AASS Seminar: The quest towards a continuous decision process integrating task plan generation, execution, active diagnosis and active learning: work done and potential avenues
28 maj 2026 13:00 – 14:00 Hörsal L2, Långhuset
The research centre AASS arranges a seminar with Thierry Vidal, Associate Professor from University of Technology of Tarbes - Research fellow, LAAS-CNRS Toulouse.
Who: Thierry Vidal, Associate Professor, University of Technology of Tarbes - Research fellow, LAAS-CNRS Toulouse
What: Seminar: "The quest towards a continuous decision process integrating task plan generation, execution, active diagnosis and active learning: work done and potential avenues"
When: Thursday May 28, at 13:00
Where: Hörsal L2, Långhuset
Remote attendance: https://oru-se.zoom.us/j/65096623293
Abstract: Autonomous agents or automated services need to build their plans of activities and execute them in a real world. We are neither interested here in the task planning or scheduling search, nor in the real-time constraints of the execution controller, but somewhere in-between, i.e., how a task plan should and might be designed and managed in order to better cope with the uncertainties of the real world. We will first quickly expose some past works around this theme: how to check the temporal consistency of a plan against temporal uncertainties, depending on the execution and observation semantics of the agent; a taxonomy of the different ways to interleave planning and execution when facing different types of uncertainties; how to relate diagnosability and repairability of a system with non-observable faults. Then we will focus on recent contributions around the design of a global multi-agent architecture in which agents with interdependent plans must ensure their temporal coordination: how to model it; how to check the global consistency in a distributed way; and when checking fails, how to repair the plans through negotiations on activity durations without having to replan. Finally, we will take a step back and examine the potential of integrating (re)active planning, sensing, diagnosis and learning in proactive autonomous agents, thereby paving the way for some potential and currently explored research collaboration with CRS lab.