Distributed Spatial Predictive Formation Control

Abstract

This thesis describes the development of the Delft Center for Systems and Control (DCSC) Networked Embedded Robotics Lab (NER) and a novel distributed formation control algorithm to showcase the lab’s experimental capabilities. The lab environment is built to support ground as well as aerial robotic platforms featuring a netted environment and state-of-the-art camera localization system. Both the hardware and software infrastructure considered in this work were chosen to provide an easy interface for students to run distributed experiments with a focus on code reusability. To this end, the Robot Operating System (ROS) was chosen as a software framework as it provides drivers for most robotics platforms, libraries for common tasks such as localization and navigation, communication between software nodes on different hardware, and its widespread use and acceptance by the robotics community. The developed algorithm to showcase the multi-robot coordination capabilities of the laboratory uses a model predictive based formation controller with a coordinate transform from Euclidean to spatial coordinates where the agents’ position is expressed in terms of traveled path length and path deviation. The method was implemented on multiple iRobot Create platforms with independent computing power. A distributed formation controller was achieved by applying a consensus algorithm on the traveled path length. Results show that the desired formations can be achieved even while tracking a target path. Implementation of the complete experiment in ROS illustrates that the developed laboratory setup and its components (hardware and software architecture) are capable of running distributed robotics experiments.