"Maximize cumulative rewards" - rewards paid to what, or whom, and in what form, energy (In any form, via money, for example, or energy saved to travel further) ?

If this is what you were doing, then you are just teaching machines to be like us - driven by and for profit, which is the fundamental reason we are tied to the now depleting extractable energy of the planet, pushing up entropy rather than moving to the energy of the sun, reducing entropy like we could be, if we were not enslaved by profit.

So at first what you might see is benefits, in the earlier stages of the competition in which machines are competing with one another and us, traffic might flow more freely, but wait until the competition starts to progress towards the winners becoming ever fewer and the rewards become ever more slight, sooner or later they have to start to fight, and then what is the effect on traffic, and safety, when the only reason anything left still driving, is to get the reward it might need to just carry on surviving... is that not just what we are seeing right now, in the final stages of capitalism dying?