Session: Research Posters

Paper Number: 173219

Experimental Cluster Space Control of Four Unmanned Aerial Systems 

This poster will provide an overview of the cluster space control methodology, define the tetrahedron formation’s cluster space geometry, describe the testbed, and present initial results. The Robotics Systems Laboratory (RSL) at Santa Clara University has developed the cluster space control framework, which describes multi-robot formations as a single virtual mechanism with attributes such as position, orientation, and geometry. The framework provides an operator or a higher-level controller a convenient interface through which to dictate a formation’s motion and geometry. The technique is general, and can be applied to land, marine, or aerial multi-robot systems. The RSL has demonstrated the cluster space technique extensively for planar formations, via ground vehicles on land and surface vehicles on water. However, cluster space’s application to 3-dimensional formations, which are of interest for marine and aerial multi-robot systems, has been more limited. This work implements and demonstrates experimental cluster space control of a 3-dimensional formation: four quadcopters in a tetrahedron-shaped formation. 

The implementation is built on top of a testbed developed by collaborators at Lawrence-Livermore National Laboratory. The testbed consists of commercially available and open-source hardware and software, enabling replication. Each quadcopter is fitted with a flight controller, a GPS unit, and a barometer. The flight control software, ArduPilot, provides position control and sensor fusion for localization. Telemetry and commands are communicated via Wi-Fi radio. A simple proportional control law is used. Control was first verified using software-in-the-loop simulation provided by the ArduPilot project, followed by experiments using the testbed. 

Experiments were performed in a flight area roughly 75 meters wide and 85 meters long; altitude was limited to 50 meters. The cluster had 15-meter side lengths. Preliminary results show satisfactory control. In the first set of tests, the cluster space variables were controlled independently (i.e. varying one variable’s setpoint while regulating the others). Though there are 24 cluster space pose variables defining the formation’s pose, 12 define the orientations of individual robots, which are not of immediate interest. (Indeed, in the case of quadcopters, pitch and roll cannot be controlled independently of position). The remaining 12 cluster space variables define cluster origin location, cluster orientation, side lengths, and internal angles. Individual tests for 10 of the 12 cluster space variables of interest have been completed. In each, the controller was able to maintain a constant error offset for a linear ramp input to the variable being tested. As expected, the variables not being tested were reasonably regulated to their initial values. Additionally, a similar test was conducted in which the side lengths were changed in concert, i.e., resizing the cluster. Future tests include tests for the final two cluster space variables (internal angles), and further tests commanding setpoints for multiple variables simultaneously. Pending these future tests, successful cluster space control of a 3-dimensional formation will be demonstrated.

Presenting Author: Caleb Sparks Santa Clara University

Presenting Author Biography: Caleb Sparks received a B.S. degree in mechanical engineering from the Georgia Institute of Technology. He is currently pursing a M.S. in mechanical engineering at Santa Clara University as a part of Santa Clara's Robotic Systems Laboratory.

Authors:

Caleb Sparks Santa Clara University
Christopher Kitts Santa Clara University
Michael Neumann Santa Clara University