Road infrastructure has been designed, standardised, and optimised around human perception for decades—but as vehicles become increasingly automated, this human-centric foundation is being put to the test. Are current road markings, signage, and traffic systems fit for machine interpretation, or are they introducing unnecessary complexity for perception systems?
At the centre of this debate is whether autonomy demands a shift toward new reference models, including the concept of a non-biological reference observer, and how emerging or evolving standards such as P2020 should respond. Machine vision does not see the road in the same way humans do, particularly under challenging lighting and weather conditions, raising questions around how sensor stacks, redundancy, and perception pipelines can compensate—or whether infrastructure itself must evolve.
The gap between ADAS and fully autonomous systems further complicates the picture, with differing requirements for reliability, interpretability, and scale. As mixed-traffic environments persist, the path toward autonomy may depend as much on adapting the road as the vehicle, from incremental changes to signage through to V2X-enabled infrastructure.
