In direct response to NASA’s need for perception systems for interior environment monitoring, modeling, and navigation and for operational subsystems that increase robot autonomy, TRACLabs proposes to create a collection of software components that will facilitate sharing of environmental and internal state information to enable safe, efficient navigation and manipulation tasks by heterogeneous robot teams working in a shared workspace. We call this software collection PLUMMRS (A Collection of Plan Ledgers and Unified Maps for Multi-Robot Safety). The goal of PLUMMRS is to provide simple APIs for existing single-agent planning and execution systems to leverage. This will allow previously individualistic robots to be used safely in multi-agent contexts. That is, PLUMMRS can be used by any individual robot in a group of
robots to contribute to, and benefit from, a unified model of not only geometric and semantic perception data but also of expected and currently executing motion- and task-level plan data.
PLUMMRS is not a planning framework, as each robot is expected to have its own “black-box” motion- and task-planning capabilities; however, PLUMMRS can be used by both motion- and task-planning & control systems as (1) an oracle of
shared knowledge, (2) a safety monitor based on the shared knowledge representation, and (3) an arbiter (and eventually a full scheduler) that attempts to loosely coordinate the short-term and long-term desires proposed by all individual robots, which are trying to independently complete their tasks while sharing geometric space and physical resources.
Any NASA application involving at least two colocated robots that interact in collaborative tasks are potentially direct beneficiaries of the results of this project, such as:
Industrial applications involving multiple, heterogeneous robots that require sharing information between them are applicable use cases for this project. Specific examples include: