Anthropomorphic Robotic Grasping with Object Shape Completion
In this project we designed an end-to-end framework to grasp partially visible objects from any desired region using human-like grasps. We do so using a system conformed of an RGB-D camera, 7-DOF robotic arm and a 7-DOF multi-fingered robotic hand.
Grasp Surface Detection, Tracking & Analysis
In this project we designed a framework to capture individualized human hand-object interactions. With our framework, we are able to accurately extract kinematics, hand shape, and contact surfaces. This framework serves as a baseline for understanding how humans manipulate objects, which in turn benefits the robotics grasping community.
In this project we devised an automatic framework capable of producing tactile sensors for objects of arbitrary shapes. We tested our approach with 3D printed rigid objects, human, and fingers. The robustness of the approach allowed us to use the sensor signals to control a robot hand with high accuracy.
In this project we designed an approach to predict grasping postures on the entire geometry of unseen objects. Our deep learning approach requires minimal object samples, as it leverages local geometric features and relates them to grasping postures, which were initially analyzed from human hand-object interactions. This approach has considerable potential within the field of robotics and prosthesis development.
Our framework is capable of designing a robot hand tendon routing capabale of reproducing any desired kinematic synergy (motion). Our results can be directly applied to robot hand design concepts, which has the potential to bring robot grasping and dexterous manipulation a step closer to real-world applications.
In this project we devised a methodology to improve the manufacturing accuracy of 5-axis CNCs. Our method was able to enhance contouring accuracy by up to 87%. This project was done in cooperation with the Industrial Technology Research Institute of Taiwan.
Mechanism generator for predefined positions
In this project we developed a python based program capable of generating planar mechanisms which satisfy two or three position constraints given by the user. This open source easy-to-use program can save mechanism designers considerable hours, by returning accurate results which can be directly applied to CAD simulations. You can download the code here.
Analysis of mechanical 3D structures
In this project we developed a set of software, which can determine the effects arbitrary forces have on arbitrary structures. This is done following the principle of virtual work, effectively determining forces in the entirety of the structure. Mechanical parameters can be easily adapted to the user application. You can download the software here.
Design and implementation of an electronic cam for user-defined motions
In this project we designed a novel electronic cam concept. Traditionally mechanical cams have been used to generate complex linear motions in industrial settings. Nevertheless, motions are limited and manufacturing costs are high. With our approach we can generate arbitrary linear motions for virtually any application with a single system. Our results showed the applicability in real-world scenarios.
In this project we designed a machine to dry quinoa grains. Drying is an essential step for the treatment of quinoa. In andean communities, where this grain is produced, this is mainly a manual process. With our design we intend to provide an applicable solution to reduce the manual input needed for this process. We designed the full mechatronics aspect of the machine, including mechanical design, electronics, control and interfacing. Feel free to see our well documented report.
Robotics competition 2013
In 2013, we organized a robotics competition at the University of the Armed Forces in Ecuador. We developed a laberynth where students could test their mobile robots. We designed the whole concept, including automatic point counting, intelligent obstacles, among others. This was the first robotics competition in the university. Nowadays this is a regular event.
Big thanks to my good friends and former colleagues.