A collaborative work with the Institute for Collaborative Biotechnologies (ICB) at the University of California Santa Barbara, as well as with Toyon Research (small business performing technology development and defense systems analysis), addresses the development of bio-inspired estimation algorithms. This research explores the application of convex optimization methods to multi-agent systems with the main objective of allowing a group of camera-equipped UAVs to cooperatively estimate the position of target agents moving on ground. The solution proposed is based on a sum of norms minimization, which makes it appropriate for dealing with problems encountered in vision-based sensing and multi-agent systems e.g., impulsive noise, disturbances, and measurement outliers.

The picture on the left shows a target tracking scenario considering a pair of UAVs (labeled UAV 1 and UAV 2) cooperatively tracking a target vehicle (labeled TARGET) moving on ground, with respect to a reference coordinate system (labeled AGENT 0). The estimation problem can be represented by a graph where the nodes correspond to the states xi, i={0, 1, 2, 3} of the four agents AGENT 0, UAV 1, UAV 2, TARGET at 5 consecutive time instants {t0, t1, t2, t3, t4}. The edges of the graph correspond to measurements (dashed arrows) and models contraints (dash-dotted arrows). This example highlights the ability to represent heterogenous sensing: at times t1, t4 both UAVs acquire relative measurements between their positions and the target, at time t0 only UAV 2 can “see” the target, and at time t3 none of the UAVs “sees” the target; the UAVs only “see” each other at times t1, t2, t4.

This method was tested in successful field tests in April, July, and November 2012, at the Center for Interdisciplinary Remotely-Piloted Aircraft Studies (CIRPAS) in Camp Roberts, California.

This research addresses the problem of enabling a quad-rotor UAV to perform the tasks of hover flight and translational velocity regulation with the main objective of allowing the vehicle to navigate autonomously. For this purpose, a vision system has been implemented in order to estimate the vehicle's relative altitude, lateral position, and forward velocity during flights. It is shown that, using visual information, it is possible to develop control strategies for different kinds of flying modes, such as hover flight and forward flight at constant velocity. A hierarchical control strategy is developed and implemented, and the local stability of the controller is also proven. Real-time experimental results consisting of autonomous hover and forward flight at constant velocity were successfully achieved, validating the proposed visual algorithm and control law.

Publications on this work can be found here.

This research addresses the problem of road following using a quad-rotor equipped with an imaging system. The main objective consists of estimating and tracking a road without a priori knowledge of the path to be tracked, as well as obtaining efficient controllers for dealing with situations when the road is not detected in the camera's image. For this purpose, two operational regions are defined: one for the case when the road is detected, and the other for when it is not. Switching between the measurements of imaging and inertial sensors enables estimation of the required vehicle's parameters in both regions. Also, for dealing with both aforementioned cases, a switching control strategy which stabilizes the vehicle's lateral position is proposed. The system's stability is proved not only in the two regions, but also in the switching boundaries between them. The performance of the switching control is tested in real time experiments, successfully demonstrating the effectiveness of the proposed approach.

Publications on this work can be found here.