Reasoning and Learning in Real-World Systems for Long-Term Autonomy
(LTA 2018)

AAAI 2018 Fall Symposium

October 18-19, 2018

Arlington, VA, U.S.A.


  • August 2018 - Papers have been accepted: 12 long and 6 short!
  • July 2018 - Invited speakers include Nick Hawes, Tom Wagner, and Peter Wurman!
  • July 2018 - Nissan Research Center Silicon Valley is sponsoring the symposium!
  • June 2018 - Program committee has been formed.
  • May 2018 - The LTA 2018 website has been launched.


Over the past decade, decision-making agents have been increasingly deployed in industrial settings, consumer products, healthcare, education, and entertainment. The development of drone delivery services, virtual assistants, and autonomous vehicles have highlighted numerous challenges surrounding the operation of autonomous systems in unstructured environments. This includes mechanisms to support autonomous operations over extended periods of time, techniques that facilitate the use of human assistance in learning and decision-making, learning to reduce the reliance on humans over time, addressing the practical scalability of existing methods, relaxing unrealistic assumptions, and alleviating safety concerns about deploying these systems.

This symposium aims to identify the challenges and bridge the gaps between theoretical frameworks for planning and learning in autonomous agents and the requirements imposed by deployment in the real world. Our goal is to help identify research avenues that can move the AI community beyond highly theoretical results for simple domains or highly engineered one-shot solutions for realistic applications. We seek papers that find a common middle ground between theory and applications, and analyze the lessons learned from these efforts, particularly with respect to long-term autonomy.

The symposium combines invited talks, presentations, and discussions from both an AI and a robotics perspective.


Topics of interest include, but are not limited to, the following:

  • Decision-making representations, models, and algorithms for the real world
  • Hierarchical and multi-objective solutions for scalable planning and learning
  • Efficient integrations of task and motion planning
  • Integrating planning, reasoning, and learning for long-term deployments
  • Safety in real-world decision-making and learning
  • Scalable multiagent and human-in-the-loop techniques
  • Proactively incorporating human feedback in decision-making
  • Leveraging the complimentary capabilities of humans and robots in real-world tasks
  • Evaluation metrics for long-term autonomy
  • Case studies and descriptions of deployed autonomous systems
  • Lessons learned from deployed applications of autonomous systems

Important Dates

Paper Submission Deadline:
Tuesday, July 31, 2018
Notification of Acceptance:
Tuesday, August 14, 2018
Camera-Ready Submission:
Friday, September 14, 2018
Symposium Date:
Thursday, October 18, 2018 to
Friday, October 19, 2018

Invited Talks

Nick Hawes Speaker: Dr. Nick Hawes, Associate Professor of Robotics at the University of Oxford
Title: Learning From Four Years of Mobile Autonomy
Abstract: In this talk I will look back over four years of long-term deployments of autonomous mobile robots in everyday environments. From this I will present examples of the kinds of things that mobile robots can learn over long autonomous operations in such environments, including navigation information, human activities, object models, and mission schedules. Following this I will explore the issues (software, hardware, and social) that impacted upon the autonomy of our deployed robots, and look at what we can learn from these experiences as both AI practitioners and as engineers deploying robots in real environments.

Tom Wagner Speaker: Dr. Tom Wagner, CEO of Berkshire Grey, Former CTO of iRobot
Title: Challenges and Opportunities for Long-Term Autonomy in Deployed Systems
Abstract: Artificial Intelligence applied to both virtual agents and physical agents (robotics) is transforming how work is performed in cyberspace and in the physical world. Joining technology often considered “research” with real-world deployed systems opens the door to questions about what can and should happen with such systems over time. Should such systems change over time? Should they not? Are there situations where the work task cannot be performed unless the systems adapt and change as they operate? What are the expectations and desires of the stakeholders in these circumstances and what are the implications of embedding reasoning, learning, planning, etc., subsystems in commercial applications? In this talk we will walk through example applications based on experiences with industrial systems, consumer systems, and military systems. Via these examples and their associated AI elements, together we will explore issues such as scoping, challenges, opportunities, and expectations which may help further application of our technologies to such problems.

Peter Wurman Speaker: Dr. Peter Wurman, VP of Engineering at Cogitai, Former Co-founder of Kiva Systems
Title: How Kiva Robots Disrupted Warehousing
Abstract: Kiva Systems introduced swarms of agile robots into an industry dominated by stationary conveyor systems. The path from concept through successful startup and eventual acquisition involved challenges on all fronts. In this talk I’ll explain the business problem that motivated the innovation, Kiva technology and the benefits it brought to customers, and the future of applications of robotics in warehouses.

