BAS Optimized ELM for KUKA iiwa Robot Learning

In this brief, an enhanced robotic learning interface has been investigated using Beetle Antennae Search (BAS) and Extreme Learning Machine (ELM). The initial values of learning weights and bias of the network have significant effect on the performance of the ELM, hence, BAS algorithm was employed to optimize the initial values of learning weights and bias. Kinect v2 camera sensor was applied to obtain the endpoint's position of the upper limb, MYO armband was used to measure the corresponding joint angle values. Those aforementioned data formed the dataset to be trained by ELM and after training the ELM model was able to generate angle values by only giving position as input without a need to carry out kinematic calculations. The proposed method has been validated by conducting series of experimental studies on a KUKA iiwa robot.

Automated summary

The study investigates an enhanced robotic learning interface using BAS algorithm and ELM model to optimize learning weights and bias, achieving the generation of angle values from position input without kinematic calculations. Experimental validation on a KUKA iiwa robot confirms the effectiveness of the proposed method.