Month: September 2015

Multi-channel iTubot-BAT

  By ,

Objective

The overall objective of this project is to develop and evaluate the key technologies of Transnasal Artificial Tentacles, a new approach involving research on the tentacle-like robotic and sensing technologies. This is aiming to create a compliant, flexible, multiple-sensorized, immersive, intraoperative targeting and guidance system for constrained transnasal endoscopic surgeries in a confined workspace.
Minimally invasive surgery-based nasopharyngeal cancer treatment is promising, but currently, it is not a common treatment choice because of the absence of suitable tools. In this project, a multi-channel concentric tube robot is proposed for the treatment of nasopharyngeal cancer based on natural orifice translumenal endoscopic surgery. The proposed system has three channels, i.e. two manipulation channels and one vision channel, and all the three channels are confined by a 10 mm active sheath. The robot is controlled by human-in-the-loop bimanual teleoperation under active endoscopic guidance. The reduced sheath diameter and the steerable vision channel improve the functionality of the system and distinguish our design from the prior art. The feasibility of the system has been evaluated through a series of simulations and experiments. Results show that the proposed system is capable to conduct cooperative tasks in a confined space and the miniaturized manipulator is suitable for transnasal procedures. Besides, comparisons with other types of flexible surgical robots are discussed to further demonstrate the superiority of the proposed system in the target clinical applications.

Funding

Singapore Millennium Foundation
PI: Jan/11/2015 – Jan/10/2018

PUBLICATIONS

Yu H, Wu L, Wu K, Lim CM and Ren H (2016), “Development of a Multi-Channel Concentric Tube Robotic System With Active Vision for Transnasal Nasopharyngeal Carcinoma Procedures”, IEEE Robotics and Automation Letters., Jan, 2016. Vol. 1(2), pp. 1172-1178.

Motion planning of continuum tubular robots based on features extracted from statistical atlas

  By

Abstract

Continuum tubular robots, which are constructed by telescoping pre-curved elastic tubes, are capable of balancing the force application and steerability during minimally invasive surgeries. These devices are able to reach the desired surgical sites in body cavities without colliding with critical blood vessels, nerves and tissues. However, the motion planning of continuum tubular robots is quite challenging because of their complicated kinematics as well as the high dimensional configuration space. In this paper, a sampling-based motion planning method is proposed based on the Rapidly-exploring Random Tree (RRT) algorithm for continuum tubular robots in 3D environments, such as medullary cavities. The proposed motion planner enables a continuum tubular robot to maneuver roughly along the central axis of the statistical humerus atlas in an approximate follow-the-leader manner. The experiment results have demonstrated the effectiveness and superiority of the proposed motion planning algorithm.

full text

Hierarchical Recognition System for Target Recognition from Sparse Representations

  By

Abstract

A hierarchical recognition system (HRS) based on constrained Deep Belief Network (DBN) is proposed for SAR Automatic Target Recognition (SAR ATR). As a classical Deep Learning method, DBN has shown great performance on data reconstruction, big data mining, and classification. However, few works have been carried out to solve small data problems (like SAR ATR) by Deep Learning method. In HRS, the deep structure and pattern classifier are combined to solve small data classification problems. After building the DBN with multiple Restricted Boltzmann Machines (RBMs), hierarchical features can be obtained, and then they are fed to classifier directly. To obtain more natural sparse feature representation, the Constrained RBM (CRBM) is proposed with solving a generalized optimization problem. Three RBM variants, L 1 -RNM, L 2 -RBM, and L 1 / 2 -RBM, are presented and introduced to HRS in this paper. The experiments on MSTAR public dataset show that the performance of the proposed HRS with CRBM outperforms current pattern recognition methods in SAR ATR, like PCA + SVM, LDA + SVM, and NMF + SVM.

full text