Version 3 of the official K3lso Quadruped by Robin Fröjd.

First connection between #k3lso of @r_frojd and #ROS 🥳. Publishing joint states and visualizing them on #rviz in realtime. (some remote programming to face my toothache).#k3lso #quadruped #quad #robotics #mit #cheetah #UPXtreme #mjbots @UPbridge_gap pic.twitter.com/ivxGLPF34e

— Camilo Camacho (@camiloim93) April 26, 2021

Quadruped robot based on the k3lso v2 prototype by Robin Fröjd.

#k3lso walking using #champ. https://t.co/zTQonfmzAM#dynamixel #ROS #robotics #quadruped #spotmicroai #upsquared @UPbridge_gap @ROBOTIS @r_frojd pic.twitter.com/O16Bj5CHLh

— Camilo Camacho (@camiloim93) November 3, 2020

ROS packages for visual 3D mapping and localization, based on RTAB-Map, and using Realsense Cámeras (D435 and T256), on iRobot Create 2 robotic platform.

Robot developed for the participacion in the Latin American Robotics Competition, in the IEEE-Open category. We (GED-USTA Team) got the first place.

This research project aims to develop, implement and evaluate algorithms and methods of supervised learning, in order to create models that allow imitating the movement of upper limbs from video demonstrations, and thus replicate them in humanoid robots such as DarwinOP or the Poppy Torso robotic platfotm. The development will be carried out in three main stages: the contextualization of the state of the art and the methods currently used, the implementation and training of the models based on the proposed algorithms, and the evaluation of the results obtained from the samples extracted from video.

The goal is for the robot to learn to move its upper limbs from video demonstrations with people. This entails two challenges, how to make the transition from the state space of people to that of the robot and how to generalize the policy. To solve the problem, it is proposed that the robot, based on reverse reinforcement learning, infer the mathematical representation that describes the task shown in the demonstrations. In this way, the robot can learn a generalized policy similar to that of the expert. For learning and validation, the use of humanoid robots, such as the DarwinOP or the open-source platform PoppyTorso, is planned.

Propper implementation of the Poppy Torso robot. (Original Project). We use the original hardware and 3D models, but the firmware and software are being developed by us.

Real-time Urban Traffic Monitoring and Classification System through Digital Image Processing

Performance Evaluation of Controllers for Temperature and Level Regulation in Cascade Tanks

Software to record RGB-D datasets via Realsense D435

OpenPose implementation using tensorflow.

Dijkstra algorithm based pacman solution.