Collaborative Robots in Surgical Research: a Low-Cost Adaptation

This work demonstrates the adaptation of an industrial robotic system to an affordable and accessible open platform for education and research through rapid prototyping techniques. The ABB YuMi collaborative robot is adapted using a low-cost 3D printed gripper extension for surgical tools. The robot is controlled using an intuitive virtual reality teleoperation system using the HTC VIVE controllers. The design and assessment of three surgical tools in two mock surgical procedures are showcased in this work. The surgical tasks involved tissue removal with the designed cutting tools, where their effectiveness and completion time are assessed. We conclude from these results, that the perpendicular scalpel tool is preferred for faster completion time, but the scissors are preferred for small tissue removal in terms of effectiveness.

HRI 2018         Poster

One-Shot Gesture Recognition: One Step Towards Adaptive Learning

As a visiting scholar at Purdue, I worked with Dr. Cabrera on the problem of one-shot gesture recognition.  The framework was based on extracting a set of salient points within the gesture trajectory and finding a compact representation of each gesture class. This representation is then used to augment the number of examples of each gesture class artificially, maintaining intrinsic characteristics of the gestures within that class. The gestures used are a subset of the publicly available data set of Microsoft Research MSRC-12. Both humans and classifiers are used to assess gestures performed artificially by a robot. We used four classifiers commonly used in the state-of-the-art for gesture recognition to evaluate the effect on the overall performance (HMM, SVM, CRF, DTW). We measure the overall accuracy for humans and classifiers and apply a second metric to measure the level of coherence between the performance of the classifiers and human observers.

FG 2017

Imitation learning for surgical robots with virtual demonstrations

I created Vrep simulators for the ABB YuMi robot and SRI Taurus robot. The robots can be controlled in the simulation environment using the HTC VIVE virtual reality headset and controllers. This system allows the user to collect abundant data through an intuitive interface that provides an immersive experience. Demonstrations will be used to train different robot learning approaches for mock-up surgical tasks.

Design of a passive walker

At the University of the Andes, I developed a low-cost passive walker with my team D-WALKERS, under the supervision of Professor Carlos Francisco Rodriguez Herrera. The walker can stably walk down a slight slope (~2 - 5°) leveraging gravitational potential energy. The prototype is developed with two stiff aluminum legs joined at the hip by a transversal steel rod holding movable weights. The feet are assembled from laser-cut parts made of MDF wood, and the overall manufacturing costs of the walker are under $30. The manufacturing manual for the walker model can be found here.