Sensors

I looked into what would be the essential parts my new robot will need in order to learn balancing and move around. I decided I will need a gyroscope with accelerometer. The gyroscope will help provide orientation while the accelerometer will provide momentum. Knowing momentum will be used to determine how strong of a correction the robot will need to make when it sense a change in movement.

Frame

Next I would need to construct a lightweight frame. I spent a lot of time in this area. There are so many ways I could try to make it. I went with making the frame out of aluminum bars. This provided a cheap, lightweight, strong, and easy to work with material. Instead of actually engineering the frame, I can sketch up a concept and make the parts as I go.

The overall concept was modeled after a robot toy. I looked at each body part and the relative ratio to one another. I then thought about the amount of freedom each joint would need. During my research I saw others do as little as 16 degrees of freedom (DOF) to 28. I decided to start with 20 DOF for this robot. Some joints have 1 degree of motion like the knee and others have 3 degrees like the hip. I will add more later as I the robot progresses.

The hardest part of making the frame was linking multiple servos together. In the image above you can see the servos come with gear on one side. To make a stable connection each servo need a frame. The frame creates a second connection point on the back side of the servo. Below you can see the knee joint servo with a frame and leg extension attached.

The knee joint is relatively easy to make, but when I got to joints with 2 or 3 DOF. It got a lot more complicated. To keep the joint size as small as possible multiple servos will need to be combined into one part. Below is how I combined 2 and created a 2 DOF. Then when adding a third became nearly impossible. To make it work I had to attach it to the end of one. This depended on how I wanted the joint to function and where it made the most sense for this third axis.

Then comes the feet. I decided to go with wood on this, since it was easier to attach wood than created another custom metal frame. It ended up looking more like sandals than feet.

Putting it all together was the fun part. Took quite a bit of time to assemble and adjust things as needed. Yes the the robot is still missing a head and has no spine movement. Those would be future addition. The next part of this series will be the electronic portion. Where I will show how I wired up all the servos and control system. I hope the servos will be able to handle the weight of this design. If not then I am going to need bigger servos!

I hope you all look forward to my finished project. My goal is to have the robot learn to stand up and walk on its own. Then progressively add interactive features to it.

Resources

Servo

• https://learn.adafruit.com/circuitpython-essentials/circuitpython-servo
• https://learn.adafruit.com/adafruit-16-channel-servo-driver-with-raspberry-pi/hooking-it-up

Gyroscope + Accelerometer

• https://learn.adafruit.com/mpu6050-6-dof-accelerometer-and-gyro/overview
• https://learn.adafruit.com/mpu6050-6-dof-accelerometer-and-gyro/python-and-circuitpython

Parts

• https://www.adafruit.com/product/155
• https://www.adafruit.com/product/3886
• https://www.adafruit.com/product/815

The post Making the Humanoid Robot Part 2 appeared first on Botmation's personal blog for creating robots and ai.

Hello,

After the long holiday session I am finally back on track to start my next project. I think it’s about time I try tackling humanoid robot making. I been bouncing ideas with my kid. He originally wanted a robot dog which can play, talk, and clean up his toys. After a few weeks designing out the details then he comes back with, let’s make a transformer robot. I of course said well that sounds even more awesome than the last!

I was reading up the latest from CES this year and came upon a kick starter company that made just that. The robot is called T9 by Robosen and it was originally suppose to be Optimus Prime, but I assume they didn’t want to pay royalties to Hasbro and ended up with a generic truck look. However, the functions are still there. After seeing this I starting going down the youtube rabbit hole and ended up watching humanoid robot competitions.

This is when I learned that we have come a long way from the early walking robots. I have never built a humanoid robot before, but after seeing how so many people built really agile robots I am inspired to try it for myself.
Having my own humanoid robot is one of my dreams. All sorts of ideas is flowing through my mind. I can make the robot talk, walk around the house, maybe clean up toys, and put away the groceries.
Of course all that will take a lot of time and development. Which I do not have too much of with a working full time and raising a kid. So I plan to make this a long series, but great mini stories as I go along.

Planning

To build a humanoid robot from scratch requires a ton of up front planning and research. Some things I ask myself are:

How big is this robot?
What kind of parts do I need?
How much are the parts?

How are the joints connected?
How am I going to program it?
What are the features I need it to function?
Are there a pre-made kits or instructions?
What are the current techniques for training the robot?
How do I make the robot walk?

