This thesis explores two significant aspects of robotic-assisted surgeries. The first part focuses on a procedure called Robotic Catheter Ablation, which utilizes robots to treat a heart condition known as Atrial fibrillation. The procedure aims to determine the contact force at the robot's tip without the direct use of a force sensor, ensuring proper contact with the patient's heart. Recent findings have highlighted the importance of considering blood flow within the heart, as it can significantly impact the procedure's success, often neglected in similar studies. This research examines the experimental effects of blood flow on the movement of soft robots and demonstrates how disregarding this effect can lead to less effective treatments. Simultaneously, this study focuses on enhancing soft robots, introducing a method for determining their shape when subjected to varying forces, similar to the way blood flow affects them. This is accomplished through mathematical modeling, employing Euler-Bernoulli beam theory and cubic Bézier curves. The model's accuracy is confirmed through finite element method testing. Additionally, the research goes beyond shape prediction by including an approach known as Bayesian neural network (BNN). This BNN significantly improves our ability to update information rapidly, valuable for various real-time applications. In the final part of this research, applying the reshaped dataset from the shape reconstruction phase, it is demonstrated that the results obtained from the mathematical model can estimate forces using a new neural network, validated through experimentation. The proposed model holds potential for use in different fields of soft robotics, crucial for surgical procedures.