Autonomous driving technology company Argo AI has partnered with the League of American Cyclists (LAB) advocacy group to develop guidelines on how autonomous vehicles should identify and interact with cyclists. The goal is to set a standard for other AV companies in the industry to follow, especially as the autonomous driving industry moves from testing to commercialization and will become more mainstream in the years to come.

The World Health Organization estimates that 41,000 cyclists are killed each year in road traffic incidents. While autonomous vehicles are expected to significantly reduce collisions, much of this anticipated safety is the result of good coding up front. Self-driving cars learn from huge amounts of databases that categorize and identify objects and situations that might arise, and Argo’s guidelines emphasize training its models in a way that specifically notes cyclists, cycling infrastructure and cycling laws.

“The creation of these guidelines is part of Argo’s commitment to build trust with community members and to develop an autonomous driving system that provides a level of comfort to riders, behaving in a consistent and safe manner. “Said Peter Rander, chairman and co-founder of Argo AI, in a statement. “We are encouraging other developers of autonomous vehicles to adopt them as well to build confidence among vulnerable road users.”

Argo, which currently operates autonomous test vehicles in the United States and parts of Germany, said it has worked with the LAB community to learn about common rider behaviors and vehicle interactions. Together, Argo and LAB have developed six technical guidelines for self-steering systems to detect cyclists, predict their behavior and drive consistently.

Cyclists must be a separate object class

Treating cyclists as a separate class and labeling them as such will create a diverse set of bicycle images that an autonomous riding system can draw inspiration from. Systems should be trained on images of cyclists from a variety of positions, orientations, points of view and speeds. Argo says it will also help the system take into account the different shapes and sizes of bikes and riders.

“Due to the unique behaviors of cyclists that distinguish them from scooter riders or pedestrians, a self-driving system (or “SDS”) should designate cyclists as a central object representation within its perceptual system in order to to detect cyclists with precision, ”according to a statement from Argo.

Typical cycling behavior is to be expected

Cyclists can be quite unpredictable. They can split up, walk on their mounts, make quick and jerky movements to avoid obstacles in the road, give in to stop signs, jump off the sidewalk and enter the street. A good autonomous driving system must not only be able to predict their intentions, but also be ready to react accordingly.

“An SDS must use specialized, rider-specific motion prediction models that account for a variety of rider behaviors. Thus, when the autonomous vehicle encounters a cyclist, it generates several possible trajectories capturing the potential options of a cyclist’s path, thus allowing the SDS to better predict and respond to the actions of the cyclist.

Map cycling infrastructure and local laws

Autonomous driving systems often rely on high-definition 3D maps to understand its surrounding environment. Part of that environment should be cycling infrastructure and local and national cycling laws, says Argo. This will help the self-drive system to anticipate cyclists’ movements – such as blending into traffic to prevent parked cars blocking the cycle lane or at red lights if there is no traffic – and to maintain a safe distance from the cycle lane.

The system must act in a consistent, understandable and extremely safe manner around cyclists

Autonomous driving technology needs to work in a way that feels natural so that the intentions of the AV are clearly understood by riders, which includes things like using the turn signals and adjusting the position of the vehicle while at the same time. staying in the same lane if they are preparing to pass, merge or turn.

Additionally, if you are driving near cyclists, the system should “Aim for safe and appropriate speeds in accordance with local speed limits and margins at or above local laws, and only overtake a rider when he can maintain those margins and speeds for the entire maneuver, ”says Argo.

The self-driving system should also give cyclists plenty of room in the event of a fall, so that they can swerve or brake.

Prepare for uncertain situations and proactively slow down

Autonomous driving systems should take into account the uncertainty in a rider’s intention, direction and speed, says Argo. The company gave the example of a cyclist traveling in the opposite direction of the vehicle, but in the same lane, suggesting that the vehicle be trained to slow down in this circumstance.

In fact, in most uncertain circumstances, the self-drive system should reduce vehicle speed and, where possible, allow more space between the vehicle and the rider. Slowing down speeds when the system is uncertain is already fairly standard in the world of AV developers, although not always specifically aimed at cyclists.

Continue to test cycling scenarios

The best way to justify the safety of VAs is to keep testing them. Argo and LAB suggest that developers of autonomous driving technologies continue with virtual and physical tests specifically dedicated to cyclists.

“A virtual test program should consist of three main test methodologies: simulation, resimulation and playforward to test a comprehensive permutation of interactions between autonomous vehicles and cyclists on a daily basis,” the company said. “These scenarios should capture both the varying behavior of vehicles and cyclists as well as changes in social context, road structure and visibility.”

Physical testing, which is typically done on closed courses and then on public roads, allows developers to validate the simulation and ensure that the technology behaves the same in the real world as it does in the virtual world. Argo says developers should test AVs on probable scenarios as well as “extreme cases” or rare situations. Testing on multiple public roads in many cities to give the system a diverse set of urban environments to learn from can generate rare and common cases.

In pursuit of public acceptance … and safety, of course

Social acceptance is one of the biggest barriers to the arrival of more AV vehicles on the roads, and many people are still unconvinced of the safety of autonomous vehicles. In fact, nearly half of those polled by market research firm Morning Consult say light commercial vehicles are either a little less safe or much less safe than cars driven by humans.

Making a vehicle safe for all road users is only half the battle. Companies like Argo AI also need to make sure people think their vehicles are safe, and standardizing safety practices in the industry could be one way to do that.