PhD Thesis Defense

Title: Compressive Cooperative Obstacle Mapping in Mobile Networks

By: Mr. Alejandro Gonzalez Ruiz

Advisor: Dr. Yasamin Mostofi

Date: Aug 9th 2012, 1:00 PM

Location: ECE, Room 118

Mobile intelligent networks can play a key role in many different areas from emergency response, surveillance and security, and battlefield operations to smart homes and factories and environmental monitoring. Accurate mapping of the obstacles/objects in the environment is key to the robust operation of unmanned autonomous networks as it is an integral part of navigation, motion planning, surveillance and environment monitoring. In the robotics community, the problem of mapping has been widely explored. However, in the existing mapping approaches, only areas that can be directly sensed by the sensors are mapped. In several scenarios, it can be necessary to have see-through capabilities and map the objects without direct sensing. For instance, the robots may need to build an understanding of the objects occluded inside a room, before entering it. Having see-through capabilities also allows the robots to detect the obstacles for navigation purposes, without direct sensing, resulting in a reduced amount of time and energy spent on mapping.

This dissertation aims to provide a framework for wireless-based through the wall mapping. We consider a mobile robotic network that is tasked with building a map of the objects/obstacles in an environment. We propose a novel framework that enables the robots to build the map non-invasively and based on a small number of wireless channel measurements. This allows the robots to efficiently map areas of the workspace that can not be directly seen by sensors such as laser scanners or sonar. By using the recent results in the area of compressive sensing, we show how a group of robots can exploit the sparse representation of the map in space, wavelet or spatial variations, in order to build it with minimal sensing. We discuss three mapping strategies based on frequency sampling, coordinated space and random space measurements and show the underlying tradeoffs of the possible sampling, sparsity and reconstruction techniques using both simulation and experimental results. We furthermore discuss the optimum number of angular motion directions of the robots, as well as the choice of the angles, to distribute a given number of wireless measurements. We establish that the total number of available channel measurements should be distributed along a small number of angles, that is bigger than or equal to the number of jump (discontinuity) angles of the structure, with a preference given to the angles of jumps.

After laying the foundations of our wireless-based mapping framework, we show how to integrate it with existing mapping techniques such as occupancy grid mapping. We propose an integrated framework that uses laser measurements to map the visible parts of the environment and also uses wireless measurements to map the occluded parts. We furthermore propose an adaptive exploration strategy which enables a pair of robots to efficiently collect wireless measurements that improve the see-through performance considerably.

Finally, we also show how to design an experimental robotic platform in order to implement the proposed approach. Most importantly, we show the performance of our framework in efficiently mapping a number of real obstacles (including blocked ones) on our campus. Our experimental results show the feasibility of the proposed frameworks for mapping structures which include occluded parts, in realistic fading environments.