(Georgia Institute of Technology, 2012-10)
Dantam, Neil; Essa, Irfan; Stilman, Mike
We demonstrate the automatic transfer of an
assembly task from human to robot. This work extends efforts
showing the utility of linguistic models in verifiable robot
control policies by now performing real visual analysis of
human demonstrations to automatically extract a policy for the
task. This method tokenizes each human demonstration into a
sequence of object connection symbols, then transforms the set
of sequences from all demonstrations into an automaton, which
represents the task-language for assembling a desired object.
Finally, we combine this assembly automaton with a kinematic
model of a robot arm to reproduce the demonstrated task.
(Georgia Institute of Technology, 2012)
Dantam, Neil; Essa, Irfan; Stilman, Mike
We describe several algorithms used for the inference of linguistic robot policies from human
demonstration. First, tracking and match objects using the Hungarian Algorithm. Then, we convert Regular Expressions to Nondeterministic Finite Automata (NFA) using the McNaughton-Yamada-Thompson
Algorithm. Next, we use Subset Construction to convert to a Deterministic Finite Automaton. Finally, we
minimize finite automata using either Hopcroft's Algorithm or Brzozowski's Algorithm.
(Georgia Institute of Technology, 2011)
Dantam, Neil; Stilman, Mike
We present a new Inter-Process Communication (IPC) mechanism and library. Ach is uniquely suited for coordinating perception, control drivers, and algorithms in real-time systems that sample data from physical processes. Ach eliminates the Head-of-Line Blocking problem for applications that always require access to the newest message. Ach is efficient, robust, and formally verified. It has been tested and demonstrated on a variety of physical robotic systems. Finally, the source code for Ach is available under an Open Source BSD-style license.
(Georgia Institute of Technology, 2010)
Dantam, Neil; Stilman, Mike
We present the Motion Grammar: a novel unified representation
for task decomposition, perception, planning, and hybrid control
that provides a computationally tractable way to control robots in
uncertain environments with guarantees on completeness and correctness.
The grammar represents a policy for the task which is
parsed in real-time based on perceptual input. Branches of the syntax
tree form the levels of a hierarchical decomposition, and the
individual robot sensor readings are given by tokens. We implement
this approach in the interactive game of Yamakuzushi on a
physical robot resulting in a system that repeatably competes with
a human opponent in sustained game-play for matches up to six
minutes.