Automatic Planning Lab
Our research revolves around automatic planning in all of its forms: time management, robots with autonomous capabilities, AI assistants (Automated Operations Department), etc. Our team is currently working on automated time management, that is, developing optimal schedules based on statistical data.
Current problem
A user has several blocks of free time available (for example, one per day). He has to complete N tasks, maximizing the remaining free time. Our problem: given the graph of dependencies between tasks "OrderGraph", which defines a partial order of task execution (for instance, if the lecture must precede the assignment, this graph will include a directed edge from lecture to assignment) and the performance influence graph “InfluenceGraph”, which indicates how task execution times are influenced by tasks previously completed, create an optimal schedule. Additionally, we need to create an “InfluenceGraph” based on the user's statistical data and adjust it based on the accuracy of our predictions.
Notation
In the OrderGraph, the directed edge (a, b) indicates that task "a" must be completed before task "b". In the InfluenceGraph, the directed edge (a, b, w_sameday, w_sameweek) denotes the following relationships:
1) If "a" gets completed the same block before "b", then the modifier w_same (w_same > 0) is applied to the execution time of "b"
2) If "a" gets completed the same cycle before "b", then the modifier w_sameweek (w_sameweek > 0) is applied to the execution time of "b".
P.S.: For example, we might consider blocks to be daily 3-hour periods and cycles to be weeks.