Use this URL to cite or link to this record in EThOS:
Title: Multi-robot collaborative visual navigation with micro aerial vehicles
Author: Williams, Richard Michael
Awarding Body: University of Liverpool
Current Institution: University of Liverpool
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Micro Aerial Vehicles (MAVs), particularly multi-rotor MAVs have gained significant popularity in the autonomous robotics research field. The small size and agility of these aircraft makes them safe to use in contained environments. As such MAVs have numerous applications with respect to both the commercial and research fields, such as Search and Rescue (SaR), surveillance, inspection and aerial mapping. In order for an autonomous MAV to safely and reliably navigate within a given environment the control system must be able to determine the state of the aircraft at any given moment. The state consists of a number of extrinsic variables such as the position, velocity and attitude of the MAV. The most common approach for outdoor operations is the Global Positioning System (GPS). While GPS has been widely used for long range navigation in open environments, its performance degrades significantly in constrained environments and is unusable indoors. As a result state estimation for MAVs in such constrained environments is a popular and exciting research area. Many successful solutions have been developed using laser-range finder sensors. These sensors provide very accurate measurements at the cost of increased power and weight requirements. Cameras offer an attractive alternative state estimation sensor; they offer high information content per image coupled with light weight and low power consumption. As a result much recent work has focused on state estimation on MAVs where a camera is the only exteroceptive sensor. Much of this recent work focuses on single MAVs, however it is the author's belief that the full potential and benefits of the MAV platform can only be realised when teams of MAVs are able to cooperatively perform tasks such as SaR or mapping. Therefore the work presented in this thesis focuses on the problem of vision-based navigation for MAVs from a multi-robot perspective. Multi-robot visual navigation presents a number of challenges, as not only must the MAVs be able to estimate their state from visual observations of the environment but they must also be able to share the information they gain about their environment with other members of the team in a meaningful fashion. The meaningful sharing of observations is achieved when the MAVs have a common frame of reference for both positioning and observations. Such meaningful information sharing is key to achieving cooperative multi-robot navigation. In this thesis two main ideas are explored to address these issues. Firstly the idea of appearance based (re)-localisation is explored as a means of establishing a common reference frame for multiple MAVs. This approach allows a team of MAVs to very easily establish a common frame of reference prior to starting their mission. The common reference frame allows all subsequent operations, such as surveillance or mapping, to proceed with direct cooperative between all MAVs. The second idea focuses on the structure and nature of the inter-robot communication with respect to visual navigation; the thesis explores how a partially distributed architecture can be used to vastly improve the scalability and robustness of a multi-MAV visual navigation framework. A navigation framework would not be complete without a means of control. In the multi-robot setting the control problem is complicated by the need for inter-robot collision avoidance. This thesis presents a MAV trajectory controller based on a combination of classical control theory and distributed Velocity Obstacle (VO) based collision avoidance. Once a means of control is established an autonomous multi-MAV team requires a mission. One such mission is the task of exploration; that is exploration of a previously unknown environment in order to produce a map and/or search for objects of interest. This thesis also addressed the problem of multi-robot exploration using only the sparse interest-point data collected from the visual navigation system. In a multi-MAV exploration scenario the problem of task allocation, assigning areas to each MAV to explore, can be a challenging one. An auction-based protocol is considered to address the task allocation problem. The two applications discussed, VO-based trajectory control and auction-based environment exploration, form two case studies which serve as the partial basis of the evaluation of the navigation solutions presented in this thesis. In summary the visual navigation systems presented in this thesis allow MAVs to cooperatively perform task such as collision avoidance and environment exploration in a robust and efficient manner, with large teams of MAVs. The work presented is a step in the direction of fully autonomous teams of MAVs performing complex, dangerous and useful tasks in the real world.
Supervisor: Konev, B. ; Coenen, F. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral