Kairos
Timeline (last 90 days)
Kairos is a framework that selects heterogeneous planning engines according to the semantics of each task, covering a broad range from STRIPS to PDDL+.
Kairos generates a semantic profile for each task, selects an applicable engine from five engines based on it, and schedules under a unified time and memory budget.
All candidate plans in Kairos are checked by a validator that is independent of the planner translation, and only plans that pass validation are returned.
Kairos reports "unsolvable" only when some search process exhausts space and its exhaustiveness constitutes a proof, in order to avoid misjudging a timeout as unsolvable.
According to the paper, on 1,756 tasks from nine comparison suites, Kairos solved more tasks under full resource limits than the strongest evaluated planner in each suite, with the differences most pronounced in numeric, temporal, PDDL3, and PDDL+ planning.
Under strict time limits of 100 milliseconds and 1 second, Kairos also solved the most tasks.
After resampling the domains, the gain was greater than zero in six of the nine suites; on classical planning, compared with LAMA, the difference is within the noise range.
An additional 450 feature-oriented tasks and 26 single-feature probes show that Kairos can handle all seventeen probed features, while none of the other evaluated planners can handle more than eight.
Ablation experiments show that compressed-time search and novelty fallback are the main contributors to Kairos's performance.
Comparison with the VAL validator traces every disagreement to a PDDL+ convention or a known defect of VAL.
2 more timeline events
All Facts (12)
Kairos is a framework that selects heterogeneous planning engines according to the semantics of each task, covering a broad range from STRIPS to PDDL+.
50%UnverifiedKairos generates a semantic profile for each task, selects an applicable engine from five engines based on it, and schedules under a unified time and memory budget.
50%UnverifiedAll candidate plans in Kairos are checked by a validator that is independent of the planner translation, and only plans that pass validation are returned.
50%UnverifiedKairos reports "unsolvable" only when some search process exhausts space and its exhaustiveness constitutes a proof, in order to avoid misjudging a timeout as unsolvable.
50%UnverifiedAccording to the paper, on 1,756 tasks from nine comparison suites, Kairos solved more tasks under full resource limits than the strongest evaluated planner in each suite, with the differences most pronounced in numeric, temporal, PDDL3, and PDDL+ planning.
50%UnverifiedUnder strict time limits of 100 milliseconds and 1 second, Kairos also solved the most tasks.
50%UnverifiedAfter resampling the domains, the gain was greater than zero in six of the nine suites; on classical planning, compared with LAMA, the difference is within the noise range.
50%UnverifiedAn additional 450 feature-oriented tasks and 26 single-feature probes show that Kairos can handle all seventeen probed features, while none of the other evaluated planners can handle more than eight.
50%UnverifiedAblation experiments show that compressed-time search and novelty fallback are the main contributors to Kairos's performance.
50%UnverifiedComparison with the VAL validator traces every disagreement to a PDDL+ convention or a known defect of VAL.
50%UnverifiedKairos's planners and benchmarks are both open-sourced.
50%UnverifiedThe current engine set of Kairos is fixed at five engines.
50%