Derek Long | King's College London (original) (raw)
After 50 years in full-time education (including school!), I joined Schlumberger as a Scientific Advisor, supporting them in using Planning and Automated Execution to achieve automation of their drilling systems. We have deployed plan-based control systems using "PDDL planning" (the standard modelling language for academic planners is PDDL) on "big iron" - rigs and other large physical hardware systems.
less
Uploads
Papers by Derek Long
Artificial Intelligence, Dec 1, 2015
In this paper we explore the deployment of planning techniques to solve a new class of metric tem... more In this paper we explore the deployment of planning techniques to solve a new class of metric temporal planning problems, characterised by the need to manage both plan trajectory constraints and uncontrollable numeric events. This combination gives rise to challenges not previously solved in state-of-the-art planners. We introduce new planning methods to handle these challenges, and demonstrate our approach using a real application domain: voltage control in Alternating Current (AC) electrical networks. Embedding electricity networks in a domain description presents important modelling challenges. We introduce an encapsulated type, Network, the implementation of which is hidden from the planner. The effects of actions trigger complex updates to the state of the network. We distinguish between the direct effects of planned actions, and the indirect effects triggered by them, and we propose a method for integrating a specialised external AC power equation solver with a planner. We consider a new heuristic function that takes into account the next uncontrollable event, and its interaction with active trajectory constraints, when determining the actions that are helpful in a state. This lookahead heuristic also exploits an abstraction of the encapsulated Network type to obtain more informative distance estimates. We conduct experiments to evaluate the benefits of the lookahead heuristic, showing that our approach scales very well with the size of the network and the number of controllable components of the network.
Proceedings of the International Conference on Automated Planning and Scheduling, Mar 30, 2016
International Conference on Artificial Intelligence, Jul 25, 2015
EarthArXiv (California Digital Library), Sep 2, 2022
Proceedings of the International Conference on Automated Planning and Scheduling, Mar 30, 2016
OCEANS 2016 - Shanghai, Apr 1, 2016
Proceedings of the International Conference on Automated Planning and Scheduling, Jun 5, 2017
Proceedings of the ... AAAI Conference on Artificial Intelligence, Mar 5, 2016
Proceedings of the International Conference on Automated Planning and Scheduling, May 25, 2021
Proceedings of the International Conference on Automated Planning and Scheduling, May 14, 2012
Artificial Intelligence, Dec 1, 2015
In this paper we explore the deployment of planning techniques to solve a new class of metric tem... more In this paper we explore the deployment of planning techniques to solve a new class of metric temporal planning problems, characterised by the need to manage both plan trajectory constraints and uncontrollable numeric events. This combination gives rise to challenges not previously solved in state-of-the-art planners. We introduce new planning methods to handle these challenges, and demonstrate our approach using a real application domain: voltage control in Alternating Current (AC) electrical networks. Embedding electricity networks in a domain description presents important modelling challenges. We introduce an encapsulated type, Network, the implementation of which is hidden from the planner. The effects of actions trigger complex updates to the state of the network. We distinguish between the direct effects of planned actions, and the indirect effects triggered by them, and we propose a method for integrating a specialised external AC power equation solver with a planner. We consider a new heuristic function that takes into account the next uncontrollable event, and its interaction with active trajectory constraints, when determining the actions that are helpful in a state. This lookahead heuristic also exploits an abstraction of the encapsulated Network type to obtain more informative distance estimates. We conduct experiments to evaluate the benefits of the lookahead heuristic, showing that our approach scales very well with the size of the network and the number of controllable components of the network.
Proceedings of the International Conference on Automated Planning and Scheduling, Mar 30, 2016
International Conference on Artificial Intelligence, Jul 25, 2015
EarthArXiv (California Digital Library), Sep 2, 2022
Proceedings of the International Conference on Automated Planning and Scheduling, Mar 30, 2016
OCEANS 2016 - Shanghai, Apr 1, 2016
Proceedings of the International Conference on Automated Planning and Scheduling, Jun 5, 2017
Proceedings of the ... AAAI Conference on Artificial Intelligence, Mar 5, 2016
Proceedings of the International Conference on Automated Planning and Scheduling, May 25, 2021
Proceedings of the International Conference on Automated Planning and Scheduling, May 14, 2012