Accepted Papers

Partial Policy Re-use in Connected Health Systems
Matthew Saponaro and Keith Decker

Learning and Generalisation of Primitives Skills Towards Robust Dual-arm Manipulation
Èric Pairet, Michael Mistry, and Frank Broz

Towards Robust Grasps: Using the Environment Semantics for Robotic Object Affordances
Paola Ardón Ramírez, Subramanian Ramamoorthy, and Katrin Solveig Lohan

Behavior Modeling for Autonomous Driving
Aniket Bera and Dinesh Manocha

SOMA: A Framework for Understanding Change in Everyday Environments Using Semantic Object Maps
Lars Kunze, Hakan Karaoguz, Jay Young, Ferdian Jovan, John Folkesson, Patric Jensfelt, and Nick Hawes

Risk-Aware Planning by Extracting Uncertainty from Deep Learning-Based Perception
Maymoonah Toubeh and Pratap Tokekar

SocialAnnotator: Annotator Selection Using Activity and Social Context
H. M. Sajjad Hossain and Nirmalya Roy

Deep CNN and Probabilistic DL Reasoning for Contextual Affordances
Hazem Abdelkawy, Sandro Rama Fiorini, Abdelghani Chibani, Naouel Ayari, and Yacine Amirat

Predictions, Surprise, and Predictions of Surprise in General Value Function Architectures
Johannes Günther, Alex Kearney, Michael R. Dawson, Craig Sherstan, and Patrick M. Pilarski

Multi-Fidelity Model-Free Reinforcement Learning with Gaussian Processes
Varun Suryan, Nahush Gondhalekar, and Pratap Tokekar

Towards Perception Aware Task-Motion Planning
Antony Thomas, Sunny Amatya, Fulvio Mastrogiovanni, and Marco Baglietto

Evaluating Predictive Knowledge
Alex Kearney, Anna Koop, Craig Sherstan, Johannes Günther, Richard Sutton, Patrick Pilarski, and Matthew Taylor

LAAIR: A Layered Architecture for Autonomous Interactive Robots
Yuqian Jiang, Nick Walker, Minkyu Kim, Nicolas Brissonneau, Daniel S. Brown, Justin W. Hart, Scott Niekum, Luis Sentis, and Peter Stone

Big Data and Deep Learning Models for Automatic Dependent Surveillance Broadcast (ADS-B)
Ying Zhao, Richard Wu, Matthew Xi, Andrew Polk, and Tony Kendall

From Abstract to Executable Models for Multi-Agent Path Finding on Real Robots
Roman Barták, Jiří Švancara, Věra Škopková, and David Nohejl

Using hierarchical expectations grounded in perception for failure reasoning during task execution
Priyam Parashar, Henrik I. Christensen, and Ashok K. Goel

A Practical Distributed Knowledge-Based Reasoning and Decision-Theoretic Planning for Multi-robot Service Systems
Abdel-Illah Mouaddib and Laurent Jeanpierre

Policy Networks for Reasoning in Long-Term Autonomy
Kyle Hollins Wray and Shlomo Zilberstein

Paper Presentations

Papers will be presented either in a long format (25 minute) or short format (15 minute) based on paper length and relevance to long-term autonomy (LTA).

To focus the discussion on the symposium's topic, all presentations should reserve 3-5 minutes at the end of their talk to relate their paper directly to long-term autonomy. They can provide their brief perspectives on the topic if appropriate as well.

Thus, long presentations should be 15 minutes of content, 5 minutes of relations to LTA, and 5 minutes of Q&A. Short presentations should be 10 minutes of content, 3 minutes of relations to LTA, and 2 minutes of Q&A.

Panel Discussion

The panel discussion will conclude the symposium with reflections and perspectives from the invite talks and papers on the state of long-term autonomy overall.

Moderator: Joydeep Biswas, University of Massachusetts Amherst

The panel members include: TBD.

Program Schedule

Thursday, October 18

9:00 am - 10:30 am Session #1 - Paper Presentations: 3 Long (25 min), 1 Short (15 min)
10:30 am - 11:00 am Coffee Break
11:00 am - 12:00 pm Invited Talk #1 - Dr. Nick Hawes
12:00 pm - 1:30 pm Lunch Break
1:30 pm - 3:30 pm Session #2 - Paper Presentations: 4 Long (25 min), 1 Short (15 min)
3:30 pm - 4:00 pm Coffee Break
4:00 pm - 5:00 pm Invited Talk #2 - Dr. Tom Wagner
6:00 pm - 7:00 pm AAAI Symposia Reception

Friday, October 19

9:00 am - 10:30 am Session #3 - Paper Presentations: 3 Long (25 min), 1 Short (15 min)
10:30 am - 11:00 am Coffee Break
11:00 am - 12:00 pm Invited Talk #3 - Dr. Peter Wurman
12:00 pm - 1:30 pm Lunch Break
1:30 pm - 3:30 pm Session #4 - Paper Presentations: 4 Long (25 min), 1 Short (15 min)
3:30 pm - 4:00 pm Coffee Break
4:00 pm - 5:00 pm Panel Discussion
6:00 pm - 7:00 pm AAAI Plenary Session

Saturday, October 20

This symposium will not hold anything on the final half-day.



Program Committee:


Please direct questions regarding the symposium to "longtermautonomy2018 'at'".