I applaud your goal. I wish we, as scientists, had a better answer to your question. There have been various suggestions as how the police might test, but all the chemical tests suffer from a number of drawbacks:
1) the relationship between THC (the psychoactive ingredient in marijuana) levels in blood and impairment of eye-hand coordination, reaction time and other components of driving skill is not a straightforward one. Also, individual differences of impairment among different users are so great that it would be very difficult to set a fair legal standard of impairment that would apply to everyone.
2) Another problem is getting a precise THC quantification for purposes of a legal charge. You cannot get this with a breath test. A blood sample would be needed as there's no THC equivalent to the breathalyzer, which gives a very accurate estimate of the blood alcohol level. In the end, what one could conclude from such a measure would be pretty doubtful anyway.
My own feeling is that people who are impaired for any reason should not be driving. Moderate sleep deprivation, for example, causes about the same degree of impairment of driving skills as the legal blood alcohol limit. So, I'd rather see us go to roadside behavioural tests that are more sophisticated than the ones now employed, instead of trying to rely on a chemical "quick fix."
3) Unlike alcohol, whose metabolism in the body progresses in such a way that if you know someone's blood alcohol concentration (BAC) at a given time (say two hours after the time of an accident), you can accurately extrapolate back in time to know what the BAC would have been at the time of the accident and make some estimate of impairment at that time. A scientist named Marilyn Heustis has devised some calculation methods with blood levels of THC that try to get around these limitiations with marijuana, but they are far from universally agreed upon in the field of psychopharmacology. It's doubtful whether they'd stand up in court the way breathalyzer results do.