Deep-Learning-Based Compliant Motion Control of a Pneumatically-Driven Robotic Catheter

In cardiovascular interventions, when steering catheters and especially robotic catheters, great care should be paid to prevent applying too large forces on the vessel walls as this could dislodge calcifications, induce scars or even cause perforation. To address challenge, paper presents a novel compliant motion control algorithm that relies solely position sensing of catheter tip knowledge catheter’s behavior. The proposed features data-driven controller. controller is trained based so-called control Long Short-Term Memory Network (control-LSTM). Trajectory following experiments are conducted validate quality control-LSTM. Results demonstrated superior positioning capability with sub-degree precision new approach in presence severe rate-dependent hysteresis. Experiments both simplified setup well an aortic phantom further show allows reducing interaction environment by around 70%. This work shows how deep learning can exploited advantageously avoid tedious modeling would needed precisely steer continuum robots constrained environments such patient’s vasculature.