January 23, 2015, Webb 1100

The successful deployment of mobile multi-robot systems in tasks such as search and rescue, environmental monitoring, and oceanic exploration depends on the accurate localization of these robots. In these applications, the environment is often uncharted, dynamic, and may not be accessible a priori. Thus, the classical beacon-based localization algorithms or fixed feature-based Simultaneous Localization and Mapping algorithms may not be applicable. Fully or intermittently GPS-denied environments also deprive these applications from exploiting GPS navigation. A technique that can work the best for such multi-robot systems is a Cooperative Localization strategy. This technique uses relative measurements among the robots as a feedback signal to jointly estimate the location of team members, resulting in increased pose accuracy for the entire team. However, as a result of these relative measurement updates, after a while the pose estimates of team members become fully coupled to each other. This coupling creates a great challenge in developing decentralized cooperative localization algorithms that do not require an all-to-all communication at each time-step. In this talk, we present our preliminary work on the design of a novel decentralized recursive cooperative localization algorithm that is a decentralized implementation of a centralized Extended Kalman Filter (EKF) for cooperative localization. In this algorithm, by introducing appropriate intermediate variables, we completely decouple the propagation stage of the EKF. We also enable each robot to update its own pose estimate in a decentralized manner using a communication message received from the robot that has taken the relative measurement at that particular time-step.

Solmaz Kia is an Assistant Professor at the University of California, Irvine. She received her PhD in Mechanical and Aerospace Engineering at the University of California, Irvine in 2009. Her research interests include networked control systems, decentralized/distributed algorithm design for multi-agent systems, nonlinear control theory and robotics. In 2014 she won Best Session Presentation Award at the American Control Conference, and the Univeristy of California President's Postdoctoral Fellowship from 2012-2014.