The League of American Cyclists and the company that makes self-driving cars collaborated to come up with guidelines for how self-driving vehicles should interact with cyclists. As the self-driving industry moves away from testing and toward commercialization, the goal is to set a standard for other companies in the industry to follow.
According to the World Health Organization, 41,000 cyclists are killed in road traffic-related incidents every year. Good coding at the beginning of self-driving vehicles is expected to reduce safety. Argo's guidelines emphasize training its models in a way that specifically notes cyclists, cycling infrastructure and cycling laws, and self-driving cars learn from massive databases that categorize and identify objects.
The creation of these guidelines is part of the dedication to building trust with community members and developing a self-driving system that provides a level of comfort to cyclists. We encourage other car developers to adopt them as well.
The company collaborated with the community to hear about cyclist behaviors and interactions with vehicles. The technical guidelines for self-driving systems were created by LAB and Argo.
In California, the public can now ride in the vehicles of Argo AI.
Cyclists are an object class.
Defining cyclists as a distinct class will create a diverse set of bicycle imagery for a self-driving system to learn from. Systems should be trained on images of cyclists from a variety of positions. The system will be able to account for the different shapes and sizes of bikes.
According to a statement from Argo, a self-driving system should designate cyclists as a core object representation within its perception system in order to detect cyclists accurately.
Cyclists behavior should be expected.
Cyclists can be very unpredictable. They might lane split, walk their steed, make quick movements to avoid obstacles on the road, yield at stop signs, hop off the sidewalk and into the street. A good self-driving system should be able to predict their intentions and be prepared to react.
When a self-driving vehicle encounters a cyclist, it should use specialized, cyclist-specific motion forecasting models that account for a variety of cyclist behaviors, so that it can better predict and respond to the cyclist.
Map cycling infrastructure and local laws.
High-definition 3D maps are often used by self-driving systems. Local and state cycling laws should be part of the environment. It will help the self-driving system to anticipate cyclists' movements, like merging into traffic to avoid parked cars blocking the bike lane or running red lights if there's no traffic.
The system should act in a way that is understandable and safe.
The intentions of the self-driving technology should be understood by cyclists, which includes things like using turn signals and adjusting vehicle position while still in one lane if preparing to pass, merge or turn.
The system should only pass a cyclist when it can maintain its margins and speeds for the entire maneuver, and only if the local speed limits are in line with the system's conservative and appropriate speeds.
Cyclists should be given a wide berth in case they fall, so that the self-driving system can stop or swerve.
Slow down when there are uncertain situations.
Uncertainty in a cyclist's intent, direction and speed should be accounted for by self-driving systems. The company suggested that the vehicle be trained to slow down in that situation if the cyclist traveled in the opposite direction but in the same lane.
When possible, the self-driving system should lower the vehicle's speed and give more space to the cyclist. Slowing down speeds when the system is uncertain is standard practice in the developer world.
Continue to test cycling scenarios.
Keeping testing them is the best way to make the safety case. Both virtual and physical testing should be done for cyclists.
Simulation, resimulation, and play forward are three main test methodologies that should be used in a virtual testing program. These scenarios should capture both the differing vehicle and cyclist behavior as well as changes in social context.
Testing on closed courses and public roads allows developers to verify simulation and ensure tech behaves the same in the real world as it did in virtual. Developers should test their AVs on likely scenarios and edge cases. Testing on multiple public roads in many cities to give the system a diverse set of urban environments to learn from can generate both rare and common cases.
Pursuing public acceptance and safety.
Social acceptance is one of the main obstacles to bringing more vehicles to the roads. Morning Consult found that nearly half of those surveyed said that cars driven by humans are more safe than those driven by audiovisuals.
Making a vehicle safe for all road users is only part of the battle. Standardizing safety practices across the industry might be one way to make people believe their vehicles are safe.
