Argo AI has partnered with advocacy group the League of American Bicyclists (LAB) to create guidelines for how self-driving vehicles should identify and interact with cyclists. The goal is to set a standard for other AV companies in the industry to follow, especially The self-driving industry moves away from testing and toward commercialization and will become more popular in the coming years.
The World Health Organization estimates that 41,000 cyclists die in road traffic-related incidents each year. While self-driving vehicles are expected to significantly reduce collisions, much of that expected safety is the result of good coding in the first place. Self-driving cars learn from a huge database that categorizes and identifies possible objects and situations, and Argo’s instruction emphasizes training its models in a unique way. Note cyclists, cycling infrastructure and cycling laws.
“The creation of these guidelines is part of Argo’s dedication to building trust with community members and developing self-driving systems that provide a level of comfort for cyclists, by behave consistently and securely,” said Peter Rander, president and co-founder of Argo AI, in a statement. “We encourage other autonomous vehicle developers to adopt them as well to further build trust among vulnerable road users.”
Argo, which is currently running tests of self-driving cars across the US and parts of Germany, says it has partnered with LAB’s community to hear about common cyclists’ behaviors and interactions with vehicles. Together, Argo and LAB have come up with six technical guidelines for self-driving systems to detect cyclists, predict cyclist behavior, and drive consistently.
Cyclists must be a separate object class
Treating cyclists as a class and labeling them as such creates a diverse set of bike images for the self-driving system to learn from. The system should be trained on images of cyclists from various positions, directions, viewpoints, and speeds. This will also help the system calculate different shapes and sizes of bicycles and cyclists, Argo says.
“Since the unique behaviors of cyclists distinguish them from scooter users or pedestrians, self-driving systems (or ‘SDS’) should designate cyclists as core audience representatives in its cognitive system to accurately detect cyclists,” according to a statement from Argo.
Typical cyclist behavior should be expected
Cyclists can be quite unpredictable. They can separate lanes, walk on horseback, move quickly, jerk to avoid obstacles in the road, give way at stop signs, jump off sidewalks and cross the street. A good self-driving system should not only be able to predict their intentions, but also be prepared to react accordingly.
“SDS should use specialized, cyclist-specific motion prediction models that account for a variety of cyclist behaviors, so that when self-driving cars meet cyclists, it will create multiple trajectories that can capture potential cyclists’ path options, thus allowing SDS to better predict and respond to cyclist actions.”
Map cycling infrastructure and local laws
Self-driving systems often rely on high-definition 3D maps to understand their surroundings. Part of that environment is cycling infrastructure and local and state cycling laws, Argo said. This will help the self-driving system predict a cyclist’s movements – like merging with traffic to avoid a parked car blocking a bike lane or running a red light if there’s no traffic – and keep a safe distance from the bike lane.
The system must operate in a consistent, understandable and safe manner for cyclists
Self-driving technology should work in a way that seems natural for cyclists to understand the AV’s intentions, including things like using turn signals and adjusting vehicle position while still in a lane if prepare to pass, enter, or turn.
In addition, if driving near cyclists, the system will “aim for a moderate and appropriate speed according to local speed limits and boundaries equal to or greater than local laws and only overtake a cyclist when he or she can maintain that boundary and speed throughout the entire journey. program,” said Argo.
The self-driving system should also provide cyclists with a wide stop in case they fall, so it can change direction or stop.
Prepare for unforeseen situations and proactively slow down
Self-driving systems should account for uncertainty about the cyclist’s intentions, direction and speed, Argo said. The company gave an example of a cyclist traveling in the opposite direction of the vehicle but in the same lane, suggesting that he or she be trained to slow down in that case.
In fact, in most uncertain cases, self-driving systems should reduce the vehicle’s speed and, when possible, create more distance between the vehicle and the cyclist. Reducing speed when the system is uncertain is the norm in the AV developer world, even if it’s not always specifically targeted at cyclists.
Keep testing cycling situations
The best way to create a safe case for AVs is to keep testing them. Argo and LAB suggest that developers of self-driving technology should continue to experiment with both virtual and physical, which are exclusive to cyclists.
A virtual test program should be made up of three main testing methods: simulation, re-simulation, and simple play to test the comprehensive permutation of interactions between autonomous vehicles, the company said. and daily cyclists”. “These scenarios will capture both the different behaviors of vehicles and cyclists as well as changes in social context, road structure and visibility.”
Physical testing, often done on closed courses and then on public roads, allows developers to validate simulations and ensure the technology works just as much in the real world as it does in the virtual world. . Argo says that developers should test AV on possible scenarios as well as “edge cases” or rare scenarios. Experiment on multiple public roads in multiple cities to provide the system with a diverse set of urban environments to learn from that can generate both rare and common.
Pursue public acceptance…and safety, of course
Social acceptance is one of the key barriers to getting more AV out into the streets, and many people are still unconvinced about the safety of their devices. self-driving car. In fact, nearly half of those polled by market research firm Morning Consult says AV is somewhat less or much less secure than human-driven cars.
Making a vehicle safe for all road users is only half the battle. Companies like Argo AI must also make sure people trust their vehicles are safe and standardizing safety practices across the industry can be one way to do that.
