202111091028 Task scheduling is a fuzzy process

Some tasks are bounded on time. We don’t do the dishes every day at exactly 8:00 PM, we do it after we’re done making new dishes and before bed.

Some tasks are done relative to other constraints. We don’t mow if it’s raining or we mow early when we know it might rain when we’d normally do it. It’s front-loaded to not go over, so we allow sooner openings.

This will be relevant in making systems and AI that can act autonomously and accomplish variable, competing tasks.

These can best be described as systems of constraints. Likely we can make a system of equations that describe a range of acceptable times that things can be done, and then choose times within that span. This would allow for layered sophistication.

For instance, with the dishes, we know it has to be done after we're done making dishes, but before bed. We can then add more soft requirements to prefer when no one else is in the kitchen, but if that's not possible, it will still do it regardless of the person in the kitchen. We can see how this can be progressively enhanced via learning methods to create gradients and optimal maxima in n-dimensional space from the n-dimensional system of constraint equations.