Robust dynamic planning and scheduling

People

Supervisor

Research areas

Description

Plans and schedules are the foundation on which much of human activity is organised. Yet, activities often do not go according to plan, or stay on schedule. Therefore, we aim to develop algorithms for monitoring plan/schedule execution, and dynamically replanning when necessary. These algorithms will be integrated into decision support systems.

Monitoring algorithms should be able to extrapolate from the current situation to determine the impact that deviations from the plan/schedule (for example, delays) will have on requirements, such as deadlines, later in time. This implies a tight integration between monitoring and planning/scheduling. Monitoring must also be able to work with limited, delayed and possibly incorrect observations of the plan's execution (for instance, a job may have been finished on time, but not reported as finished until much later), requiring diagnostic inference. Planning/scheduling algorithms, on the other hand, should be able to optimise for robustness against future deviations, or quantify the deviations (delays, resource unavailability and other contingencies) within which plan success is guaranteed.

Goals

We aim to build on our current work in diagnosis, planning and temporal reasoning, extending in many directions. The scope of a project on this topic is variable, depending on whether it is a single-semester undergraduate, honours or PhD project.

Requirements

A strong background in computer science, and very good programming skills.

Background Literature

  • Jing Cui, Peng Yu, Cheng Fang, Patrik Haslum, Brian C. Williams. Optimising Bounds in Simple Temporal Networks with Uncertainty under Dynamic Controllability Constraints. International Conference on Automated Planning and Scheduling, p. 52-60, 2015 (paper available from AAAI).
  • Jing Cui, Patrik Haslum. Dynamic Controllability of Controllable Conditional Temporal Problems with Uncertainty. International Conference on Automated Planning and Scheduling, p. 61-69, 2017 (paper available from AAAI).
  • Patrik Haslum and Alban Grastien. Diagnosis as Planning: Two Case Studies. ICAPS'11 Scheduling and Planning Applications Workshop, 2011. Available from the workshop web page.

Updated:  1 June 2019/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing