Self-evolving idea for smart materials design platform

Code：

Minimization algorithm

If you are not familiar with machine learning or have no coding experience, it is recommended to start with the minimization function compiled by scipy.optimize.minimize. Because in the whole loop built, it is a process of narrowing the gap between the intime state and the target state, and some simple problems can be solved with the simplest minimization algorithm. If you are more familiar with machine learning algorithms, you can use some more streamlined and faster algorithms.

Electronics：

Raspberry pi
Arduino

Using Raspberry pi to connect with multiple Arduinos can achieve the purpose of quickly controlling multiple voltage ports.

Raspberry pi ports

Connect Raspberry with Arduinos

##code part in python

from smbus import SMBus

addr1 = 0x04 #address of Arduino uno
addr2 = 0x05 #address of Arduino mega
bus = SMBus(1)

bus.write_block_data(addr1, 10, [3,5,6,9,10,11])

# pwm ports of Arduino uno

time.sleep(0.1)
bus.write_block_data(addr2, 10, list(range(2,12)))

# pwm ports of Arduino mega
time.sleep(0.1)

Thus you can individually control 16 ports which can be a simple version of multiplexing.

Aim:

Let LEDs self-evolve into any random brightness.

Required components:

Raspberry pi; Arduino; LEDs; resistance; jumper wires

Schematic diagram:

self-evolving schematic diagram

In order to demonstrate this idea, the first simple case is using LEDs. In this case, the target is 5 LED with different brightness which I don’t know. And 5 LED’s initial state is set at 0. The microcontroller is used to adjust the input current into the LEDs. The LEDs will try to change the brightness itself, whenever they change, a picture will be taken, the image analysis will tell it the difference between the intime brightness and target one, so it can figure itself the way to approach the target in this feedback loop. Finally, when it evolves pretty much like the target, it goanna stop and generate the output. This is the real experimental pic that I try, I think this case is quite fulfilled now.

Brightness detection:

brightness detection demo code:

def ReadLEDLightness(File):

    frame = cv2.imread(File)

    print(‘read image file: ‘ + File)

    PosX=[217, 273, 322, 388, 409]  #Position the Red LED

    PosY=[297, 301, 303, 300, 307]

    RectangleX = 10

    RectangleY = 10

    LEDLightness=[0 for ii in range(0,PortNum)]

    fout=open(“TestLED.txt”, “w”)

    for jj in range(0,PortNum):

        HAve=0.0

        LAve=0.0

        SAve=0.0

        for kk in range(-RectangleY,RectangleY+1):

            for ll in range(-RectangleX,RectangleX+1):

                RGB_Red=frame[PosY[jj]+kk,PosX[jj]+ll,2]/255.0

                RGB_Green=frame[PosY[jj]+kk,PosX[jj]+ll,1]/255.0

                RGB_Blue=frame[PosY[jj]+kk,PosX[jj]+ll,0]/255.0

                (H,L,S)=colorsys.rgb_to_hls(RGB_Red,RGB_Green,RGB_Blue)

                HAve=HAve+H

                LAve=LAve+L

                SAve=SAve+S

        HAve=HAve/((2*RectangleX+1)*(2*RectangleY+1))

        LAve=LAve/((2*RectangleX+1)*(2*RectangleY+1))

        SAve=SAve/((2*RectangleX+1)*(2*RectangleY+1))

        LEDLightness[jj]=LAve 

        for ii in range(0,3):

            fout.write(‘%8d’ % frame[PosY[jj],PosX[jj],ii])

        fout.write(‘%15.5f %15.5f %15.5f’ % (HAve,LAve,SAve))

        fout.write(‘\n’)

    fout.close()       

    return LEDLightness

Results:

For each function evaluation, it will take 0.5s, after 5 iterations, the brightness will reach the target.

Extra practice:

Using RGB LEDs to tune the color of your LEDs.