Abstract. We describe further progress towards the development of a MAV (micro aerial vehicle) designed as an enabling tool to investigate aerial flocking. Our research focuses on the use of low cost off the shelf vehicles and sensors to enable fast prototyping and to reduce develop- ment costs. Details on the design of the embedded electronics and the modification of the chosen toy helicopter are presented, and the tech- nique used for state estimation is described. The fusion of inertial data through an unscented Kalman filter is used to estimate the helicopter’s state, and this forms the main input to the control system. Since no de- tailed dynamic model of the helicopter in use is available, a method is proposed for automated system identification, and for subsequent con- troller design based on artificial evolution. Preliminary results obtained with a dynamic simulator of a helicopter are reported, along with some encouraging results for tackling the problem of flocking.68719

1 Introduction

Swarm robotics is nowadays an established field of research; it offers the ad- vantages of scalability, robustness through redundancy, flexibility, and reduced complexity of the inpidual robots. Within swarm intelligence, the topic of flock- ing deals with methods for controlling the motion of a group of agents (in real or virtual space) using rules directly inspired by ethological observations of real flocks of birds or schools of fish.

Since the seminal treatment of flocking developed by Reynolds [1] several researchers have explored the idea of flocking in real platforms or simulations. Most of the work involving flocking in real robots has concentrated on wheeled robots [2] [3] or airborne robots with limited dynamic capabilities [4]; due to the intrinsic dynamic and sensory limitations of the platforms used, none of these examples achieved really good-looking fluid flocking. Crowther addressed the problem of vehicles with more complex dynamics in his simulations of a flock of aircraft [5]. His research successfully demonstrated the potential usability of flocking as a decentralised traffic control method. In particular it showed that by simply changing the weights associated with Reynolds’ rules, phase transi- tions appeared in the flock structure. However, omnidirectional perception was

assumed, and the presence of noise was neglected. Recently a development pro- gram carried out at the NASA Dryden Flight Research Center [6] demonstrated the coordination of two UAVs (in the form of two instrumented model aircraft) using Reynolds’ flocking rules. GPS information was used to determine the rel- ative positions of the aircraft, and this was sufficient to guarantee coordination. Unfortunately further details about this project are still unavailable. Another interesting application was developed by Atair Aerospace Inc. [7] in the domain of guided parafoils; a behaviour based algorithm inspired by flocking is used to ensure that all the payload-carrying parafoils will land together in the same area. In the last few years, the problems of flocking and the distributed control of agents have gained popularity among the control system community [8][9][10]. The problems of stability, robustness, and the effects of sensing or communication delays are now being considered; see [11] for a more extensive review in the field. In a recent paper [12], Olfati Saber presents a theoretical approach to flocking; a single distance dependent potential function is defined to achieve both cohesion and separation. A particularly good definition of the potential function results in a smooth pairwise potential with a finite cut off that greatly simplifies the stability analysis. Since cohesion-separation and alignment can lead to fragmentation, an additional contribution to the control is added in the form of navigational feedback from progress towards a target point. The paper also presents an obstacle avoidance behaviour obtained by introducing fictitious