Use this URL to cite or link to this record in EThOS:
Title: Enhanced heuristics for numeric temporal planning
Author: Piacentini, Chiara
ISNI:       0000 0004 5369 1351
Awarding Body: King's College London (University of London)
Current Institution: King's College London (University of London)
Date of Award: 2015
Availability of Full Text:
Access from EThOS:
Access from Institution:
After 50 years of fundamental research, domain independent planning has recently started to be applied to numerous real world problems. However, this has shown that the techniques developed until now are not completely mature: improvements can be made in different directions, such as in the area of metric temporal planning. This PhD research is focused on how we can use more sophisticated and informative heuristics in the general context of automated planning, when numeric and temporal constraints are a significant part of the problem. As a starting point, we will use as a reference example the voltage control problem in distributed electricity networks of power systems. This domain is a real world application of planning in which non-linear numeric effects and exogenous events are combined, challenging the state of the art planners. A power system is a nation-wide infrastructure which delivers electricity from suppliers to consumers. Technical and economic considerations impose constraints on the different elements of the network and subsequently on control and controlling variables of interest. One of the main parameters is the voltage, which must lie in strict boundaries. The voltage, as well as all the other physical quantities involved, is subject to the variation of the electrical output that changes through time. Effects of these changes propagate all across the network. This comes with a substantial computational burden and calls for extensions to be developed to enable the application of automated planning. In addition to this, the presence of events representing predicted load and supply over time require the planner to interact with uncontrollable events. The voltage control problem is the starting point for our investigation on how the standard delete-relaxation behaves in the presence of numeric and temporal constraints. Fully relaxing all the negative effects can result in a too poorly informed heuristic. In this thesis we explore different ways to enhance the heuristic, adding selected negative effects, while not compromising too much the efficiency of the heuristic computation. In particular we have studied the numeric temporal heuristic of the planner popf2, based on the Temporal Relaxed Planning Graph (TRPG), and propose a way to take into account numeric effects that are calculated by external modules connected to the planner. Negative effects of predictable numeric exogenous events in the presence of trajectory constraints are also taken into account in the heuristic.
Supervisor: Fox, Maria; Long, Derek Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available