A Force Sensing-based Pneumatics for Robotic Surgery using Neural Network

Nowadays, the use of minimally invasive surgery (MIS) is popular due to its small incision and faster recovery compared to open surgery. However, small working envelopes due to the use of multiple trocars restrict the use of MIS instruments. Performing a complicated surgical operation requires a very skillful surgeon. Many researches on the use of robotic surgery were proposed and developed recently. With robots, operation accuracy can improve significantly, reducing labor intensity and minimizing the effect of human errors. One of the critical parameters of robotic surgery is force sensing. There are several force sensing types and one of them is sensing with the use of pneumatic cylinder. This force sensing utilizes compressible air characteristics and does not require any sensor on the instrument tip. This method is superior to other sensing techniques in terms of structure. Its only drawback is the modeling and computational complexity. Pneumatics has high non-linearity generated by air compressibility. In order to sense force accurately, a complex model needs to be established. In this study, neural network was used to estimate the external force on a pneumatic cylinder. The pneumatic system model was developed in MATLAB R2018a that takes into consideration the compressibility and the friction. To simulate the network model, a direct external force was applied to the pneumatic cylinder. Results provided a high accuracy of force estimation using the proposed model.