First I looked for existing works with guides to jump start my project. I found one project with some details of how it was built. This build had 17 servos controlling multiple joints. Each servo contributes to what is referred as degree of freedom (DOF) for short. The robot also had two other key components, a gyroscope accelerometer and servo driver. The gyroscope and accelerometer will give the robot a sense of balance and force. The server driver allows multiple servos be controlled from a single interface to something like a raspberry pi.

My next steps which I will talk about in the next post, will be the design of the robot. Such as where are my joints going to be and what kind of frame will I use to put all the servos together.

The post Making a Humanoid Robot Part 1 appeared first on Botmation's personal blog for creating robots and ai.

Thinking through the idea I want the plant to be able to say something to potential residents to know the moisture level is getting low. The plant could say things like “Hello, Yo, I need water, or Ahem”

Well this sounds annoying you say? well how about adding a feature to only go off during certain times of the day? Better yet lets add facial recognition to only bug the person responsible for watering the plant.

This idea is a great spring time activity to get your futuristic talking garden you always wanted.

I have uploaded the video to my youtube page.

For this project you will need the following.

• Speaker

Optionally for facial recognition you will need the following

• Raspberry Pi Camera V2

• Jetson Nano

STEP 1 Wiring Pi

Solder the pins to the ADS1115 board and connect to raspberry pi board on the following pins.

VDD – (Rpi 3.3v)

GND – (Rpi Ground)

SCL – (Rpi pin 3)

SDA – (Rpi pin 2)

Next wire the moisture sensor

VCC – (Rpi 3.3v)

GND – (Rpi Ground)

Aout – (ADS115 A0)

Once the wiring is complete go ahead and hook up the usb audio adapter and speaker.

STEP 2 Run the code

Power up the raspberry pi.

Go to my github page and get the latest files for this project. Download the Plant.py file for the raspberry pi.

https://github.com/Botmation

You should now see current reading from the sensor. In the air you may see values above 1200. If you dip the sensor into water you will see values below 600.

The code also includes mqtt function which listens for alerts from the nano. This is optional and won’t interfere if you do not have the nano.

Facial Recognition with Nano

If you like to use the facial recognition feature download the nano.py from github site.

For initial setup of the Nano please watch my video which walks you through the setup and training the program to recognize your face.

Once you train the program you can then run the Nano.py using the trained file as reference.

You will need to edit the code and update the network IP address to point to your raspberry pi.

The post Talking Thirsty Plant appeared first on Botmation's personal blog for creating robots and ai.

Video https://youtu.be/PttoKt6TMDk

GOAL: Running facial recognition in OpenCV 4.1 with Contrib lib over python using Nano and camera.

Disclaimer: This will uninstall OpenCV 3.3 that is preloaded on Jetpack for Nano and upgrade it to OpenCV 4.1. Worst case is it will break OpenCV and you will have to reload image. If you do not want facial recognition or other libraries in the Contrib build then you do not need to upgrade to OpenCV 4.1.


What you will need.

• Jetson Nano
• 32 Gb MicroSD card
• 4 Amp power supply
  • https://www.adafruit.com/product/1466
• Raspberry Pi Camera
  • https://www.raspberrypi.org/products/camera-module-v2/
• Cable Jumper Female to Female or jumper pin
  • https://www.amazon.com/LGQing-Cables-Jumper-Breadboard-Colorful/dp/B07RL12YFT/ref=sr_1_19?keywords=gpio+jumper+female+to+female&qid=1558471201&s=gateway&sr=8-19
  • https://www.amazon.com/TOTOT-Standard-Connection-Motherboards-Accessories/dp/B07H5F1JLC/ref=sr_1_9?keywords=circuit+board+jumper+pin&qid=1558471256&s=gateway&sr=8-9
• Mouse
• Monitor
• Keyboard
• HDMI cable

Preparing MicroSD card You will need to prepare the SD card using software from this link. https://www.sdcard.org/downloads/formatter/ Download and install the software. Then perform a quick format of the SD card with no name.
Install the Jetson Image Download the image from Nvidia web site. https://developer.nvidia.com/embedded/dlc/jetson-nano-dev-kit-sd-card-image
Download the image writing software and install on your computer. https://www.balena.io/etcher/
Use the Etcher software and load the image to the SD card. No need to unzip Jetson image.
Connect Camera to Nano Now connect the Raspberry Pi camera to the Nano.
Starting up Nano Insert the SD card into the Nano. Set the jumper on the Nano to use 5V power supply rather than microSD.

