An Engineer's Approach to the Application of Knowledge Based Planning and Scheduling Techniques to Logistics

O-Plan is a command, planning and control architecture with an open modular structure intended to allow experimentation on, or replacement of, various components. The research is seeking to determine which functions are generally required in a number of application areas and across a number of di erent command, planning, scheduling and control systems. O-Plan aims to demonstrate how a planner, situated in a task assignment and plan execution (command and control) environment, and using extensive domain knowledge, can allow for exible, distributed, collaborative, and mixed-initiative planning. The research is seeking to verify this total systems approach by studying a simpli ed three-level model with separable task assignment, plan generation and plan execution agents. O-Plan has been applied to logistics tasks that require exible response in changing situations. The O-Plan research has achieved a clearer understanding of the components necessary in a exible planning system, and has shown how such components can be combined in an open systems integration architecture. The work has determined improved ways in which the knowledge available from modelling an application domain can be used e ectively to restrict search in a planner. An improved characterisation of a plan as a set of constraints on activity opens up many possibilities for richer distributed, cooperative and mixed-initiative planning systems in the future. The project has created a prototype implementation which has been demonstrated on a class of realistic applications. iii Acknowledgements The O-Plan project began in 1984. Since that time the following people have participated: Colin Bell, Ken Currie, Je Dalton, Roberto Desimone, Brian Drabble, Mark Drummond, Anja Haman, Ken Johnson, Richard Kirby, Arthur Seaton, Judith Secker, Glen Reece, Austin Tate and Richard Tobin. Prior to 1984, work on Interplan (1972-4) and Nonlin (1975-6) was funded by the uk Science and Engineering Research Council. From 1984 to 1988, the O-Plan project was funded by the Science and Engineering Research Council on grant numbers gr/c/59178 and gr/d/58987 (ukAlvey Programme project number ikbs/151). The work was also supported by a fellowship from sd-Scicon for Austin Tate from 1984 to 1985. From 1989 to 1992, research on scheduling applications of the O-Plan architecture was funded by Hitachi Europe Ltd. A number of other research and development contracts placed with aiai have led to research progress on O-Plan. From 1989 to 1995, the O-Plan project has been supported by the us Air Force Rome Laboratory through the Air Force O ce of Scienti c Research (afosr) and their European O ce of Aerospace Research and Development by contract numbers F49620-89-C0081 (eoard/88-0044) and F49620-92-C-0042 (eoard/92-0001) monitored by Northrup Fowler iii at the usaf Rome Laboratory. The projects have been part of the arpa/Rome Laboratory Planning Initiative (arpi). Additional resources for the O-Plan projects have been provided by the Arti cial Intelligence Applications Institute through the europa (Edinburgh University Research on Planning Architectures) Institute development project. The views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing the o cial policies, either expressed or implied, of the Advanced Research Projects Agency or the U.S. Government. This research was jointly sponsored by the Advanced Research Projects Agency of the Department of Defense and by the U.S. Air Force's Rome Laboratory, and was monitored by Northrup Fowler iii, RL (C3C), 525 Brooks Rd., Gri ss AFB, NY 13441-4505, USA. The United States Government is authorised to reproduce and distribute reprints of this paper for government purposes notwithstanding any copyright notation hereon. iv Abbreviations The following abbreviations are used within the report. This section serves as a reminder of their meaning wherever the context is not clear. ads Associated Data Structure { the level of data structure in O-Plan at which a plan is represented. This is \associated" with an underlying Time Point Network (tpn). am O-Plan Agenda Manager { one of the main processes of the O-Plan system and the main part of the \Controller" which decides on what can be processed next in an O-Plan agent. arpa Advanced Research Projects Agency { earlier called darpa, the Defense Advanced Research Projects Agency. arpi arpa/Rome Laboratory Planning Initiative { the Knowledge-based Planning and Scheduling Initiative research and development programme. cpe Common Prototyping Environment { a shared framework of tools and domain information used within the arpi. coa Course of Action { military terminology for a particular plan option for soem given task and assuming certain constraints. dm O-Plan Database Manager { one of the main processes of the O-Plan system which manages the plan state and gives access to it on behalf of other modules. gop Graph Operation Processor { a support routine in O-Plan used to manipulate information in graphs or networks, e.g., in the Time Point Network (tpn). gost Goal Structure Table { used to hold conditions associated with a plan and their method of satisfaction. ifd Integrated Feasibility Demonstrator { used to demonstrate arpi technologies on military relevant problems. im O-Plan Interface Manager { one of the main processes of the O-Plan system which manages inter-module, inter-agent and user communications. Issues, Nodes, Orderings, Variables, Auxiliary Constraints Model { used to represent constraints on activity or plans. kp O-Plan Knowledge Source Platform { one of the main processes of the O-Plan system on which Knowledge Sources can be run. ks Knowledge Source { a computational capability in O-Plan. ksf Knowledge Source Framework { a proposed language for describing an agent's capabilities (it's Knowledge Sources). v mtc Modal Truth Criterion { another name adopted by other researchers for a process similar to Question Answering (qa). neo Non-combatant Evacuation Operations { military operations to evacuate civilians from a danger zone. pmo Plan Modi cation Operator { a term used to describe the abstract operation of O-Plan in which partially-speci ed plans are modi ed by \Operators" during the search for a solution to a given task. pmos correspond to Knowledge Sources in O-Plan. psv Plan State Variable { an object in a plan which is not fully de ned. psvm Plan State Variables Manager { the Constraint Manager in O-Plan which looks after Plan State Variables (psvs). precis Planning, Reactive Execution and Constraint Satisfaction domain { an experimental application domain to allow demonstration and evaluation of systems for planning, scheduling, constraint satisfaction and reactive plan execution. This domain involved neos from the ctional island of Paci ca. qa Question Answering { the O-Plan support routine which nds the ways in which a plan condition can be satis ed. rea Reactive Execution Agent { an agent designed to support the execution of plans where reaction to changing circumstances is required. rue Resource Utilisation Entry { the form of constraint information looked after by the Resource Utilisation Manager (rum). rum Resource Utilisation Manager { a constraint manager which looks after resource constraint information. tie Technology Integration Experiment { an experiment to join together two or more technologies from the arpi to evaluate some given objective. tf Task Formalism { the domain description language for the O-Plan planner. tgm tome/gost Manager { the Constraint Manager in O-Plan which looks after e ects and conditions. tome Table Of Multiple E ects { used to hold e ects associated with a plan. tpn Time Point Network { used to hold time points associated with a plan and constraints between these time points. tpnm Time Point Network Manager { the Constraint Manager in O-Plan which builds and looks after the tpn. vi