Multi-Agent Source Seeking in Unknown Environments
A hybrid adaptive feedback approach for unicycles
M.J. van der Linden (TU Delft - Mechanical Engineering)
Suad Krilašević – Mentor (TU Delft - Team Bart De Schutter)
Sergio Grammatico – Mentor (TU Delft - Team Bart De Schutter)
L. Ferranti – Graduation committee member (TU Delft - Learning & Autonomous Control)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
The application of autonomous robotic vehicles to explore unknown environments is a growing field of interest in the research community. This thesis report studies the practical implementation of employing three autonomous vehicles to navigate towards a signal emitting source in an unknown environment. The group of autonomous vehicles is assigned to drive towards the source in a coordinated formation shape whilst avoiding an obstacle. Therefore, a control algorithm is developed that is required to be suitable for a non-holonomic unicycle. As an experimental platform for the algorithm, the Husarion ROSbot is used. To evaluate the performance of the algorithm, simulations are run in a Gazebo simulator environment. To achieve the stated research objective, the problem is broken down into three sub-objectives: source seeking, obstacle avoidance, and formation control. First, two types of source seeking approaches for unicycles are implemented to investigate how each approach affects the source seeking performance. Thereafter, an obstacle avoidance algorithm is developed to combine with both source seeking approaches. [19] is used as a framework to create a hybrid adaptive feedback (HAF) law to avoid an obstacle, with the location and orientation of the obstacle w.r.t. the source is not assumed to be known a priori. Subsequently, two follower vehicles are introduced to study how to enable a group of vehicles to drive in a coordinated formation to the source while avoiding an obstacle. Finally, the algorithm developed in this thesis is empirically shown in simulations not to suffer from convergence to a detrimental line M that would occur with an artificial potential function approach, as described by [24]. The follower vehicles employ a decentralized leader-follower strategy to maintain formation shape and successfully avoid an obstacle.