Argo AI, the self-driving company that's currently working with VW on its ID.Buzz van, as well as Walmart and Ford to take down Amazon, has teamed up with the League of American Cyclists to figure out the safest combination on the road. Evolving a self-driving program is difficult, as Tesla has proven, and as more of these systems come online they're going to have to know the difference between a pedestrian, a cyclists, an animal, a self-driving vehicle and a standard vehicle, instantaneously.

The World Health Organization estimates that 41,000 cyclists are killed each year and road-related incidents. If all goes according to plan, that number should drop precipitously. Tech Crunch reports that good self-driving cars should recognize the differences, and that starts during the coding.

"The creation of these guidelines is part of Argo's dedication to building trust with community members and developing a self-driving system that provides a level of comfort to cyclists, by behaving consistently and safely," Peter Rander, president and co-founder of Argo AI, said in a statement. "We encourage other autonomous vehicle developers to adopt them as well to further build trust among vulnerable road users."

Argo partnered with LAB to hear about the community's experiences with vehicles on the road. Argo is currently testing in the US and other parts of the world. The pair came up with six guidelines for systems to keep cyclists safe, which also involves the self-driving car, driving consistently.

First, cyclists should be classified as a separate object class. Treating them like pedestrians, which is what many systems do now, makes no sense as they don't have the same speed, vectors or profile of walker. These systems should be trained to see the shape of a cyclist from all angles. They should also expect typical cyclist behavior like lane splitting, walking their bike, and avoiding obstacles.

Argo and LAB also want to codify a cycling infrastructure, that could be represented in navigational maps that autonomous vehicles use. Some states have different laws around cycling, they should be represented. And like we said at the top, self-driving cars should act in a predictable manner.

If a vehicle nears a cyclist, it 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 a cyclist when it can maintain those margins and speeds for the entire maneuver," Argo said.

Finally, Argo and LAB say these systems should proactively slow down in unfamiliar situations and that self-driving systems should continually test cycling scenarios. That's the only way these machine learning systems get better. And if they wanted to accepted by everyone else on the road, and the public in general, this is a good start.