A detailed explanation of the operational design domain (ODD) of autonomous driving

On April 30, 2021, SAE released the fourth edition of J3016 "Driving Automation Classification", which is another iterative update of J3016 after January 16, 2014, September 30, 2016, and June 15, 2018.

In July 2021, ISO released ISO 22737 "Performance requirements, system requirements and performance test specifications for low-speed automated driving (LSAD) on a predetermined route".

This series will provide a detailed interpretation of autonomous driving ODD, DDT, DDT fallback, OEDR, low-speed autonomous driving and autonomous driving classification.

Text | Wu Dongsheng

basic concept

Autonomous driving needs to work correctly in every country and region on Earth. Therefore, it is required to take into account various types of roads, road markings, traffic signs, etc., as well as environmental issues such as weather conditions. Ensuring that the autonomous vehicle performs adequately in its intended operating environment is a key part of the verification of the entire autonomous driving system. The original definition of the term Operational Design Domain (ODD) comes from the definition in SAE J3016.

SAE J3016 defines ODD as "the conditions under which a particular driving automation system or its functions are designed to operate, including but not limited to environmental, geographic and temporal constraints, and/or the presence or absence of certain traffic or roadway characteristics."[1]

Simply put, ODD is to define the conditions under which autonomous driving is possible. Without these conditions, autonomous driving cannot be guaranteed to work. Any autonomous driving vehicle must have certain limited operating conditions. These conditions can be broad or precise, and determine what scenarios the autonomous driving vehicle can handle.

For example, a car's autonomous driving system can only be used on highways. It can automatically maintain lanes, automatically overtake, automatically follow cars, automatically give way, automatically pass ETC, automatically enter and exit ramps, etc., but it cannot be fully autonomous in the city. At the same time, to ensure that the autonomous driving test and verification is complete, at least all aspects of the ODD must be ensured to ensure the safe operation of the system, or by ensuring that the system can identify beyond the scope of the ODD.

In the recommended national standard "GBT Automobile Driving Automation Classification" issued by the Ministry of Industry and Information Technology, ODD refers to the vehicle status and external environment of the driving automation function determined during design. The operational design condition (ODC) is the condition for the normal operation of the driving automation function determined during the design operation, including ODD, driver status and other necessary conditions. [2]

Dynamic Driving Task (DDT) is all real-time operational and strategic functions required to operate the vehicle in road traffic, excluding strategic functions such as trip scheduling and destination and waypoint selection. Dynamic Driving Task fallback (DDT fallback) is the user response to the implementation of DDT or the achievement of a minimum risk condition under the same circumstances (1) after a DDT performance-related system failure, or (2) when ODD is exited, or the ADS response to the achievement of a minimum risk condition.

Object and Event Detection and Response (OEDR) is a subtask of DDT that involves monitoring the driving environment (detecting, identifying, and classifying objects and events, and preparing to respond as needed) and executing appropriate responses to those objects and events (i.e., completing DDT and/or DDT takeover as needed). Low-speed Automated Driving (LSAD) is an autonomous driving system with a maximum speed of 8.89 m/s (32 km/h) and will be used for last-mile transportation, commercial area transportation, commercial or airport, port, university campus, and other low-speed environments. [3]

Levels or Categories of Driving Automation are based on the driving automation system functionality and are determined by the division of roles in the execution of DDT and DDT fallback between the function and the (human) user (if any). The manufacturer of the driving automation system functional components determines the requirements, ODD and operating characteristics of the function, including the level of driving automation.

ODD Case

(1) LSAD ODD defined in ISO 22737

Each LSAD system shall have its ODD defined by the manufacturer. A LSAD ODD-limited system shall specify at least the following attributes:

a) Low speed: The speed of the LSAD system should be equal to or less than 8.89 m/s or 32 km/h;

b) Scope of application: for example, restricted access or dedicated roads (public or private), or pedestrian/cycle lanes, or areas where access is restricted to all or certain classes of motor vehicles. Restricted access roads may be designated by lane markings or speed limits or physical demarcations;

c) Predefined routes: Routes defined within the LSAD system before the LSAD system is operational. The LSAD system can only operate on predefined routes. The predefined routes should be jointly determined by the relevant stakeholders (e.g. local authorities, service providers, manufacturers, etc.). The dispatcher should confirm that any deviation from the predefined routes will not result in hazardous situations;

d) lighting conditions in the application area;

e) weather conditions;

f) road conditions;

g) the presence or absence of VRU;

h) There may be static obstacles within the drivable area;

i) Internet connection requirements, etc.