Connect Monitor, mouse, and keyboard. Connect power supply to Nano and power it on.
Create user name and password.

Increase System Memory
In order to install OpenCV 4.1 on Nano we need roughly 4Gb of additional memory. Otherwise the program will crash. Run the code below. Ensure you are using at least a 32 Gb SD card.

sudo fallocate -l 4.0G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile

Uninstalling OpenCV

sudo apt-get purge libopencv*

Installing OpenCV 4.1
Open Terminal window and browse into a folder where you want OpenCV to download and compile. Then run the following code.

sudo apt-get update
sudo apt-get install -y build-essential cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev
sudo apt-get install -y libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev
sudo apt-get install -y python2.7-dev python3.6-dev python-dev python-numpy python3-numpy
sudo apt-get install -y libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libjasper-dev libdc1394-22-dev
sudo apt-get install -y libv4l-dev v4l-utils qv4l2 v4l2ucp
sudo apt-get install -y curl
sudo apt install -y python3-pip
sudo apt-get update

wget https://github.com/opencv/opencv/archive/4.1.0.zip -O opencv-4.1.0.zip
wget https://github.com/opencv/opencv_contrib/archive/4.1.0.zip -O opencv-contrib-4.1.0.zip
unzip opencv-4.1.0.zip
unzip opencv-contrib-4.1.0.zip
cd opencv-4.1.0/

mkdir release
cd release/
cmake -D WITH_CUDA=ON -D CUDA_ARCH_BIN="5.3" -D CUDA_ARCH_PTX="" -D  OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib-4.1.0/modules -D  WITH_GSTREAMER=ON -D WITH_LIBV4L=ON -D BUILD_opencv_python2=ON -D  BUILD_opencv_python3=ON -D BUILD_TESTS=OFF -D BUILD_PERF_TESTS=OFF -D  BUILD_EXAMPLES=OFF -D CMAKE_BUILD_TYPE=RELEASE -D  CMAKE_INSTALL_PREFIX=/usr/local ..
make -j3
sudo make install
sudo apt-get install -y python-opencv python3-opencv

sudo apt-get install -y libjpeg-dev 
pip3 install -y --user pillow

When the installation is complete. Run the code below to test the version of OpenCV you have.

python

import cv2
print cv2.getBuildInformation() 

Generate Test images
In the location where you want to store the following example code.
Create a new folder called “dataset”, but do not browse into it.
All three python scripts for generation, training, and recognition are to be saved in the same directory.
This code will take images of your face using the Raspberry pi camera. The results will be stored into the “dataset” folder. Then the next script will do training on it.
Run the following code

import cv2
 import os
 cam = cv2.VideoCapture('nvarguscamerasrc !  video/x-raw(memory:NVMM),width=1280, height=720, framerate=21/1,  format=NV12 ! nvvidconv flip-method=2 ! video/x-raw,width=960,  height=616 format=BGRx ! videoconvert ! appsink' , cv2.CAP_GSTREAMER)
 face_detector = cv2.CascadeClassifier('/usr/local/share/opencv4/haarcascades/haarcascade_frontalface_default.xml')
 # For each person, enter one numeric face id
 face_id = input('\n enter user id end press <return> ==>  ')
 print("\n [INFO] Initializing face capture. Look the camera and wait ...")
 # Initialize individual sampling face count
 count = 0
 while(True):
     ret, img = cam.read()
     img = cv2.flip(img, 1)
     gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
     faces = face_detector.detectMultiScale(gray, 1.3, 5)
     for (x,y,w,h) in faces:
         cv2.rectangle(img, (x,y), (x+w,y+h), (255,0,0), 2)    
         count += 1
         # Save the captured image into the datasets folder
         cv2.imwrite("dataset/User." + str(face_id) + '.' + str(count) + ".jpg", gray[y:y+h,x:x+w])
         cv2.imshow('image', img)
     k = cv2.waitKey(100) & 0xff # Press 'ESC' for exiting video
     if k == 27:
         break
     elif count >= 30: # Take 30 face sample and stop video
          break
 # Do a bit of cleanup
 print("\n [INFO] Exiting Program and cleanup stuff")
 cam.release()
 cv2.destroyAllWindows()

Train Test images
This script will train the machine learning program to recognize your face.
Create a new folder called “trainer”, but do not browse into it.
Run the code below.

