Elasticity compensation using explicit learning


The broad goal of the project is elasticity characterization using learning and elasticity compensation in Roombots modules.

The Roombots project at BioRob uses modular, homogeneous, and in-series robots. They are moving, attaching or detaching, and reconfiguring on a grid as single modules and meta-modules. Reconfiguration cannot always be performed because of the deflection in the structure. The hypothesis is that the main causes for the deflection are elasticity and backlash in the modules’ joints.




The goal of my internship is to characterize the elasticity of the modules, and ideally find the main cause of the module bending. The global method chosen to reach this goal consists in using a learning algorithm on experimental hardware, and simulated data. The real data came from an infrared tracking system (Vicon system), and the simulated data was acquired using a finite element analysis tool in SolidWorks.


In the final report, I am presenting results both from FEA and hardware experiments, which give a first estimation of the Roombots elasticity.