The LSAD system or the dispatcher shall select an operational value for the specified application (for the vehicle being driven by the LSAD system) within the range of predetermined values ​​of the ODD attributes based on the current ODD conditions (e.g., foggy conditions, nighttime lighting conditions). For example, the dispatcher or LSAD system may decide to limit the maximum permissible speed on a rainy day to a lower speed than on a sunny day.

(2) L3 ODD of Model B [4]

a) Geographic restrictions: Geographic ODD includes the boundaries of the geo-fence and all applicable traffic rules for the area within these boundaries. It can only be activated and operated on limited highways with continuous structural separation from oncoming vehicles, without intersections and roundabouts. This road type is characterized by a low probability of pedestrians and bicycles. Compliance with these activation conditions will be detected by on-board sensors, such as cameras monitoring traffic signs, and will also be guaranteed by information provided by high-definition maps. Taking into account various factors such as legal requirements, Level 3 ADS is designed to operate at speeds of 0 to 85mph (136.79 km/h);

b) Environmental constraints: The autonomous vehicle continuously monitors the environmental ODD to ensure safe operation under all conditions. For example, the vehicle’s speed will be adjusted based on the time of day, light conditions, if the road surface has too little friction (e.g. snow or ice on the road), or if the wind is too strong;

c) Limitations of human drivers: The driver must wear a seat belt, sit properly, and stay awake, etc.

(3) L3 ODD of Model A

a) Driving on an expressway or a two-lane or more motorway with a central divider and guardrails;

b) The distance between the vehicle in the lane and the surrounding lanes is relatively close, that is, there is a traffic jam;

c) The vehicle’s speed does not exceed 60km/h;

d) There are no traffic lights, pedestrians, cyclists, etc. within the sensor's detectable range.

NHTSA ODD Classification Framework

NHTSA uses six elements to construct the design operation domain in "A Framework for Automated Driving System Testable Cases and Scenarios", including infrastructure, driving operation restrictions, surrounding objects, interconnection, environmental conditions, and area. [5]

(1) Infrastructure Road types: divided highway, undivided highway, arterial, urban, rural, parking lot, multi-lane, single lane, high occupancy vehicle (HOV) lane, entrance/exit ramp, emergency evacuation route, one-way, turn lane, private road, two-way lane, intersection (signal, U-turn, four-way/two-way parking, roundabout, merge lane, turn lane, pedestrian crossing, toll plaza, railroad crossing) (FHWA, 2012). Road surface: asphalt, concrete, composite, grating, brick, dirt, gravel, scraped road, partial blockage, speed bump, pothole, grass (Gibbons, 1999). Road edge: marking line, temporary marking line, shoulder (paved or gravel), shoulder (grass), concrete guardrail, grating, railing, curb, cone (Sage, 2016). Road geometry: straight line, curve, hill, side peak, corner (regular, blind corner), negative obstacle, lane width (Huang, 2010).

(2) Driving Operation Limits Speed ​​limits: minimum and maximum speed limits (absolute, relative to speed limit, relative to surrounding traffic) (Elpern Waxman, 2016). Traffic conditions: minimum traffic volume, normal traffic volume, bumper-to-bumper/rush hour traffic volume, changing traffic conditions (accidents, emergency vehicles, construction, road closures, special events) (University of California Road Program, 2016).

(3) Surrounding objects Signage: Signs (e.g., stop, yield, pedestrian, railroad, school zone, etc.), traffic signals (flashing, school zone, fire department zone, etc.), crosswalks, railroad crossings, stopped buses, construction signs, emergency signals, distress signals, road user signals, hand signals (FHWA, 2012). Road users: Vehicle types (passenger cars, light trucks, large trucks, buses, motorcycles, wide loads, emergency vehicles, construction equipment, carriages/carriages), parked vehicles, moving vehicles (manual, automatic), pedestrians, bicyclists (CA DMV, 2016). Non-road user obstacles/objects: Animals (e.g., dogs, deer, etc.), shopping carts, debris (e.g., tire debris, trash, ladders), construction equipment, pedestrians, bicyclists.