As it currently stands, both lighting activation and deactivation conditions, as well as usage preferences, are effectively set in stone for each vehicle post-initialization. For real-world drivers, of course, this is not the case: Perception (thresholds) slightly vary each time a decision is made, for many reasons, and preferences are not constant either (the same driver may use headlights one night and headlights plus fog lights another).
As a partial workaround, since 1.0.1, vehicles have been getting reconfigured anew each time control is (re-)transferred to the AI. This enables a user to witness different simulated driver profiles being applied to the same vehicle. Furthermore, the actualization time under changing conditions is now dynamically re-evaluated each time a distinct transition gets scheduled.
As for implementing more dynamic decision making, however, while certainly feasible, we feel that it would not be productively invested time and effort, because the likelihood of users ever noticing, particularly when engaged in driving a vehicle of their own, would be close to zero. And ultimately, to preserve consistency, we would have to repeat the same exercise for all the other AI preferences beyond lighting as well, which would cause complexity to go through the roof.
As it currently stands, both lighting activation and deactivation conditions, as well as usage preferences, are effectively set in stone for each vehicle post-initialization. For real-world drivers, of course, this is not the case: Perception (thresholds) slightly vary each time a decision is made, for many reasons, and preferences are not constant either (the same driver may use headlights one night and headlights plus fog lights another).
As a partial workaround, since 1.0.1, vehicles have been getting reconfigured anew each time control is (re-)transferred to the AI. This enables a user to witness different simulated driver profiles being applied to the same vehicle. Furthermore, the actualization time under changing conditions is now dynamically re-evaluated each time a distinct transition gets scheduled.
As for implementing more dynamic decision making, however, while certainly feasible, we feel that it would not be productively invested time and effort, because the likelihood of users ever noticing, particularly when engaged in driving a vehicle of their own, would be close to zero. And ultimately, to preserve consistency, we would have to repeat the same exercise for all the other AI preferences beyond lighting as well, which would cause complexity to go through the roof.