Adapted Neighborhood Particle Swarm Optimization Algorithm for Kilobot in Kilombo Simulations

This paper presents the simulation results of a swarm of Kilobots optimized to find a given region of interest. In a previous paper the authors have shown that Neighborhood Particle Swarm Optimization (NPSO) is a viable candidate to deal with the constraints of Kilobots such as noise and limited communication range and have applied an NPSO control algorithm to Kilombo, a Kilobot simulator, which incorporates physical attributes of Kilobots. The physical attributes of Kilobots that had been ignored in the Matlab simulation earlier include the following: (1) The particles are assumed to be dimensionless, and hence they could overlap in space to all land at the optimum solution. (2) Kilobots can sense the range of its neighbor but lack the direct mapping capability to execute a movement using a displacement vector in the Cartesian coordinate. (3) (trilateration is possible but costly in computation and energy and hence not used). They cannot make instant moves or turns as required in the update steps of the Matlab simulations. The realistic physical constraints require adaptation of the NPSO algorithm, yielding a new NPSO-inspired movement control algorithm. Due to (1), Kilobots cannot overlap and all land at the best solution, instead, they will accumulate around the successful solution and stop the search. Therefore, the metric to determine success is adapted to represent the feasible ending state instead of exact accuracy for the Kilobots to have to all land at the best solution. (2) and (3) require that we change the discrete vector math in NPSO into a pseudo-continuous method. In this paper, we presented four algorithms by adding more and more factors to guide the Kilobot movement. Then, the simulation results in the Kilombo environment are evaluated and presented using convergence plots and success tables.