A Dynamic Interaction Simulator for Studying Macroscopic Swarm Self-organization of Autonomous Surface Watercraft

摘要

Despite the advances in autonomous watercraft, both underwater andsurface, a robust methodology for distributed control and selforganizationin swarms is lacking. In the present work we develop asimulator that allows for investigations of swarm macroscopicdynamics when the interaction forces between the constituent units ofthe swarm abide with different mathematical laws and governingequations. Every watercraft unit in the swarm is represented as amassless rigid beam of known length connecting two point masseswith, in general, different value at the ends. This defines the mass andthe moment of inertia of each vehicle in the swarm. The masses of onevehicle need not generally interact with those of other vehiclesgravitationally. In contrast the bow and stern masses can be taken to be“charged”, with identical or oppositely signed charges, resulting inattractive or repulsive forces obeying some form of easilyprogrammable, mathematical, law. Additionally, the charged masses atthe ends of each vehicle will also be subject to externally applied forcefields. Each vehicle will as a result move on a path that is notpreprogrammed, but is derived dynamically on the basis of the unit’sinteractions with other units in the swarms together with the effect ofany exogenous fields. The macroscopic dynamics of swarms ofvehicles complying with programmable artificial physics of the sortintroduced above need to be analyzed in order to be able, at asubsequent stage, to reverse engineer the observed dynamics and designgoverning laws for the interactions so that a desired pattern of swarmbehavior or self-organization emerges on the holistic level. A simulatorwill be used to support analytical findings based on the theory ofdissipative structures and self-organizing systems applied to distributedcontrol and automated dynamic planning of surface watercraft systems.A critical detail is the fact that such vehicles are restricted to moving ona plane in the simplest of case assumptions. Simulations using the toolof this work will allow further investigations, enabling the developmentof a formal synthesis technique of dynamic interaction governing lawsmeeting arbitrary macroscopic self-organization requirements.