diff --git a/content/program/tutorials/2024_t01_aiopt_scheduling.md b/content/program/tutorials/2024_t01_aiopt_scheduling.md index 11bcf4a..b44fa5c 100644 --- a/content/program/tutorials/2024_t01_aiopt_scheduling.md +++ b/content/program/tutorials/2024_t01_aiopt_scheduling.md @@ -3,8 +3,9 @@ title: "Tutorial 01 - AI and Optimization for Scheduling" date: 2024-04-16 draft: false --- +# Tutorial 01 - AI and Optimization for Scheduling -# Abstract +## Abstract Scheduling problems arise in various areas, including business, engineering, healthcare, and others. In this tutorial, we will first present several scheduling problems and case studies from various application domains, such as project @@ -18,7 +19,7 @@ demonstrate the application of these techniques in several real-world domains. - [Tutorial Website](xxxx) -# About the Presenters +## About the Presenters **Nysret Musliu** is an Associate Professor and the Head of the Christian Doppler Laboratory for AI and Optimization for Planning and Scheduling at TU Wien. His research focuses on problem solving and search in artificial intelligence, diff --git a/content/program/tutorials/2024_t02_finding_multiple_plans.md b/content/program/tutorials/2024_t02_finding_multiple_plans.md index a3b1a2b..2b4e1e1 100644 --- a/content/program/tutorials/2024_t02_finding_multiple_plans.md +++ b/content/program/tutorials/2024_t02_finding_multiple_plans.md @@ -4,58 +4,23 @@ date: 2024-04-16 draft: false --- -# Abstract +# Tutorial 02 - Finding Multiple Plans for Classical Planning Problems -The goal of the tutorial is to familiarise the audience with the theorethical and practical +## Abstract + +The goal of the tutorial is to familiarise the audience with the theoretical and practical aspects of devising multiple plans for classical planning problems. We will motivate the need for such planners from application perspective, formally define the respective computational problems, as well as describe the existing approaches to solving these problems. We will finish with a hands-on session on using the existing planners. +### Target Audience + The tutorial is aimed at ICAPS audience, familiar with classical planning. We hope that the tutorial will help both the practitioners who want to use these tools in applications and the tools builders interested in creating efficient planners. -# Official Website and Auxiliary Materials +## Official Website and Authors' Details - [Tutorial Website](https://mp-tutorial.github.io/) -# About the authors - -**Shirin Sohrabi** is a principal research scientist and research manager at IBM T.J. Watson -Research Center in New York, USA. She has over 12 years of research experience -in AI Planning in the industry and has co-authored more than 50 publications. She -served as program co-chair of ICAPS 2020, Novel Application Track co-chair of ICAPS -2018-2019, Journal Track co-chair of ICAPS 2022, and has co-chaired three editions -of SPARK and two editions of the FinPlan workshop. She has co-organized IJCAI -2021 and AAAI 2022 tutorials on “AI Planning: Theory and Practice”. She regularly -serves on the Senior Program Committees of ICAPS, IJCAI, and AAAI. She is an -ACM and AAAI senior member. She is a member of ICAPS executive council and -also serves as the diversity and inclusion chair. - -**Michael Katz** is a principal researcher at IBM T.J.Watson Research Center, NY, USA. -His primary research interest is in domain-independent planning and in the integration -of planning and RL. He received his PhD for a dissertation on heuristics for domainindependent -planning from Technion in 2010. He was a co-organizer of the 2011, -2013, 2014, 2015, 2016, 2018, and 2022 H(S)DIP workshops, as well as of five of the -six editions (including the first) of PRL workshop 2020-2023. He was a program chair -of ICAPS 2021. He co-organized IJCAI 2021 and AAAI 2022 tutorials AI Planning: -Theory and Practice. He co-authored 50+ publications in various premier research -conferences, including AAAI, ICAPS, IJCAI, IROS, and journals like JAIR and AIJ. - -**Junkyu Lee** is a research scientist at IBM T.J. Watson Research Center, NY, USA. -He received his