Jorge Cortés
Professor
Cymer Corporation Endowed Chair
Cooperative dynamic domain reduction
A. Ma, M. Ouimet, J. Cortés
Distributed Autonomous Robotic Systems:
The 14th International Symposium, ed. N. Correll, M. Schwager and M. Otte, Springer Proceedings in Advanced Robotics, vol. 9, Springer, New
York, 2019, pp. 499-512
Abstract
Unmanned vehicles (UxVs) are increasingly deployed in a
wide range of challenging scenarios, including disaster
response, surveillance, and search and rescue. Consider
a scenario where a heterogeneous swarm of UxVs are
tasked with completing a wide variety of types of
objectives that possibly require cooperation from
vehicles of differing capabilities. Our goal is to find
a framework that enables vehicles to distributively and
autonomously aid each other in the services of these
objectives. We approach this problem by extension of
Dynamic domain reduction for multi-agent planning
(DDRP), in which we created a framework that utilizes
model-based hierarchical reinforcement learning and
spatial state abstractions crafted for robotic
planning. We extend DDRP with a design for optimizing
the joint trajectories to allow multiple agents to
coordinate for the completion of cooperative
objectives. We modify of our previous framework and
introduce an algorithm that uses simulated annealing to
find joint trajectories. We call the result of the
modifications and extensions, Cooperative dynamic domain
reduction for multi-agent planning (CDDRP). Our analysis
characterizes long term convergence in probability to
the optimal set of trajectories. We provide simulations
to estimate the performance of CDDRP in the context of
swarm deployment.
pdf
Mechanical and Aerospace Engineering,
University of California, San Diego
9500 Gilman Dr,
La Jolla, California, 92093-0411
Ph: 1-858-822-7930
Fax: 1-858-822-3107
cortes at ucsd.edu
Skype id:
jorgilliyo