According to Spectrum, autonomous vehicle researchers say bikes are the most difficult to detect because they are relatively small, fast and heterogenous, meaning they don't have a lot of mass and can vary in appearance as people attach strollers or hang things off of them. They are also much less predictable because they can make sudden turns or jump out of nowhere.
Data shows that autonomous vehicle technology identifies approximately 89 percent of cars. Yet it could only spot 74 percent of bikes in testing. Technology is improving though and visual processing of roadways is being augmented with laser-scanning imagery and radar sensing, which helps cars to "see" that there is an object in the roadway, even they can't identify what it is.