- 16
- 85 426
Slawomir Korbas
Poland
Приєднався 11 лют 2013
Robotics, deep learning and programming. Projects and designs from the field of robotics, neural networks, deep learning and software architecture. Small robot projects and software applications based on Java, C++, Python and JavaScript. All the content provided is result of my work and self-learning process. For the majority of robotic projects I am using Arduino boards and Nvidia ARM64 computers (SBC).
Why robotics?
Robotics is an inspirational field where multiple domains of science and engineering meet. From digital electronic, mechanical engineering and cross-platform software solutions to artificial intelligence and machine learning. Building machines capable to perform dedicated task in human-like way is extremelty satisfiyng.
Why robotics?
Robotics is an inspirational field where multiple domains of science and engineering meet. From digital electronic, mechanical engineering and cross-platform software solutions to artificial intelligence and machine learning. Building machines capable to perform dedicated task in human-like way is extremelty satisfiyng.
Changing Dynamixel velocity in Position Control mode
Changing Dynamixel servo motor speed (velocity) while operating in the Position Control mode using Arduino dynamixel shield from Robotis. The velocity is set using PROFILE_VELOCITY property from Dynamixel Control Table (RAM area) by method writeControlTableItem from the shield's API. the unit for the new speed value must be calculated in 0.229 rpm unit. This movie concerns Dynamixel XL series servo motors.
Переглядів: 4 037
Відео
Walking quadruped robot - testing inverse kinematics and collision avoidance
Переглядів 7 тис.3 роки тому
Robot dog (quadruped) build and walking test. Hardware specification: Main processing unit: Nvidia Jetson Nano 4GB developer kit (Quad-core ARM A57 @ 1.43 GHz 128-core Maxwell GPU) Move control unit: Arduino Leonardo Dynamixel Shield(Robotis) Storage: NVMe SSD 256 GB drive SanDisk microSD (Extreme PRO) Collision detection unit: RP Lidar from SLAMTEC Computer Vision: Raspberry PI V2.1 8mpx CSI c...
Collision avoidance with LIDAR and Jetson Nano
Переглядів 7 тис.3 роки тому
Implementation of environment scanning and collision detection using RPLidar from Slamtec and Jetson Nano single board computer. Lidar is used to detect obstacles in front while the robot is walking. Also LIDAR data can be visualized in robot control Python application using websocket connection and polar chart from plotly.js library.
NVIDIA Jetson Nano and Dynamixel robot: control using Python
Переглядів 13 тис.3 роки тому
Jetson Nano robot review. Hardware description of the walking robot platform and demo of Python web application communicating with the robot. This walking dog (quadruped) robot movement is controlled by Jetson Nano developer kit computer which is connected to Arduino board using serial UART interface. Jetson computer sends commands to Arduino where another porgram resposible for movement sequen...
Deep learning and Tic Tac Toe game in Java - part 2
Переглядів 6433 роки тому
Second part regarding Tic Tac Toe game implementation using deep learning and custom artificial neural network framework written in Java language. Here I am discussing: - training data preparation - supervised learning code - tic tac toe engine performance evaluation This application is currently deployed on Heroku here: tic-tac-toe-ai-game.herokuapp.com/tictactoe Github repository: github.com/...
Jetson Nano and Arduino serial communication using Python
Переглядів 22 тис.3 роки тому
Example of data serial data communication between Nvidia Jetson Nano computer and Arduino Leonardo board. Both boards are connected using micro USB cable. The same connection is used to program Arduino board from the Jetson using Arduino IDE. Alternatively for the serial communication also pins 6, 8, 10 (GND,TX,RX) from Jetson can be used along withe UART to USB converter. On Jetson side short ...
Deep learning: Back-Propagation in artificial neural network
Переглядів 1873 роки тому
Back-propagation is the essential algorithm in neural net training. This presentation explains the base concept of machine learning process in neural network with realtion to back-propagation and stochastic gradient descent. Machine learning in artificial neural network is about tuning weights of the net so the cost function with respect to each weight reaches its minimum value. Proper tuning o...
Robot dog walking - quadruped motion test
Переглядів 3,8 тис.3 роки тому
Robot walk test. Quadruped robot walk sequence based on simple kinematic formula. Quadruped Dog robot consists of 12 servo motors. Lidar and collision detection not activated yet. Control unit based on Arduino Leonardo and Dynamixel shield from Robotis. Robot is powered by 11.1V Li-Po battery and DC/DC step down converter so the proper voltage can be delivered for Arduino board.
Walking robot test
Переглядів 1,3 тис.3 роки тому
Robotic dog initial walking sequence test of my four leg robot dog based on Arduino Leonardo, Dynamixel XL-430 servo motors and Robotis Arduino Shield. Further improvements assume: * enabling LIDAR and implementing "self-awareness" (collision avoidance) by scanning environment * get rid of labolatory power supply and using Li-Pol battery along with DC/DC converter * adding Jetson Nano and camer...
AI playing Tic-Tac-Toe game - deep learning and neural network - part 1
Переглядів 7 тис.3 роки тому
This video covers deep learning in relation to the web application capable for playing Tic-Tac-Toe game (called also: "Noughts and Crosses"). The decision making engine is based on trained neural network. The goal was to design such neural network topology and train it with proper data so the engine (TicTacToe agent) is able to win/draw over 90% games against "random" Tic-Tac-Toe player - see t...
NVIDIA Jetson Nano 2GB Developer Kit - preparing microSD Card
Переглядів 1,8 тис.3 роки тому
In this video I am showing alternative way for formatting SD card that is used to flash Ubuntu Linux distribution dedicated for Jetson Nao 2GB Developer Kit. The recommended by NVIDIA CD Card Formatter application didn't worked for me and after flashing Ubuntu image to it the Jetson Nano was not able to boot. After multiple tries and searching over the forums I used FAt32Format command line too...
Powering servo motors and Arduino board using single Li-po battery
Переглядів 1,8 тис.3 роки тому
Introduction of DYI DC/DC step-down converter capable for delivering 5, 6 and 11 Volts to power up Arduino and servo motors using single battery pack. The battery used is Li-po 11.1V. The DC step-down converter consists of two modules. One with max current 2A and the second having 3A. Both modules are followed by huge elecrolityc capacitor 4700uF. Everything is soldered into one universal PCB b...
Using software serial port from Dynamixel shield to monitor program on Arduino
Переглядів 3,1 тис.3 роки тому
Quick overview on how to use additional software serial port exposed by Dynamixel-Arduino shield to monitor or debug programs. By default the shield from Robotis is using hardware serial port (UART) to control Dynamixel servo motors. While operating in "dynamixel" mode you cannot use this port to monitor your program. This is why the shield provides additional UART socket to communicate with Ar...
Multiple Dynamixel servos and Arduino shield from Robotis
Переглядів 4,5 тис.3 роки тому
Brief explanation on how to update default factory ID of Dynamixel servo motors using the Arduino Shield library and sketch from Nikolas Lamb - github.com/nikwl. After changing IDs of Dynamixel servo motors you can address them separately and call desired actions like setGoalPosition etc. Updating factory ID (1) of Dynamixel servos is a first step before programming real application that uses t...