ‘hands-free’ and let the car handle acceleration and braking, essentially achieving Level 3 autonomy. The vehicle made use of several sensors to detect other cars and objects as well as for lane detection and for anticipating driving conditions.
Automaker GM is currently simulated driving to test its Cruise Automation Vehicle by driving 200,000 miles per day using 3D simulators. By the time it goes into production, the Origin model is expected to perform as well as an average human driver in terms of safety. GM plans to continue testing its software in real-world scenarios on the streets of San Francisco soon.

Closing thoughts

Simulated driving will come to play an ever-increasing role in the development and testing of autonomous vehicles. It offers the opportunity to negate the need for expensive prototypes in the design phase and can accelerate the time to market by providing accurate feedback on safety. As an enabling technology, simulation will play an essential role in delivering the benefits of autonomous vehicles in an affordable, safe and timely manner.