1.0 Introduction

1.1 Problem and Background 

With institutions open across the country during this pandemic, we are increasing the risk of outbreaks. On-location activity is particularly important for traders who need robust, real-time communication and sales teams that are subject to specific compliance monitoring. In the event of an outbreak, an institution may not be able to accurately identify every employee/student at risk and in turn, can be forced to shut down completely. There is a huge opportunity cost due to a loss of learning potential/revenue from shutdowns.

The ability to control an outbreak and immediately trace down which people should quarantine is crucial. Recent studies have shown that 80% of transmissions can be prevented through immediate contact tracing. Manual contact tracing can take days, reducing the efficiency all the way to 5%. New innovations by Apple & Google to help with contact tracing aren’t working either, because only 3 in 5 report a willingness to use the software. These applications have ranges of hundreds of feet and in a closed setting would not efficiently track user interactions. Our technology can be used to help schools and workplaces make data-driven decisions such as what areas to monitor, which employees to send home, and how to predict when an outbreak will happen.

1.2 Solution summary in simple terms

OPEX is a startup focusing on providing solutions for contact tracing. Our long-term vision is to provide asset tracking solutions maximizing the efficiency of manufacturing infrastructure. 

Health Check

Through Health Check, employees answer questions about symptoms they may be experiencing to indicate their risk of having COVID-19. Our survey is logged daily and is formatted based on CDC guidelines.

Contact Tracing

Our contact tracing feature reports an office's overall safety, contact risk, COVID risk, & the risks for each employee. This information enables banks to control an outbreak by notifying them which employees may have COVID & where to schedule deep cleanings. 

Demo found in 2.1

Occupancy Detection

Analyzing Bluetooth signals received from personal devices such as wearables and phones enables Opex to report the occupancy of a room. If a room occupancy exceeds limits set based on size, staff are notified. Demo:

Key Benefits

For Institutions

• Disruptively Cost-Effective Solution for Contact Tracing.
• Fast & Easy Deployment: Deploy this solution within days.
• Safe: All hardware meets US standards & certification requirements.
• Endless Possibilities: Our platform is "expandable" so you can leverage your infrastructure investment for twin solutions in inventory management, asset tracking, and more.

For Employees

• Peace of mind for employees.
• Captures location & interaction events helping employees avoid routes through highly congested office spaces.
• The system maintains user privacy by randomizing data ID's.

1.3 Solution summary in technical terms

Bluetooth signals are sent as electromagnetic waves from beacons to receivers. Beacons come in many shapes and forms, including wearables, phones, watches, and more. Using Received Signal Strength Indicator (RSSI), a relative measurement indicator of the strength of a signal we have developed an advanced Neural Network that can predict a user’s location. OPEX’s contact tracing software operates in four steps. First, our beacons advertise Bluetooth signals. Second, these signals are detected by wave scanners. Third, these readings are sent to the cloud and inserted into our SQL database. Finally, OPEX’s proprietary deep learning model is used to evaluate distance values based on these readings. Using a sophisticated filtering algorithm, we can accurately determine occupancy and communicate this on OPEX’s Connected Platform. Our occupancy detection software incorporates the same Neural Network, determining if a user exists in a room based on the movement of a signal.

1.4 State of advancement of the project

After over 6 months in development, the Occupancy Detection solution is fully functional, and Contact Tracing is 1 month away from 100% accuracy within a 6 ft range over a period of 10-15 minutes (CDC definition for “contact”). The solutions are currently being modeled at our houses, with the ability to deploy immediately. After winning two multinational hackathons, Opex has been given permission to conduct beta tests of our technology at Interlake High School, accompanying the school’s reopening in April. Furthermore, the project has generated interest from Washington State University, The University of Washington, Mazik Global, North Idaho College, and Bellevue School District. 

1.5 Project Timeline

Week 1: Acquisition of supplies: 3 New Antennas, Receiver Cases, Longer Power Cables, and Azure Cloud Credits.

Week 2: Begin Trials at Interlake High School: Setup receivers in Interlake Labs, model data over the course of the week and compare to teacher accounts. Assess the accuracy using different Antennas and models.

            Continue home trials of contact tracing algorithm: Explore RTLS (Real-Time Location System) technology, compile a        list of hardware compatible with bearing software available to determine user location more accurately.

Week 3: Meet with Potential Clients, continue work with Interlake: Continue work with Interlake, collect data on the accuracy of occupancy detection on a large scale (20+ students) compared to trials with 1-6 people. Additionally, begin taking meetings with university representatives using Interlake as a PoC offering to self-fund beta tests.

