Optimisation of swing-up control parameters for a robot gymnast using the Bees Algorithm

This paper focuses on using the Bees Algorithm to optimise the parameters of a swing-up control for a robot gymnast (Robogymnast) attached to a freely rotating high bar mounted on ball bearings. Robogymnast, which mimics the basic movements of a human acrobat swinging on a high bar, consists of three links and three joints. Its motion is manipulated by two DC motors mounted at the shoulders and hip joints. The freely rotating high bar represents the third joint to which link 1 (hands and arms as a single rigid part) is firmly attached. Although, this triple pendulum-like structure is difficult to balance at upright posture, its unpowered joint is advantageous during the swing-up phase. The ultimate challenge was to smoothly swing up Robogymnast from the downward (stable) position to the upright (unstable) configuration by finding optimum values of the parameters that regulate the amplitudes and frequencies of the sinusoidal signals applied to the two DC motors. The Bees Algorithm, a novel swarm-based optimization technique, was used as to achieve this. The simulation and experimental results showed successful swing-up of Robogymnast.