APPLICATION OF NEURAL NETWORKS IN A PARALLEL MANIPU- LATOR WITH FLEXIBLE LINKS FOR MODEL EXTRACTION

Resumo

Parallel manipulators (PMs) are a viable design alternative for industrial applications. Due to their
closed kinematic architecture, they present some advantages compared to their serial counterparts: lightness, high
speed/acceleration ratios, high rigidity, load capacity, and high compactness. However, this design option could
yield undesired vibrations due to its components‘ flexibility requiring the implementation of novel joint and task
space control strategies. Two main challenges arise when designing a control strategy for a PM: the lack of a
direct measurement of the end-effector‘s pose and their coupling dynamics. This work proposes an estimator for
assessing the end-effector‘s pose of a PM using Artificial Neural Networks using measurements from encoders,
strain gauges and camera. The encoders measure the angular displacement of the active joints of the manipulator,
the strain gauges the deformation of the links and the camera the position of the end effector. The proposal is
validated using experimental data from a 3RRR PM with flexible links. The estimator can be used in control
schemes to enhance the performance of flexible PMs.