# FBG-Based Position Estimation of Highly Deformable Continuum Manipulators: Model-Dependent vs. Data-Driven Approaches

## TL;DR

This paper introduces a data-driven approach for fiber Bragg grating-based tip position estimation of deformable continuum manipulators, outperforming traditional model-dependent methods especially during large deflections.

## Contribution

A novel regression-based, data-driven method for FBG-based tip position estimation that reduces errors compared to conventional model-dependent techniques.

## Key findings

- Data-driven method achieves lower mean absolute error (1.52 mm) than conventional methods (3.63 mm).
- Proposed approach maintains accuracy during large deflections.
- Significant improvement in real-time tip position estimation accuracy.

## Abstract

Conventional shape sensing techniques using Fiber Bragg Grating (FBG) involve finding the curvature at discrete FBG active areas and integrating curvature over the length of the continuum dexterous manipulator (CDM) for tip position estimation (TPE). However, due to limited number of sensing locations and many geometrical assumptions, these methods are prone to large error propagation especially when the CDM undergoes large deflections. In this paper, we study the complications of using the conventional TPE methods that are dependent on sensor model and propose a new data-driven method that overcomes these challenges. The proposed method consists of a regression model that takes FBG wavelength raw data as input and directly estimates the CDM's tip position. This model is pre-operatively (off-line) trained on position information from optical trackers/cameras (as the ground truth) and it intra-operatively (on-line) estimates CDM tip position using only the FBG wavelength data. The method's performance is evaluated on a CDM developed for orthopedic applications, and the results are compared to conventional model-dependent methods during large deflection bendings. Mean absolute TPE error (and standard deviation) of 1.52 (0.67) mm and 0.11 (0.1) mm with maximum absolute errors of 3.63 mm and 0.62 mm for the conventional and the proposed data-driven techniques were obtained, respectively. These results demonstrate a significant out-performance of the proposed data-driven approach versus the conventional estimation technique.