A BRIEF INTRO TO: COBOTS, LSM AND A.I. APIs
HELLO! I am Ricardo de Azambuja, Eng, MSc Today I will show you a little bit about our work using robots and artificial intelligence You can find me at github.com/ricardodeazambuja
By the end of this lecture...
By the end of this lecture... … you should be able to: Understand what is a cobot and LSM (SNN) Identify BAXTER's main components Understand how MS Cognitive Service APIs work Smile, since it is supposed to be fun too LSM: Liquid State Machine SNN: Spiking Neural Network BAXTER: cobot from Rethink Robotics, Inc
PART I: COBOTS & LSM
A cobot or co-robot (from collaborative robot) is a robot intended to physically interact with humans in a shared workspace. https://en.wikipedia.org/wiki/Cobot Before we start, I need to give you at least this definition / explanation
A Liquid State Machine (LSM) is a special type of Spiking Neural Network (SNN) Reservoir Computing (RC). Time series can be processed, using only a simple linear combination at the output. Before we start, I need to give you at least this definition / explanation
1 MOTIVATION (1) Or why do we need cobots
Make a joke: "be prepared, strong scenes where dummies get really hurt
What I have been doing with cobots 1 MOTIVATION (2) What I have been doing with cobots
BAXTER Controlled by LSM LSM: Liquid State Machine BAXTER: co-bot manufactured by Rethink Robotics, Inc. Instead of programming, you teach the LSM Explain this is part of my PhD in Robotics / AI
BAXTER Using its sense of sight OpenCV: special library for computer vision Robot detects things according to colour Tell them Davi was a MSc student from Brazil and we worked together to create this demonstration using BAXTER and OpenCV
Point to the fact the robot has NO safety fences, cages, etc and the user CAN interact. Explain OpenCV uses HUE colour values to create a mask, etc to finally find the ball's position
VERY IMPORTANT
NEVER ENTER THE ROBOT'S WORKSPACE Unless you know the robot is turned OFF or it is a SAFE cobot
EXAMPLES OF COBOTS
BAXTER Rethink Robotics, Inc (USA) Photo credits: https://www.technologyreview.com/s/513746/baxter-the-blue-collar-robot/
UR Series Universal Robots (Denmark), the most successful cobot Photo credits: https://www.universal-robots.com/products/ur-robot-benefits/
LBR iiwa KUKA (Germany), high accuracy - high price Photo credits: https://www.kuka.com/en-de/products/robot-systems/industrial-robots/lbr-iiwa
Justin DLR (Germany) Photo credits: http://www.dlr.de/rm/en/desktopdefault.aspx/tabid-5471/
GummiArm Plymouth University (UK), our local development! Photo credits: http://mstoelen.github.io/GummiArm/
BAXTER ROS based (research version) Two arms (7 DOF each), accuracy +/- 5mm. Three cameras (head and one each cuff). Torque/force sensors, IR range sensors, Accelerometers, Sonar sensors (head ring). Remember: they must learn about BAXTER's main components and to you are supposed to program it (demonstration) Talk about SEA's (safer by design), screen (face) and buttons (programming by demonstration) ROS: Robot Operating System Photo credits: https://www.technologyreview.com/s/513746/baxter-the-blue-collar-robot/
BAXTER Control Modes Position Control mode (safest) Raw Position Control mode Velocity Control mode Torque Control mode (only mechanical springs) Best place to learn is http://sdk.rethinkrobotics.com/wiki DOF : Degrees-of-freedom
BAXTER Position Control Mode Show them this mode has several safety mechanisms: All joints arrive together, max speed, collision avoidance (robot hits itself) and detection (impact) Image credits: http://sdk.rethinkrobotics.com/wiki/Arm_Control_Modes
BAXTER Raw Position Control Mode Image credits: http://sdk.rethinkrobotics.com/wiki/Arm_Control_Modes
BAXTER Velocity Control Mode Image credits: http://sdk.rethinkrobotics.com/wiki/Arm_Control_Modes
BAXTER Torque Control Mode This mode only protects the motors (max torque) and keeps gravity compensation No safety mechanisms, but the mechanical design (springy) Image credits: http://sdk.rethinkrobotics.com/wiki/Arm_Control_Modes
BAXTER simulating it Gazebo (sdk.rethinkrobotics.com/wiki/Baxter_Simulator) V-REP (www.coppeliarobotics.com) Best place to learn more about BAXTER is http://sdk.rethinkrobotics.com/wiki DOF : Degrees-of-freedom
ARTIFICIAL INTELLIGENCE API'S PART II: ARTIFICIAL INTELLIGENCE API'S … AND HOPEFULLY SOME FUN WITH A GAME!
EXAMPLES OF AI API'S: IBM Watson Google Cloud ML MS Cognitive Services Amazon Alexa BigML
EXAMPLES OF AI API'S: IBM Watson Google Cloud ML MS Cognitive Services Amazon Alexa BigML
Register and get your key first! REST API Register and get your key first! Acquire data Send Receive results Explain there's no need for a continuous connection. It's HTTP: get / put / post ...
REST API Perfect for low power systems… IoT Image credits: http://www.business2community.com/big-data/internet-things-iot-going-impact-business-01572401
LET'S TRY IT ONLINE https://www.microsoft.com/cognitive-services Use facetime
PYTHON CODE IMPLEMENTATION github.com/ricardodeazambuja/MSCogServ-Bing-Speech-API https://github.com/ricardodeazambuja/MSCogServ-Bing-Speech-API https://github.com/ricardodeazambuja/MSCogServ-Bing-Speech-API/blob/master/Bing%20Speech%20API%20Using%20IPython%20notebooks.ipynb
GAME TIME! github.com/ricardodeazambuja/Simon-Says-Game---Emotions-API
THANKS! Any questions? You can find me at github.com/ricardodeazambuja & ricardo.azambuja@gmail.com
CREDITS Presentation template by SlidesCarnival BAXTER control modes diagrams: sdk.rethinkrobotics.com