In real life, at least under clear weather, there appears to be a (slight) asymmetry between the conditions under which human drivers adjust lighting in the morning (what we model as seeing→reduced-seeing→being-seen→off transitions) and in the evening (off→being-seen→reduced-seeing→seeing transitions), with the former set of conditions tending to be "laxer" ("worse") than the latter. In other words, the same driver tends to overall be more tolerant of poor conditions in the morning than in the evening, as if, in the former case "anticipating" / "preparing for" the day, while in the latter "anticipating" / "preparing for" the night.
Hypotheses:
Mood: In the afternoon - evening, drivers are tired or expect others to be tired, and hence activate lights earlier as a precautionary measure.
Laziness: Depending on the region and time of the year, home-work commuters may constitute the majority of morning / afternoon - evening traffic. All these people would start their trips to/from work at roughly the same time. In the morning, even if it is not quite bright outside yet, a significant percentage of them might think "even though it's not quite bright outside yet, it will be become bright long before I reach work, so I'm not gonna bother turning on lights for just a portion of my trip (and risk forgetting them on for the trip's entirety)". In the evening a significant percentage might think the opposite and turn on lights before actually necessary, to avoid risking turning them on too late.
Considerations:
This project is about simulating buses and professional bus driver behavior, not average me-and-you car drivers. We can't empirically tell whether this observed asymmetry, pertaining almost exclusively to observed car traffic, also holds for buses; the speculated theory certainly doesn't.
The casual observer would be unlikely to notice an asymmetry as slight as +/- 5 degrees solar elevation, or +/- 10% insolation / brightness, were it to be introduced to the AI's perception and behavior.
In real life, at least under clear weather, there appears to be a (slight) asymmetry between the conditions under which human drivers adjust lighting in the morning (what we model as seeing→reduced-seeing→being-seen→off transitions) and in the evening (off→being-seen→reduced-seeing→seeing transitions), with the former set of conditions tending to be "laxer" ("worse") than the latter. In other words, the same driver tends to overall be more tolerant of poor conditions in the morning than in the evening, as if, in the former case "anticipating" / "preparing for" the day, while in the latter "anticipating" / "preparing for" the night.
Hypotheses:
Considerations: