Automated Planning and Scheduling is a historical and active branch of Artificial Intelligence that concerns the realization of strategies or action sequences, typically for execution by intelligent agents, autonomous robots and unmanned vehicles.
The 9th Italian Workshop on Planning and Scheduling (IPS-2021) aims at bringing together researchers interested in different aspects of planning and scheduling problems, and to introduce new researchers to the community. The workshop welcomes both theoretical and practical contributions in any aspect of automated planning and scheduling, with the purpose of encouraging the italian research community to share ideas and new trends on the field.
Although the primary target of this series of workshops is the Italian community of P&S, the aim is to attract an international gathering, thus expecting contributions and participations from around the world. Submissions to this event are here solicited. Each contribution will be reviewed by members of a strong international Program Committee. Original and already published papers will be made available via this workshop web-site.
An Interactive System for Mission Activity Planning
David E. Smith
Activity planning for many spacecraft and planetary missions is a highly interactive process. The users put activities into the plan and want to see the consequences – the processes that get invoked, and the states and resources over time. What makes this interesting is that for much of the planning process, the plan is incomplete and often invalid – it may have unsatisfied conditions, unsatisfied constraints, conflicting action effects, and violated variable bounds. Furthermore, throughout this planning process, the users may change their minds about what objectives to achieve, the order in which to do them, and what their preferences are. The planing problem itself is therefore under-specified, over-subscribed, and constantly evolving.
The Planning and Scheduling Group at NASA Ames Research Center has recently embarked on a project to build a next-generation mission planning system with the ability to facilitate this kind of interactive plan creation and evolution. The language is quite rich and allows the description of temporal activities and processes with continuous change and uncertainty, and complex constraints such as duty-cycle limitations. Key components of the system are: a web-based user interface for plan display and modification; a permissive simulator that constructs state and resource chronicles for a partial plan, and reports violations; and a plan repair module that attempts to fix violations in the plan by moving activities around, and inserting new activities to achieve unsatisfied conditions. I will describe the objectives, challenges, and approach of this project, the rich language used to describe activities, processes, and constraints, and go into some technical detail about some of the components, including the simulator, violation detection, and plan repair.
About the speaker
David E. Smith received his Ph.D. in AI from Stanford University way back in the days of the expert systems boom. However, he resisted the allure of expert systems startups and big bucks, and instead found “true religion” working on AI planning. He spent time as a Research Associate at Stanford University, Research Scientist at the Rockwell Palo Alto Science Center, and Visiting Scholar at the University of Washington before joining NASA Ames Research Center in 1997. He served as lead of the Planning and Scheduling Group there for six years before abdicating the throne and going back to scribbling incomprehensible stuff on whiteboards. He retired from NASA in 2018, but doesn’t seem to really understand the meaning of the term. He is currently serving as Secretary-Treasurer of AAAI, is tormenting several Ph.D. students at Arizona State and Kings College London with his ideas on explainable planning and plan explanation, and got sucked into consulting on this Mission Activity Planning project, where he keeps pushing the project into areas that are way too difficult to solve.