Automated planning and scheduling: Difference between revisions

Revision as of 03:48, 3 July 2019

Revision as of 16:43, 3 July 2019

Line 83:

An early example of a conditional planner is “Warplan-C” which was introduced in the mid 1970s.<ref>{{cite conference |title=Conditional nonlinear planning |author=Peot, Mark A and Smith, David E |conference=Artificial Intelligence Planning Systems |pages=189--197 |year=1992 |publisher=Elsevier}}</ref> Until now, the question was not answered what the difference is between a normal sequence and a complicated plan, which contains if-then-statements. It has to do with uncertainty at [[Run time (program lifecycle phase)|runtime]] of a plan. The idea is, that a plan can react to [[Soft sensor|sensor signals]] which are unknown for the planner. The planner generates two choices in advance. For example, if an object was detected, then action A is executed, if an object is missing, then action B is executed.<ref>{{cite conference |title=Conditional progressive planning under uncertainty |author=Karlsson, Lars |conference=IJCAI |pages=431--438 |year=2001}}</ref> A major advantage of conditional planning is the ability to handle [[Partial-order planning|partial plans]].<ref>{{cite techreport |title=A survey of planning in intelligent agents: from externally motivated to internally motivated systems |author=Liu, Daphne Hao |year=2008 |institution=Technical Report TR-2008-936, Department of Computer Science, University of Rochester}}</ref> An agent is not forced to plan everything from start to finish but can divide the problem into [[Chunking (computational linguistics)|chunks]]. This helps to reduce the state space and solves much more complex problems.

==== Contingent planning ====

We speak of "contingent planning" when the environment is observable through sensors, which can be at faulty. It is thus a situation where the planning agent acts under incomplete information. For a contingent planning problem, a plan is no longer a sequence of actions but a [[decision tree]] because each step of the plan is represented by a set of states rather than a single perfectly observable state, as in the case of classical planning<ref>{{Cite conference|conference=International Joint Conference of Artificial Intelligence (IJCAI)|orig-year=2009|author1= Alexandre Albore| author2 = Hector Palacios| author3 = Hector Geffner| title =A Translation-Based Approach to Contingent Planning| publisher=AAAI|___location =Pasadena, CA|url=http://www.aaai.org/ocs/index.php/IJCAI/IJCAI-09/paper/download/587/852}}</ref>. The selected actions depend on the state of the system. For example, if it rains, the agent chooses to take the umbrella, and if it doesn't, they may choose not to take it.

Mikael L. Littman showed in 1998 that with branching actions, the planning problem becomes [[EXPTIME]]-complete<ref>{{cite conference|first1=Michael L.|last1=Littman|title=Probabilistic Propositional Planning: Representations and Complexity|conference=Fourteenth National Conference on Artificial Intelligence|publisher=MIT Press|year=1997|url=http://citeseer.ist.psu.edu/viewdoc/summary?doi=10.1.1.38.3076|access-date=2019-02-10|pages=748–754}}</ref><ref name="rintanen04">{{cite conference| author1 = Jussi Rintanen| title = Complexity of Planning with Partial Observability| publisher=AAAI | year =2004|url=http://www.aaai.org/Papers/ICAPS/2004/ICAPS04-041.pdf|conference=Int. Conf. Automated Planning and Scheduling}}</ref>. A particular case of contiguous planning is represented by FOND problems - for "fully-observable and non-deterministic". If the goal is specified in LTLf (linear time logic on finite trace) then the problem is always EXPTIME-complete<ref>{{cite conference|first1=Giuseppe |last1=De Giacomo|first2=Sasha|last2=Rubin|title=Automata-Theoretic Foundations of FOND Planning for LTLf and LDLf Goals|year=2018|conference=IJCAI|url=https://www.ijcai.org/proceedings/2018/657|access-date=2018-07-17}}</ref> and 2EXPTIME-complete if the goal is specified with LDLf.

==== Conformant planning ====

Conformant planning is when the agent is uncertain about the state of the system, and it cannot make any observations. The agent then has beliefs about the real world, but cannot verify them with sensing actions, for instance. These problems are solved by techniques similar to those of classical planning<ref>{{Cite journal|title=Compiling uncertainty away in conformant planning problems with bounded width|journal=Journal of Artificial Intelligence Research|last=Palacios|first=Hector|volume=35|pages=623-675|last2=Geffner|first2=Hector|year=2009}}</ref><ref>{{Cite conference|title=Effective heuristics and belief tracking for planning with incomplete information|conference=Twenty-First International Conference on Automated Planning and Scheduling (ICAPS)|last=Albore|first=Alexandre|last2=Ramírez|first2=Miquel|year=2011|last3=Geffner|first3=Hector}}</ref>, but where the state space is exponential in the size of the problem, because of the uncertainty about the current state. A solution for a conformant planning problem is a sequence of actions. Haslum and Jonsson have demonstrated that the problem of conformant planning is [[EXPSPACE]]-complete<ref>{{cite journal|first1=Patrik|last1=Haslum|first2=Peter|last2=Jonsson|title=Some Results on the Complexity of Planning with Incomplete Information|journal=Recent Advances in AI Planning|series=Lecture Notes in Computer Science|publisher=Springer Berlin Heidelberg|year=2000|isbn=9783540446576|doi=10.1007/10720246_24|url=https://link.springer.com/chapter/10.1007/10720246_24|access-date=2019-02-10|pages=308–318}}</ref>, and 2EXPTIME-complete when the initial situation is uncertain, and there is non-deternimism in the actions outcomes<ref name="rintanen04"/>.

== Deployment of planning systems ==