Authors:
Eniko T. Enikov
1
and
Juan-Antonio Escareno
2
Affiliations:
1
University of Arizona, United States
;
2
Institut Polytechnique des Sciences Avancées (IPSA), France
Keyword(s):
Micro-Air Vehicles, Artificial Neural Network, Path Planning, Body Schema, Cognitive Robotics.
Related
Ontology
Subjects/Areas/Topics:
Agents
;
Artificial Intelligence
;
Cognitive Robotics
;
Control and Supervision Systems
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Neural Networks Based Control Systems
;
Perception and Awareness
;
Robot Design, Development and Control
;
Robotics and Automation
Abstract:
To date, most autonomous micro air vehicles (MAV-s) operate in a controlled environment, where the location
of and attitude of the aircraft are measured be dedicated high-power computers with IR tracking capability. If
MAV-s are to ever exit the lab and carry out autonomous missions, their flight control systems needs to utilize
on-board sensors and high-efficiency attitude determination algorithms. To address this need, we investigate
the feasibility of using body schemas to carry out path planning in the vision space of the MAV. Body schemas
are a biologically-inspired approach, emulating the plasticity of the animal brains, allowing efficient representation
of non-linear mapping between the body configuration space, i.e. its generalized coordinates and the
resulting sensory outputs. This paper presents a numerical experiment of generating landing trajectories of
a miniature rotor-craft using the notion of body and image schemas. More specifically, we demonstrate how
a tra
jectory planning can be executed in the image space using a pseudo-potential functions and a gradient-based
maximum seeking algorithm. It is demonstrated that a neural-gas type neural network, trained through
Hebbian-type learning algorithm can learn a mapping between the rotor-craft position/attitude and the output
of its vision sensors. Numerical simulations of the landing performance of a physical model is also presented,
The resulting trajectory tracking errors are less than 8 %.
(More)