Skip to main content
eScholarship
Open Access Publications from the University of California

UC Riverside

UC Riverside Electronic Theses and Dissertations bannerUC Riverside

Multi-Robot Cooperative Localization and Target Tracking

Abstract

Sensor networks with the ability of communication and perception has a wide range of applications. They can be utilized to estimate a target's pose even if some sensors are blind to the target. This problem is termed as distributed state estimation (DSE) which has been widely studied. However, existing works are limited to 2-D scenarios with the assumption of fixed and known sensor states. This manuscript addresses the limitations and extends DSE to the mobile robot case where the robots use onboard sensors to track the target's state.

In particular, in Chapter 2 we study the problem of joint localization and target tracking (JLATT) in 2-D situations. A team of robots simultaneously localize themselves and track multiple targets. Instead of treating localization and target tracking as two separate problems, we explicitly account for the influence of one to the other and exploit it to improve performance in a distributed context. We introduce a fully distributed algorithm that is applicable to generic robot motion, target process and measurement models, and is robust to time-varying sensing and communication typologies.

In the following three chapters, we work on the 3-D scenarios and use the most popular sensor rig -- the visual-inertial sensor. Specifically, we first focus on the target tracking and robot localization separately and then work on the visual-inertial JLATT. In chapter 3, a static camera network is used to cooperatively estimate the six degree-of-freedom (6-DoF) pose of a moving object. A novel distributed Kalman filter (DKF) is introduced for a general nonlinear system. In chapter 4, we present a multi-robot visual-inertial navigation system (VINS) which achieves cooperative localization (CL) by efficiently fuses environmental features. The algorithm enables drift-free estimation through the use of loop-closure constraints to other robots’ historical poses without a significant increase in computational cost. Finally, in chapter 5, we present an algorithm to track a target’s state by utilizing a heterogeneous robot network. Rather than assuming a known common global frame for all the robots, we allow each robot to perform motion estimation locally. For localization, one robot builds a prior map and then the map is used to bound the long-term drifts of the visual-inertial odometry (VIO) running on the other robots. The novel DKF is employed to track the pose of the object which is represented as a point cloud.

The research presented in this dissertation aims at extending the application of multi-robots by improving the performance in self-localization and target tracking. The proposed algorithms are demonstrated in Monte-Carlo simulations and experiments.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View