PhD at University of California, Irvine in 2020, and had over 6 years -of experience in systems engineering in the industry. His primary research interest -is causal reasoning and graphical model algorithms for sequential decision making -and augmenting foundation models with reasoning capability. He co-authored over -10 publications in AAAI, UAI, IJCAI, NeuRIPS, and regularly serves as a Program -Committee of AAAI, IJCAI, NeuRIPS, ICML, ICLR, and others. - -**David Speck** is a postdoctoral researcher at Link¨oping University in Sweden, specializing -in domain-independent planning. He received his PhD from the University of -Freiburg in Germany in 2022 and has authored 19 peer-reviewed publications at prestigious -conferences such as AAAI, ICAPS, and ECAI. Serving as a program committee -member and reviewer for renowned AI conferences such as AAAI, ICAPS, IJCAI, -and ECAI shows his active involvement in the field. In addition, he co-organized the -HSDIP workshop at ICAPS in 2020 and 2021. Finally, he has won multiple competitions -in the field of AI, including the Probabilistic International Planning Competition -(IPC) in 2018 and the Optimal Track of the IPC 2023. + diff --git a/content/program/tutorials/2024_t03_scikit_decide.md b/content/program/tutorials/2024_t03_scikit_decide.md index 9a09819..79c40bd 100644 --- a/content/program/tutorials/2024_t03_scikit_decide.md +++ b/content/program/tutorials/2024_t03_scikit_decide.md @@ -4,7 +4,9 @@ date: 2024-04-16 draft: false --- -# Abstract +# Tutorial 03 - A Hands-On Tutorial on scikit-decide, the Open-Source C++ and Python Library for Planning, Scheduling and Reinforcement Learning + +## Abstract Scikit-decide is an open-source library for modeling and solving planning, scheduling and reinforcement learning problems within a common API which helps break technical silos between different @@ -23,11 +25,11 @@ The half-day tutorial will show how to model and solve the same problems using a different communities, and how to extend the libraries with new domains and solvers in a few lines of code. It will alternate presentations and live Python coding sessions -# Official Website and Auxiliary Materials +## Official Website and Auxiliary Materials - [Tutorial Website](xxxx) -# About the authors +## About the authors **Florent Teichteil-Königsbuch** is an expert in AI Decision-Making and Combinatorial Optimization in Airbus Central Research and Technology. After graduating as a PhD in Artificial diff --git a/content/program/tutorials/2024_t04_storytelling.md b/content/program/tutorials/2024_t04_storytelling.md index c259553..4358c5c 100644 --- a/content/program/tutorials/2024_t04_storytelling.md +++ b/content/program/tutorials/2024_t04_storytelling.md @@ -4,7 +4,9 @@ date: 2024-04-16 draft: false --- -# Abstract +# Tutorial 04 - Planning for Storytelling + +## Abstract In this tutorial we will demonstrate the role that planning, or planning-based representations, can play in narrative generation methods. The plan becomes the representation of the story and also that of the story world, and thus we use @@ -12,11 +14,11 @@ Planning to create logical, believable, and coherent stories (narratives) in a v techniques, including modern approaches that make use of Large Language Models (LLMs) and provide the opportunity for attendees to play with the technology themselves live. -# Official Website and Auxiliary Materials +## Official Website and Auxiliary Materials - [Tutorial Website](xxxx) -# About the authors +## About the authors **Nisha Simon** is a PhD candidate at Queen’s University under the supervision of Prof. Christian Muise, and a Teaching Assistant with the School of Computing. Her interests are in researching structured learning from unstructured diff --git a/content/program/tutorials/2024_t05_rl_coach.md b/content/program/tutorials/2024_t05_rl_coach.md index 0220c55..6a15e9a 100644 --- a/content/program/tutorials/2024_t05_rl_coach.md +++ b/content/program/tutorials/2024_t05_rl_coach.md @@ -3,43 +3,25 @@ title: "Tutorial 05 - Orchestrating autonomous agents: Reinforcement Learning To date: 2024-04-16 draft: false --- +# Tutorial 05 - Orchestrating