December 6, 2021
Argo AI has announced the release of technical guidelines it applies to ensure safe interactions between autonomous vehicles and cyclists. The guidelines were created in collaboration with the League of American Bicyclists, a national advocacy group, and are intended as a foundation for further innovation and improvement among companies developing self-driving technology, the company said.
“Argo AI is focused on developing self-driving technology that makes cities safer for everyone – in particular cyclists and other vulnerable road users,” said Peter Rander, president and co-founder of Argo AI. “These technical guidelines deliver on our commitment to developing a self-driving system that is trusted by cyclists and enhances the safety of the communities in which we operate.”
To understand the concerns among cyclists when sharing the road, Argo said it collaborated and engaged with the cycling community. The League of American Bicyclists consulted with the company to give them common cyclist behaviors and typical interactions with vehicles. Together, the group outlined six technical guidelines for the manner in which a self-driving system should accurately detect cyclists, predict their behavior, and drive in a consistent way to effectively and safely share the road. These include:
- Cyclists should be a distinct object class. Because they have unique behaviors different from scooter users or pedestrians, a self-driving system should designate cyclists as a core object representation within its perception system. “By treating cyclists as a distinct class and labeling a diverse set of bicycling imagery, a self-driving system detects cyclists in a variety of positions and orientations, from a variety of viewpoints, and at a variety of speeds,” the guideline states. Systems should also account for the different shapes and sizes of bikes, such as recumbent bikes, bikes with trailers, electric bikes and unicycles.
- Typical cyclist behaviors should be expected. Systems should understand potential cyclist patterns of movement to predict their intentions and prepare the vehicle’s actions. For example, a cyclist may yield at stop sights, walk a bicycle, or make quick deliberate lateral movements to avoid obstacles, like the swinging open of a car door. A self-driving system needs to “utilize specialized, cyclist-specific motion forecasting models that account for a variety of cyclist behaviors, so when the self-driving vehicle encounters a cyclist, it generates multiple possible trajectories capturing the potential options of a cyclist’s path” to enable the system to better predict and respond.
- Cycling infrastructure and local laws should be mapped. This includes marking where dedicated bike lanes are located, and include all local and state cycling laws to make sure self-driving vehicles are compliant. This can include monitoring behavior such as merging into traffic to avoid parked cars blocking a bike lane or treating a red light as a stop sign, known as an “Idaho stop”, which is legal in some states.
- A self-driving system should drive in a consistent and understandable way. These vehicles should operate naturally so the intentions of an autonomous vehicle is understood by other road users. “In the presence of nearby cyclists or when passing or driving behind cyclists, a SDS should target conservative and appropriate speeds in accordance with local speed limits, and margins that are equal to or greater than local laws, and only pass cyclists when it can maintain those margins and speeds for the entire maneuver.”
- Prepare for uncertain situations and proactively slow down. This includes accounting for uncertainty in cyclists’ intent, direction and speed, such as when a cyclist is traveling itn the opposite direction of the vehicle in the same lane.”When there is uncertainty, the self-driving system should lower the vehicle’s speed and, when possible, increase the margin of distance to create more time and space between the self-driving vehicle and the cyclist and drive in a naturalistic way.”
- Cyclist scenarios should be tested continuously. This includes both virtual testing and physical testing, in all phases of development.
The two organizations said the development of these guidelines are intended for adoption of industry best practices that promote special consideration of cyclist behavior and interactions. “We encourage other autonomous vehicle developers to adopt them as well to further build trust among vulnerable road users,” said Rander. The guidelines build upon the six principles that Argo outlined last year for the development of a self-driving system that prioritizes safe interactions with vulnerable road users.
“Argo AI and the League of American Bicyclists share a common goal to improve the safety of streets for all road users,” said Ken McLeod, policy director for the league. “We appreciate Argo’s proactive approach to researching, developing, and testing for the safety of people outside of vehicles. Roads have gotten significantly less safe for people outside of vehicles in the last decade, and by addressing interactions with bicyclists now, Argo is demonstrating a commitment to the role of automated technology in reversing that deadly trend.”
Argo AI said it currently operates self-driving test vehicles in Miami; Austin, Texas; Washington, DC.; Pittsburgh, Detroit; and Palo Alto, Calif. The company also recently expanded on-road testing in Munich and Hamburg, Germany. For more information about Argo AI, visit its website here.