The last installment of this series illustrated how open-source software can amplify the benefits of autonomous driving technology. This installment focuses on operational design domains (ODDs) – specifications of the conditions and environments under which autonomous driving software is intended to operate.

In the context of autonomous driving development, operational design domain is a relatively common term. At TIER IV, we use three categories to guide the development of autonomous driving systems: ODD1, ODD2, and ODD3.

• ODD1: Road environments dedicated exclusively to autonomous vehicles.
• ODD2: Roads managed by an authority where rules are enforced to prevent incidents like sudden pedestrian crossings or vehicle intrusions.
• ODD3: Public roads where people and vehicles share the space.

In addition to these three types, a future classification might include vehicle-only roads where people are not permitted. Categorizing ODDs conceptually is extremely useful for aligning the development and operational frameworks of autonomous driving software.

If you’re a driver, you probably have a general sense of the roads and environments where you can drive. This might include the route from your home to your workplace, the shopping mall, or the way to a friend’s house. This mental map of where you can drive is the essence of an ODD. It’s important to note that TIER IV’s ODD categorizations do not represent a definition of specific ODDs. The definition of an ODD in the context of autonomous driving is quite complex.

Even on familiar roads, factors such as weather can influence your decision to drive. For instance, foggy conditions might prompt you to postpone your trip, or you might wait for snow to clear before heading out. The drivability of a road can be affected by such environmental factors. Additionally, even if two roads have the same width and curvature, the type of vehicle you’re in – whether it’s a passenger car or a large truck – will affect which roads are navigable.

The skill level of the person behind the wheel, whether a beginner or an experienced driver, also influences the range of drivable roads. Things like lane restrictions due to roadworks, road closures due to accidents, and other factors must also be considered. These are all part of the concept of the operational design domain, which encompasses various elements of the road environment.

Autonomous vehicles drive autonomously within a defined ODD. They have to stop autonomous operations if they move outside the ODD. However, with Level 4 autonomous driving systems there is no need for a human to be in the driving seat, so it's not possible to simply switch control from an autonomous driving system to a human driver if the vehicle exits the ODD.

Conditions like heavy rain or fog could push an autonomous vehicle outside its operational design domain. How to handle situations when a vehicle veers outside an ODD presents significant challenges for autonomous driving systems.

It can be unsafe for a vehicle to stop abruptly on the road in an emergency. It might be necessary to pull over to the shoulder, for example. In cases where there is no shoulder, it might be necessary to reduce speed and drive to the nearest parking area. On the other hand, if an accident occurs right in front of a vehicle, an immediate stop might be the only safe option. When an autonomous vehicle exits an ODD, different responses are required to balance risk and safety depending on the road and traffic conditions.

In simpler words, the ODD refers to the area within which an autonomous driving system is designed to operate. Autonomous vehicles must determine whether they are within or outside their designated ODD to decide whether to continue operating autonomously. Therefore, it is essential to define the boundaries of the ODD, which depends not only on the road environment but also on the vehicle's handling capabilities and the autonomous driving system's performance.

In the case of Level 4 autonomous driving, where no driver is available to take over if a vehicle moves outside its operational design domain, defining the ODD becomes even more challenging. This makes it crucial to have strategies for managing such situations. In the next installment, we’ll explore minimum risk maneuvers (MRM), which involve taking actions as safely as possible to mitigate risks when an autonomous vehicle leaves its ODD.

Toshihide Ando | TIER IV Fellow
Toshihide joined TIER IV in 2019, where he has served in roles including vice president of technology before assuming his current position as fellow. Previous experience includes software development and R&D at a major Japanese automotive parts manufacturer.