import cv2
 import numpy as np
 from PIL import Image
 import os
 # Path for face image database
 path = 'dataset'
 recognizer = cv2.face.LBPHFaceRecognizer_create()
 detector = cv2.CascadeClassifier("/usr/local/share/opencv4/haarcascades/haarcascade_frontalface_default.xml");
 # function to get the images and label data
 def getImagesAndLabels(path):
     imagePaths = [os.path.join(path,f) for f in os.listdir(path)]    
     faceSamples=[]
     ids = []
     for imagePath in imagePaths:
         PIL_img = Image.open(imagePath).convert('L') # convert it to grayscale
         img_numpy = np.array(PIL_img,'uint8')
         id = int(os.path.split(imagePath)[-1].split(".")[1])
         faces = detector.detectMultiScale(img_numpy)
         for (x,y,w,h) in faces:
             faceSamples.append(img_numpy[y:y+h,x:x+w])
             ids.append(id)
     return faceSamples,ids
 print ("\n [INFO] Training faces. It will take a few seconds. Wait ...")
 faces,ids = getImagesAndLabels(path)
 recognizer.train(faces, np.array(ids))
 # Save the model into trainer/trainer.yml
 recognizer.write('trainer/trainer.yml') # recognizer.save() worked on Mac, but not on Pi
 # Print the numer of faces trained and end program
 print("\n [INFO] {0} faces trained. Exiting Program".format(len(np.unique(ids))))

Running Facial Recognition
Now comes the good part. Your final file directory should look like this.

Run the following script in the base directory. The program should now recognize your face.

import time
 import sys
 import cv2
 import numpy as np
 import os


 print ("OpenCV "+cv2.__version__)
 recognizer = cv2.face.LBPHFaceRecognizer_create()
 recognizer.read('trainer/trainer.yml')
 cascadePath = "/usr/local/share/opencv4/haarcascades/haarcascade_frontalface_default.xml"
 faceCascade = cv2.CascadeClassifier(cascadePath);
 font = cv2.FONT_HERSHEY_SIMPLEX
 #id counter
 id = 0
 # names related to ids: example ==> yourname: id=1,  etc
 names = ['None', 'Botmation']
 cam = cv2.VideoCapture('nvarguscamerasrc !  video/x-raw(memory:NVMM),width=1280, height=720, framerate=21/1,  format=NV12 ! nvvidconv flip-method=2 ! video/x-raw,width=960,  height=616 format=BGRx ! videoconvert ! appsink' , cv2.CAP_GSTREAMER)
 # Define min window size to be recognized as a face
 minW = 0.1*cam.get(3)
 minH = 0.1*cam.get(4)
 ret, img=cam.read()
 while ret:
 while (delaytime < 0.5)&ret:
  ret, img=cam.read()
 #ret, img =cam.read()
 img = cv2.flip(img, 1) # Flip vertically
 gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
 faces = faceCascade.detectMultiScale(
  gray,
  scaleFactor = 1.2,
  minNeighbors = 5,
  minSize = (int(minW), int(minH)),
        )
 for(x,y,w,h) in faces:
  cv2.rectangle(img, (x,y), (x+w,y+h), (0,255,0), 2)
  id, confidence = recognizer.predict(gray[y:y+h,x:x+w])
  #Looks for a specific person
  # Check if confidence is less them 100 ==> "0" is perfect match
  if (confidence < 40):
   id = names[id]
   confidence = "  {0}%".format(round(100 - confidence))
     else:
   id = "unknown"
   confidence = "  {0}%".format(round(100 - confidence))
   cv2.putText(img, str(id), (x+5,y-5), font, 1, (255,255,255), 2)
  cv2.putText(img, str(confidence), (x+5,y+h-5), font, 1, (255,255,0), 1)
   cv2.imshow('camera',img)

 k = cv2.waitKey(10) & 0xff # Press 'ESC' for exiting video
 if k == 27:
 break
 # Do a bit of cleanup
 print("\n [INFO] Exiting Program and cleanup stuff")
 cam.release()
 cv2.destroyAllWindows()

Additional resources and References
Facial Recognition on OpenCV
https://www.hackster.io/mjrobot/real-time-face-recognition-an-end-to-end-project-a10826

Jetson Nano quick start guide
https://developer.nvidia.com/embedded/learn/get-started-jetson-nano-devkit

The post OpenCV 4.1 Contrib Lib on Jetson Nano Quick Start Guide with Facial Recognition appeared first on Botmation's personal blog for creating robots and ai.