Collaborative motion planning for multi-agent systems

Abstract

The main theme of this dissertation is autonomous navigation of multi-agent systems, with focus on entertainment applications.The goals are to study methods for safe navigation in cluttered and dynamic environments among many decision-making agents, and to explore how a designer or artist could seamlessly design the motion characteristics of a large team of mobile robots.Beside these two main results, methods for real-time interaction with a swarm of robots are also described.Novel algorithms are proposed and extensive experiments are performed with a large team of small pixel robots, robotic wheelchairs and quadrotor helicopters.This work paves the road towards multi-robot applications as well as autonomous navigation in dynamic environments.From a technological standpoint, a system formed by up to one hundred small mobile robots is also described, creating a novel display in which each pixel is a mobile robot of controllable color.Methods for real-time control are described that result in collision-free trajectories and that scale well to very large groups of agents.