Weekly report: January 11 - January 18, 2018

Transcription

1 Weekly report: January 11 - January 18, 2018 Yerbol Aussat January 18, Activities Set up SPI communication between Onion Omega2 and AMN PIR motion sensor. Made a standalone sensing module on Onion Omega2 Collected Current vs Illuminance vs Control value (PWM) data for a DC bulb. Conducted two tests. Looked into integration of computer vision capabilities into the smart lighting system Walked through the current state of the smart lighting project with Sasha. Demonstrated how to use Raspberry Pi, Onion Omega2 and Node-RED. 2 What I learned (insights) This week I was focusing on three areas: Standalone sensing module on Onion Omega2 Power vs Illuminance measurements Computer-vision-based approaches 2.1 Standalone sensing module running on Onion Omega2 This week Sasha and I worked on getting PIR sensor readings with Onion Omega2 microcomputer. Since Onion Omega2 has only digital inputs, and AMN is an analog sensor, we used MCP3008 analog-to-digital converter was used, which communicates with devises via SPI protocol. The SPI communication was successfully set up, even though we ran into a number of problems during the process due to the poor documentation of this subject on the Onion s website and lack of examples on the Internet. OnionSpi python library was used for establishing SPI communication [2]. However, this library is a low level library that doesn t provide a high level 1

2 abstraction. It s API only provides functions for writing an dreading bytes to specific addresses on an SPI device. Sasha asked a question on the Onion Community forum [3] regarding SPI addressing, and within one day we got a response with a great explanation and sample Python code. Such a quick response rate was quite surprising and reassuring at the same time. Another issue we ran was due to misleading documentation of onionspi library on the Onion s website. According to the website, any GPIO s can be configured as SPI GPIO s. It turned out that only specific GPIO s can be used for SPI purposes. As shown in Figure 1, default SPI GPIO s on Onion Omega2 are 6,7,8 and 9. GPIO-6 can be used for CS/SS signal, GPIO-7 - for SCK/CLK, GPIO-8 - for MOSI, and GPIO-9 - for MISO. However, the expansion dock that we use doens t have GPIO 9 (Figure 2). By trials and errors, it was determined that GPIO 12 is the only GPIO that can be used for MISO. Figure 1: Default SPI ports on Onion Omega2 [1] Figure 2: Onion Omega2 used in the lab After figuring it out, we were able to successfully get readings from AMN with Onion Omega2. The standalone per desk sensing module is shown on Figure 3. 2

3 Figure 3: Standalone Sensing Module that is capable of sensing occupancy and light level 2.2 Power vs Illuminance measurements Costin instructed me on how to measure AC current drawn by the bulbs. It is done by using an Ampere-Meter to measure the current drawn by bulbs and the Onion Omega (F109) microcomputer that sends pwm signals to the DC bulbs. For measuring light levels REED ST luxmeter was used. In the future experiments the TSL light sensor will be calibrated and used for light level measurements. Figure 4: Setup for current measurements Next, two experiments of measuring Power vs Illuminance vs Control value (PWM Duty cycle) were conducted for one of the DC bulbs. The results of the experiment 1 are shown in Figures 5, 6 and 7. A similar result was obtained in experiment 2. The goal of this experiment was to get familiar with the procedure of getting measurements, and determine the relationships between control value, power and illuminance. 3

4 Figure 5: Illuminance vs Power Figure 6: Power vs PWM Duty Cycle Figure 7: Ulluminance vs PWM Duty Cycle 4

5 2.3 Computer-vision-based approaches I looked into integration of computer vision capabilities into the smart lighting system. Firstly I installed python-camera library on Raspberry Pi, and was successfully able to take pictures and videos with it. Also, the camera has infrared filter, which makes it good for taking pictures in low light environments. In the next few weeks I am planning to explore two approaches: the traditional approach using OpenCV computer vision library, and a deep learning - based approached using the Movidius neural compute stick. This week I was able to successfully install and run the latest version of OpenCV python library, using the guide from [4]. I was surprised that OpenCV can be run on Raspberry Pi.The big advantage of OpenCV is that it is very well documented. Also, I found numerous projects on the Internet on different kinds of object detection, recognition and tracking applications using OpenCV and Rasperry Pi. I think at the moment it is the most popular library for computer vision applications. Movidius neural compute stick and SDK is a very recent technology, and its documentation is not as rich. Though it is a very promising approach. TensorFlow library can be used on Movidious NCS. 3 Next Set up TCP/IP communication between a client on Node-RED (Raspberry Pi) and a python server on Onion Omega2, so that Raspberry Pi receives all sensor readings from the sensing module. I think this would be a good task for Sasha to work on. Through this task he would be able to build up some experience in using Raspberry Pi, Onion Omega as well as creating a socket server with python and a client on Node-RED. Conduct Power vs Illuminance vs Control value experiments for the Philip Hue bulbs Come up with a plan for experiments, i.e. decide on which bulbs to use in the experiments and at which distances to measure the light level values. Possibly conducting experiments with combinations of bulbs. Further explore uses of OpenCV and Movidius on Raspberry Pi 4 References. [1] Communicating with SPI Devices [ [2] SPI Python Module [ 5

6 [3] Onion Community forum [ spi-addressing-and-0x00-returned-values] [4] OpenCV installation guide [ 6