Long-Term Continual Planning for Remote Human-Robot Teaming in Open Worlds

Project: Research project

Description

2 Project Description HSAP/URAP students will conduct research on the human-robot teaming problem along two themes: A. Information Elicitation and Processing: It is recognized that in a human-robot team, the human holds a lot of information about the scenario that is not known initially to the robot. This is often due to the difficulty of extracting such information and representing it formally before the team is forced into execution. To deal with this problem, issue of communication need to be considered. In particular, the planner needs to be aware of the kinds of utterances and dialogue that might come from the human, as well as the kind of questions that must be asked of the human executive to elicit the information that it needs to continue execution or to generate plans to satisfy a given objective. B. Model Construction: The other aspect of the open world that the team must execute in is the fact that the model used by the planner to represent the robot's capabilities is incomplete in varying degrees. An important part of the human-robot teaming problem is the construction of this model, and testing that model out under various scenarios. Subsequently, updates to the model that was created must also be accommodated. 2.1 Research Issues HSAP/URAP students will investigate the following specific topics within the laboratory setting: (a) Simulation of Instances: Students will simulate various problem instances that are in the mold of the human-robot teaming problem, first in human-only teams as walkthroughs and then using simulators and humanoid NAO Robots from Aldebaran Robotics (that we already have available). (b) Planning for Elicitation: Based on the results of human-only experiments, students will encode the kinds of knowledge that may be expected from a human executive, and various triggers for the planner to request or otherwise explicitly elicit such knowledge. (c) Model Creation: Students will study the NAO Robots from Aldebaran Robotics and attempt to come up with a model of the robot's various effectors, sensors, and actions. This model will then be used with planning systems in order to grant the robots some degree of autonomy when acting in human-robot teams

Description

Project Abstract Long-Term Continual Planning for Remote Human-Robot Teaming in Open Worlds PI: Subbarao Kambhampati An increasing number of military applications demand that humans and robots/machines team and work together remotely over long periods to solve complex problems over open worlds. Examples of such tasks include search and rescue applications, and executive command and guidance of unmanned vehicles. Robots operating in such remote human-robot teams need to engage in goaldirected reasoning with partial models of world and objectives, while responding to state, goal and objective updates that come to them both from a rich, dynamic world as well as from the human commanders that exercise control over them. Typical reactive robotic architectures are inadequate in such scenarios since they come with hard-wired implicit goals. Instead, teaming robots require more explicit planning components that can take new requirements and directives into consideration. Most existing work on decisionmaking for human-robot teams focuses on teams working in proximity that do not have to deal with the challenges of partial and evolving domain models. Similarly, most existing work in automated planning ignores the humans in the loop, and assumes complete knowledge of models and objectives. Finally, pure learning-based approaches that attempt to first learn the complete models before using them are not well suited, as the robot does not have the luxury of waiting until the models become complete. The broad aim of the proposed research is to understand the challenges faced by a planner that guides a robot in such remote teaming scenarios, and to develop effective frameworks for handling those challenges. The challenges stem both from the long-term nature of teaming tasks and the open-world nature of the environment. These in turn demand the ability to deal with incompletely specified models, uncertain objectives, open and dynamically changing worlds, as well as the ability to take continual human instructions (including those that change and/or modify goals and action models). In this research, we propose to address these challenges. Specifically, we propose to undertake research tasks aimed respectively at handling incomplete models through generation of robust plans; handling uncertain objectives and partial preferences through diverse plans and open world conditional goals; and handling continual state, goal and model-updates with the help of commitment and opportunity sensitive replanning. The proposed work leverages and builds on promising preliminary results from the PIs recent research on the state-of-the-art partial satisfaction planning, replanning and temporal planning. It is expected to make fundamental contributions to automated planning as well as decision-making for human-robot teams. C-1

Description

Project Summary HSAP / URAP Research: Enabling Natural Language Model Updates in a Human-Robot Team As the extent and reliability of artificial intelligence (AI) technology grows, tasks that put the most valuable military assets humans in the path of danger are increasingly being sought out for automation via robotic agents. Many of these are teaming scenarios, typically characterized by the lack of complete and clearly demarcated information about the model, goals or setting in which the team must perform their operations. Our main project aims at studying the planning challenges in such human-robot teaming scenarios. This HASP/URAP project requests funds for supporting two High School/Undergraduate students to take part in mentored research experience in the context of this project. The research being proposed here builds on work conducted as part of a previous HSAP/URAP project, in 2013. In that project, a humanoid NAO robot was integrated with an automated planning system, with the aim of producing a fully autonomous agent that could interact with humans and achieve specified goals. The proposed research project can be broken down into three main phases. The first phase will involve the processing of the human operators speech into a PDDL action and its attendant requirements and effects. The second phase will consist of building up a database of actions that can be pulled from and used or deleted based on utterances from the human teammate. The final phase will be using the PDDL actions name in order to give the robots codebase access to the new action on a lower level. Once all three phases are complete, we will create a rigorous test demo for the robot to ensure that it can handle a variety of initial conditions, and dynamic changes to the world. This project will also prepare students to participate in STEM competitions, give presentations, and encourage them to continue their pursuit of an undergraduate program in STEM (for HSAP students), and in graduate programs in the DOD's STEMP for URAP students via fellowship programs such as NDSEG and SMART. The PI has significant track record and experience in mentoring undergraduate students. Two undergraduate students were mentored as part of 2013 URAP/HSAP program, and one of them has since continued his involvement in this research by securing undergraduate research funding from ASU.
StatusFinished
Effective start/end date1/11/131/10/16

Funding

  • DOD-ARMY-ARL: Army Research Office (ARO): $370,370.00

Fingerprint

Robots
Planning
Students
Robotics
Unmanned vehicles
Autonomous agents
Military applications
Processing
Artificial intelligence