Web-based platform to control a real anthropomorphic robotic arm remotely using artificial intelligence.
Movetherobot.live consists of an online platform that allows users to control a robotic arm remotely via an internet connection. The platform uses artificial intelligence and robotics technologies to enable the robotic arm to move precisely and naturally in response to real-time commands from the user.
The goal of this project is to democratize access to robotic technology and enable more people to experiment with robots and learn about the technology in a safe and accessible environment.
Web platform for collaborative database generation for training of intelligent algorithms based on myoelectrical signals.
OpenROBi arises from my final degree work in the search for a myoelectric signal acquisition protocol for use in the creation of algorithms for motion prediction. Particularly, the models used in the thesis achieved excellent metrics with the collected data, but when adding new data, the "online effect" was present. It turns out that small modifications in the the position of the electrodes, the change in the intensity of the ambient noise and the variations inherent to the muscle state at the moment, are factors that directly affect the predictions.
OpenROBi introduces the protocol used in the thesis, and a system whereby anyone can perform exactly the same data acquisition exercise to collaborate with the overall database.
Investigation of different artificial intelligence algorithms applied to the electrical muscle activity of the human arm in the search of a model that predicts its movements!
In this project, the research of different artificial intelligence algorithms for the improvement of control systems based on electromyography was carried out, whose motivation arises from the study and application of robotics for humans in the rehabilitation area, specifically for a user with transhumeral amputation. For the realization of this project, it was necessary the materialization of a prototype of an anthropomorphic robotic arm and an angular datalogger for the generation of a database.
Design, materialization and control of an anthropomorphic robotic arm.
An interesting detail about this project is that the final actuator was designed based on a human hand by biomodeling from a CT scan. The hand has a system of internal channels that are responsible for allowing the passage of the tensors that transmit the movement. This complex and laborious design is intended to take advantage of the biofeedback of mirror neurons when training for the use of the robotic arm using biopotentials!
This video summarizes the robotics stage of my final degree project.
Computer vision mapping of hand motions with AI for robotic hand control
In this project, the hand is mapped with artificial intelligence, and then, the relative angles of the finger joints are calculated to be subsequently used in the control of a robotic hand. The whole implementation is done in real time and for each finger separately. Soon I hope to bring this concept to the whole arm!
Holter prototype for cardiac signal monitoring using AI processed acceleration signals for patient context detection.
The particularity of this project is that the prototype had to be as cheap as possible, in order to make the device accessible. The essence behind this Holter is to democratize the diagnosis and monitoring of the electrical signals of the heart, trying that anyone can replicate the project with little money and at home, currently this prototype costs around 20 USD. Currently it only analyzes one channel, and the AI processing must be done offline (like most of the commercial holters) but the idea is a constant connection with a database, to generate the corresponding alarms depending on the patient's condition. As this project is essentially open source when I reach a good progress I will upload the designs and the code!
Use of object detection models as home automation system control. Incorporation of intelligent alarms for people with disabilities.
Glove-shaped device to control the mouse cursor in a different and fun way!
Domotics system applied to people with disabilities
With my teammates Cristian and Matias we developed this home automation system applied to people with disabilities. This system was designed to generalize and scale the problems of disability, taking into account that diseases evolve over time, in the same way Mi Cassa evolves allowing to incorporate new devices and methods of assistance!
This project is open source and is hosted on the Posta-Project platform.
Co-author on: "Artificial intelligence architecture for detecting patterns in scanned documents." Vol. 8 No. 11 (2022): SID 2022 - Simposio Argentino de Informática y Derecho - December 14, 2022.
With my colleagues Cristian and Matías, in collaboration with UBA IALAB, we developed a computer vision system to detect visual patterns in police files in order to classify the risk level of victims of gender violence. This project emerged during the covid pandemic, due to the increase of gender violence cases as a consequence of confinement. The interesting detail was the use of different computer vision systems instead of the classical approach using natural language processing techniques. As a result, the accuracy of detection of the risk level increased from 60-70% to more than 95%.
Check out the paper!
https://publicaciones.sadio.org.ar/index.php/JAIIO/article/view/285