Flocking + Formation Control + Path Following

Robots like the one presented in the picture above are equiped with cameras so the cooperation among the group replicates the behavior of a scanner covering a large terrain. Hence, a controller is conceived so that the group of robots fall into a particular formation (the shape of the scanner-formation), while attaining zero relative velocity among the members of the group (flocking) as the group follows a prescribed path. A simulation with 5 robots displaying the behaviors just described is presented on the following video.

Note that all the behaviors take place simultaneously and the robots do not collide with one another. In the following video we can see the experimental realization with two robots (we only had two).

This was a colaborative effort with a fellow graduate student at the time where there is a machine vision component. For more details please consult the paper ''Flocking, Formation Control and Path Following for a Group of Nonholonomic Robots''.

Close Window

Biometrics Data Acquisition in Real Time

A portable device located in the chest of a driver capable of reporting blood alcohol content (BAC) in real time has been formulated. The correct BAC is estimated from alcohol levels present in the driver's exhalation. This device takes reading form ethanol and CO2 sensors as well as a heart rate sensor that serves to estimate if the driver might be falling asleep.

The device sends the estimated BAC and state of the driver to a device capable of GPRS communication using GARMIN protocol. Additionally, the device emits sound and light signals acording to the current state of the driver.

Digital filters are implemented on an ATMEGA328-PU-ND microcontroller.The blue signal corresponds to ethanol while the red one is the CO2 measurement. The circles mark when someone (me) under alcohol influence (two 20 ounces beers, 5.5% alcohol) exhalated.

Close Window

Splay state on a group of ground and aerial vehicles

Inspired by the concepts of syncronization, we have formulated a control law where two groups of ground and aerial vehicle fall into a circular trajectory while the phases of both ground and aerial vehicles align. Additionally, the groups of robots are distributed uniformly along the circunference, which is know as splay-state. This is also an interdisciplinary effort. For more information, please consult ''Stable Formation of Groups of Robots via Synchronization''

Simulation results.

Experimental validation.

Close Window

Self-balancing robot

A self-balancing robot made with materials that I had readily in my lab. I am using bluetooth modules to keep track of the readings given by the MPU6050 accelerometer. See the code implemented on an Arduino MEGA.

HC-05 Bluetooth modules are included to record the readings in angular velocity and position.

Close Window

Quadcopter overhaul

Taking photogrammetry as the end goal for this platform, I installed a LIDAR to have more precise altitude readings. The processing unit of the quadcopter is an ArduPilot and there were no more inputs available for SPI communication. As I2C was still available on the ArduPilot, I implemented a simple code on an Arduino UNO that works as a simple interface between the LIDAR and the ArduPilot.

The implementation of the LIDAR-Lite sensor inside the ArduPilot project is done in AP_RangeFinder_LidarLiteLL905PIN02.cpp and AP_RangeFinder_LidarLiteLL905PIN02.h .

Navigating through the ArduCopter project with limited documentation was easier with this python code. This function helped me a lot to put me up to speed in understanding the ArduCopter project, or at least the part that I needed to incorporate the LIDAR.

Close Window

IsoGlide at National University of Colombia

This is the project that gave me the opportunity to conduct graduate studies on USA. This is a parallel manipulator intended to perform 3-D printing of ceramic tissue. This manipulator has three translational degrees of freedom and keeps the orientation of the end effector (the table) always fixed. The following document presents more details about the platform. It is in Spanish, I am working on the translation

The driver code for the arm is presented in the following code.

A conventional DC motor was equiped with a power stage, an encoder and a driver to be a servomotor.

A man in the middle, this is the IsoGlide controller. It is the communciation bridge between the computer and the three arms.

Customized optos to work as switches.

The IsoGlide and me.

Direct-Inverse kinematics simulation. Joint parameters are generated to produce a particular position for the table (end effector). The resulting motion generates a pyramid(cyan). Note that the motion was formulated to see that both forward and backward kinematics were working correctly. It is necessary to invert the pyramid so it is printed on the table.

One of the arms following a velocity profile to reduce jerk in the end effector.

The IsoGlide conducting a start-up maneuver.

Close Window

Splay state on dynamic model

Close Window

Drivers for mobile robots

I have written programs in Python that uses a program written in C++ that provides localization services to a bio-inspired robot known as OctoROACH. The C++ program is a position server that communicates with a local positioning system (Vicon system). The boost library is used to generate a C++ wrap file for Python to use, I take advantage here and conduct some matrix transformations to provide more convenient position reports. The python program transmits the legs velocities to the OctoROACH with the aid of an Xbee module. Although all this functionality can easily be implemented in ROS, my implementation is lightweight and ideal for the OctoROACH computing power.

In an attempt to create an optical communication link between different robotic platforms, I found in the need to create drivers for mobile robots and actuators to show a proof of concept. As the prototype was too heavy for the OctoROACH to handle, I decided to make a ground robot out of a little tank that was around. The code of the drivers for the little tank are shown in panzer.cpp and panzer.h The code of the drivers for the gimbal are shown in driver_gimbal.cpp and driver_gimbal.h.

The ground robot has the receiver. The receiver is composed of 4 photodetectors.

The gimbal points the laser (visible light) to the position of the receiver in the ground robot.

Gumstix, it controls the gimbal and the laser to send messages.

Everything put together.

Close Window

Artificial potential field

Application of the ideas developed in the paper ''Hybrid potential field based control of differential drive mobile robots''. This is a simulation of a group of clients requesting printed jobs to be handed to them by a robot. The robot goes to the printer station, retrieves the printed job with a robotic arm and delivers it to the user who did the request.

Dry-run test.

Close Window

Serial Manipulators

A Stanford and Puma manipulators cooperating to weld the logo of the University of New Mexico.

Close Window

Concept Idea

A robot that moves autonomously over a set of solar pannels while cleaning them. This is just a concept idea.

Close Window

Miscelaneous

Animation of an OctoROACH moving on a rough terrain. The camera is tracking it.

I've been learning blender. I think blender provides a very flexible framework to conduct any (mechanical) design, it has excellent performance on Linux and it is free. It has a fluid simulator that I put into practice with the steam coming from the train.

Modeling of a Mitsubishi A6M Zero.

Modeling of a P-51 Mustang.

Use of the fluid simulator present in blender under a more notorious smoke.

Vehicle inspired on the V16 hydra schmidt roadster.

The Homer's mobile.

UNM mascot using photogrammetry.

Close Window

My 3D printing projects

With a 3D printer that I recently got, I am exploring the construction of complex machines prototypes. I have designed and built a thirdhand/press that has served me to do multiple things like holding material to cut or put bearing on my gears.

The gears with bearing are part of a compound gearbox with a ratio of 1:216, which is part of a project I am currently developing.

Currently working on the feedback capabilities.

Close Window