Autonomous Agents: Reinforcement Learning for Hierarchical Planning with COACH -# Abstract +## Abstract -In many actual applications, semi-autonomous multi-agent systems need to be controlled and -evaluated by human operators. Consider the routing of autonomous vehicles with potential communication -interference, coordinated search with quad-copters whose on-board systems execute close -quarters navigation faster than goals can be updated, or skill selection for agents engaged in complex -coordinated maneuver in a team based activity - in all of these cases on-agent systems (control based -or deep learning based) execute higher level directed plans in a semi-autonomous fashion, without -high resolution control from a directing agent. Such orchestration style planning problems lie in -a fascinating grey area between reinforcement learning, general planning, and human-on-the-loop -system design. +This tutorial will provide an introduction to [COACH](https://github.com/act3-ace/coach) - a suite of Python tools for +recasting [Gymnasium](https://gymnasium.farama.org/index.html) and +[PettingZoo](https://pettingzoo.farama.org/index.html)-compatible Multi-Agent Reinforcement Learning (MARL) problems as +orchestration-style planning problems. Traditional Reinforcement Learning (RL) focuses on training low level agents to +interact with an environment in a high frequency feedback loop. Once policies have been trained, human direction becomes +an orchestration problem, especially with large numbers of agents. COACH provides tools for researchers to bridge this +gap: given any simulation compatible with Gymnasium or PettingZoo, trained agents can be set up to interface with a +director agent who tackles high level scheduling, policy selection, or coordination for generating autonomously executed +plans. COACH was created in association with the US Air Force Research Laboratory’s Autonomy Capability Team (ACT3). -Unfortunately, the technical challenge of coordinating multiple hierarchical agents across multiple -systems with complex communication schedules makes for a high barrier of entry for practical -research on such orchastration problems. To facilitate research into this area, we will guide tutorial -participants in using open source tools to easily convert their existing simulations into semiautonomous -planning problems. +## Resources -This tutorial will provide an introduction to COACH - a suite of Python tools for recasting -Gymnasium and PettingZoo compatible Multi-Agent Reinforcement Learning (MARL) problems as -orchestration-style planning problems. Traditional Reinforcement Learning (RL) focuses on training -low level agents to interact with an environment in a high frequency feedback loop. Once policies -have been trained, human direction becomes a orchestration problem, especially with large numbers -of agents. +- [Tutorial Website](https://github.com/act3-ace/coach/blob/main/docs/icaps.md) -COACH provides tools for researchers to bridge this gap: given any simulation compatible with -Gymnasium or PettingZoo, trained agents can be set up to interface with a director agent who -tackles high level scheduling, policy selection, or coordination for generating autonomously executed -plans. COACH was created in association with the US Air Force Research Laboratory’s Autonomy -Capability Team (ACT3). - -# Resources - -- [Tutorial Website](xxxx) - -# About the authors +## About the authors **Nate Bade** is an award-winning educator and former teaching professor and program director of the MS in Applied Mathematics (MSAM) program at Northeastern University. He specialized in project based education and designed the diff --git a/content/program/tutorials/tutorial_entry_template.md b/content/program/tutorials/tutorial_entry_template.md index c5ec4e7..2969390 100644 --- a/content/program/tutorials/tutorial_entry_template.md +++ b/content/program/tutorials/tutorial_entry_template.md @@ -1,7 +1,9 @@ -# Abstract +# Tutorial Title -# Resources +## Abstract + +## Resources - [Tutorial Website](xxxx) -# About the authors \ No newline at end of file +## About the authors \ No newline at end of file