Distributional modelling and control of large-scale passive swarm robots

The overall aim of this research is to address some of the challenges for modelling and controlling a large number of particle swarm robots. The inspiration of this research was the control of microrobots, in which because of their tiny hardware space, have limited computational capabilities and are reliant on external magnetic field, controlled by a central distributional controller, for movement to fulfil their objectives. The probability distribution to represent the behaviour of a large scale swarm of robots and a representation of the potential field are defined, which are parametrised through Gaussian basis functions. A swarm distributional states estimation algorithm, modified from the Expectation-Maximisation algorithm, is also presented to calculate the swarm distributional states. To represent the swarm robots
as a probability distribution, a continuum representation of the advection-diffusion model is first considered and a reduced-order model of the advection-diffusion is derived using the Galerkin approach and parametrised with Gaussian functions. Finally, to control the robot swarm using this reduced model, a distributional controller was developed to control the trajectories of the robot swarm for open-loop and closed-loop control. The models and the control schemes were analysed using simulations for a swarm of robots to form a desired objective spatial distribution. The simulations show the potential for this novel modelling and control framework to be applicable to control the behaviour of a swarm of large number of robots.