SENSELET

SENSELET: a Sensory Network infrastructure for Scientific Lab Environments

The cloud service 4Ceed and the BRACELET network accelerates the process of making scientific discoveries by providing researchers with the convenience to upload, examine, and process their experimental data (e.g., microscope images) and metadata (e.g., microscope settings). For researchers, equally important information towards correct scientific experimentation, besides instrument raw data and metadata, is sensory data around the instruments when experiments are conducted. For example, the ability to capture and control laboratory environmental sensory information such as temperature, humidity, vibration is crucial for nanofabrication. In some laboratories, we have a few stand-alone sensors to collect humidity data. However, it is very time-consuming to manually collect and correlate those parameters with our fabrication process.

Access to environmental sensory data is also important to laboratory managers. The prices of Transmission Electron Microscopes, Scanning Electron Microscopes, Atomic Force Microscopes, and other instruments range from tens of thousands of dollars to millions of dollars. And these devices are sensible and vulnerable to environmental changes, such as overheat, water leakage, or power supply disturbance. Therefore, environment monitoring and real-time emergency alert are of vital importance.

With regard to these motivations, we design a sensor network architecture for scientific lab environments. Our goal is to (a) deploy diverse wireless and scalable sensory infrastructure in experimental labs, close to scientific instruments, and (b) correlate and synchronize sensory data with cloud-based instrument data and metadata in real-time and on-demand. SENSELET infrastructure will provide additional contextual measurements that will increase accuracy and causal relations of scientific results for scientists and better environmental monitoring and control of scientific labs for lab managers.

The SENSELET infrastructure includes (a) wireless sensors such as humidity, temperature, vibration sensors, (b) a wireless edge device with multiple wireless interfaces including Zigbee, NFC (Near Field Communication), BLE (Bluetooth Low Energy), Wifi, residing in the campus building lab, and (c) cloud service in cooperation with 4CeeD cloud to store and correlate sensory data with instrument data in real-time or on-demand. SENSELET will play a foundational role in providing crucial information. It will be closely integrated with BRACELET network infrastructure and 4CeeD cloud infrastructure, already embedded in our science labs.

 

The project consists of two parts:

Senselet Edge Device:

The edge device is located in the same room as sensors. It connects to sensors with wireless communication protocols (in this setting, Wifi). The edge collects sensory data, does pre-processing, and upload them onto cloud service. This pre-processing includes emergency detection, emergency alert, compressive sensing, and intrusion detection.

Cloud Side:

At the cloud side, a database is created to hold time series data. A visualization tool is provided for users to better monitor the sensor status and room environment. Configuration of Senselet and sensors is provided through a user interface.

Update:

Holonyak Micro & Nanotechnology Lab clean rooms will be outfitted with state of the art IOT sensors providing real-time mission-critical data for staff on humidity and temperature readings.

 

Localization:

Wi-Fi Fingerprint-based Localization System for Cleanrooms

Indoor localization has been a popular research topic for years. Many indoor localization systems have been proposed, and one of the most widely-used technologies is the Wi-Fi fingerprint-based method. The main idea is leveraging the diversity of received signal strength (RSS) from WLAN access points (APs) to create a received signal strength indicator (RSSI) fingerprint for each location. This method can achieve good localization accuracy and requires no additional infrastructure besides the fingerprint database.

For the scenario of cleanrooms, it is necessary for lab managers to have a location tracking system, which keeps the location record of lab users. This system can provide different services such as access control, equipment usage tracking, and evacuation monitoring in disasters. With this motivation, we design an indoor localization system for cleanrooms.

Localization Method

As shown in the figure, assume that the signals of multiple APs cover the user’s location. The user’s mobile device measures the RSS from each AP, and uploads the RSSI values to the server. The server estimates the user’s location, and replies with the localization result, or stores the location data for the record.

The process of localization consists of offline data collection and training phase and an online localization phase. In the offline data collection and training phase, we divide the localization region into grids with constant size and measure the RSS from the APs at the center of each grid. All the fingerprint information collected will be uploaded to a SENSELET server, and the server constructs the fingerprint maps. In the online localization phase, the lab user’s mobile device keeps measuring RSS for a certain duration. Then the mobile device uploads the measured result to the SENSELET server, and the server locates the user based on the RSSI fingerprint maps.

Handling Mobile Device Diversity

The performance of the RSS-based indoor localization methods can be impacted by the diversity of users’ mobile devices. Due to different Wi-Fi chipsets, antenna types and encapsulation materials, the RSSIs output by different devices in the same position can still differ. Since the devices used to generate the RSS fingerprint maps during the offline phase (training devices) are usually different from users’ devices during the online phase (tracking devices), the localization accuracy can drop significantly.

We focus on the scenario of cleanrooms and propose a new device diversity elimination method based on door-entering detection and linear regression. As shown by the figure, based on the RSSI output by the tracking device, our method detects the moment when the user goes through the door of a locker room. Then, we apply the linear regression between the RSSI output of the tracking device and that of the training device at the door. The result of the linear regression is used to amend the future RSS measurement of the tracking device, and thus the difference between the two devices is eliminated. This process runs during the online phase, and no extra user effort is needed. Another advantage of our method is that, besides the Wi-Fi chipsets used for the RSS measurement, we do not require any additional sensor, which is practical for more mobile devices.

Acknowledgment: This research was funded by the National Science Foundation (award number 1827126). The opinions, findings and conclusions or recommendations expressed in this paper are those of the authors and do not necessarily reflect the view of the National Science Foundation.