Incremental Timeline-Based Planning for Efficient Plan Execution and Adaptation

The increasing deployment, in real environments, of intelligent and distributed systems like robotic platforms, wearable sensors and AI-based devices, requires robust solutions that allow planned activities to converge with the emerging dynamic reality. Once a planning problem has been solved, indeed, it needs to be executed and, in the real world, things might not go as expected. While planned activities may be carried out by some underlying reactive modules, in fact, the adaptation to the surrounding environment provided by such components may not be sufficient to achieve the planned goals. Planned activities, for example, can be delayed or last longer than expected. The execution of other activities could fail threatening the achievement of the desired goals. Finally, new objectives may emerge during execution thus requiring changes to ongoing plans. This paper presents a timeline-based framework for efficiently adapting plans in order to cope with possible complications which might emerge during execution. By exploiting the information gathered during the finding solution process, the proposed framework allows, efficiently and without overturning it, to adapt the generated plan in case of unexpected events during its execution. Empirical results show that, compared to re-planning from scratch, plan adaptations can be obtained more efficiently, reducing computational costs and consequently enhancing the ability of the whole system to react quickly to unexpected events.