            Evaluate & Integrate Contact Tracing: Assess the current level of certainty of the model, integrate into Interlake high         school beta tests.

Week 4: Follow up with Potential Clients, discussing the progress at Interlake. Develop new partnerships to implement technology in reopening universities and corporate offices.

2.0 Project Implementation

2.1 Solution?

Solution addressed above, find more details/demo in this video: 

2.2 Methodology

We took over 1 million data points from 24 different distances staggered between 0-6m to train the Neural Network. The Neural Network applies a parametric nonlinear regression model.

To maximize the accuracy of the distance estimations, we adjusted the signal window from 10 to 1 second. To reduce the number of RSSI measurements our system had to process, we collected a weighted average every 15 second interval.

Then, we collected the average of each 10 datapoints per distance:

This was the excel formula used:

=AVERAGE(OFFSET($F$353,(ROW()-ROW($AK$3))*10,,10))

The formula was:

After collecting the average for each 10 data points, we collected the average of all data points per distance:

This was the excel formula used:

=AVERAGE(AK3:AK500)

The formula was:

From Weighted Average, we were able to derive 26 different factors all incorporated into the Neural Network based on separate tests to determine their relationship to distance: WeightedRssi, AverageRssi, CountSignals. FilteredWeightedRssi, FilteredAverageRssi, IqrRssi, Q1Rssi Q3Rssi RangeRssi, MedianRssi, ActualDistance, EstimatedDistance, FilteredStandardDeviation, StandardDeviation, FilteredVariation, Variation, CoeffecientOfVariationFilteredWeighted, CoeffecientOfVariationWeighted, CoeffecientOfVariationFiltered, CoeffecientOfVariation, FilteredCoeffecientOfVariationFilteredWeighted, FilteredCoeffecientOfVariationWeighted, FilteredCoeffecientOfVariationFiltered, FilteredCoeffecientOfVariation, TotalCountDevicesRatio, TotalUsersInReading, TotalCount, Min, and Max. TotalCountDevicesRatio, TotalUsersInReading, TotalCount, Min, and Max.

Program scans for readings every tenth of a second. 

• Once it finds readings, (2 readings per scan) 
• Storing devices on memory
• Every 30 seconds call API that sends data to the Cloud 
• (About 600 readings) Data being inserted into Sequel table 
• Azure Worker (Function App) Convert Raw Data into Weighted Averages in the Cloud
• Second Azure Worker is predicting estimated distance for each data point. This is added to the final data table.
• We are going to: Use another Azure Function App that will go through all the different services and trace users in services by triangulating user distance between receivers. These distances will be compiled to determine employee proximity. 
• If interactions are recorded, they will be sent to the OPEX database. These interactions will be deleted from our platform after 14 days.

Referring to “Official Trial Data for Math IA” in additional documents, we observed the Neural Network estimate the location of beacons with an accuracy of ~1 foot. However, these tests were conducted over two months ago. With the use of antennas and a better-trained model, the uncertainty has been cut in half.

2.3 Results/Expected results

Ultimately, these are the results achieved by modeling the trial data available in additional documents. Following the purchase of antennas and RTLS compatible technology, we expect that more isolated data points from 15-second intervals will begin to display similar results.

3.0 Safety, quality assurance and regulation 

3.1 What steps have you taken to ensure your solution’s safety? How advanced are you in this process (if applicable)? Please check the Biosafety and Biosecurity guideline of OpenCovid19

Not applicable, digital solution.

3.2 Have you planned the conduct of your manufacturing process that ensures quality, what are the steps you have taken? How advanced are you in this (if applicable)?

Not applicable, digital solution.

3.3 Will you need assistance with the regulation system? If not, which regulatory system do you plan on using to distribute the product? Please elaborate (please see: Regulatory-Strategies) (if applicable)

Not applicable, digital solution.

3.4 Have you talked to medical staff about the feasibility of your project? What did they say?  

Yes, we have discussed the solution with Dr.Pallavi Patel, gastroenterologist at Kaiser Permanente who connected us with her colleagues at the University of Washington Medical Laboratories to integrate our occupancy detection solution. Additionally, she gave positive feedback for our occupancy detection solution, while expressing data privacy concerns regarding contact tracing.

3.5 Have you planned the testing, verification and validation of your solution? How advanced are you? (if applicable)

Yes, we have already done live demos/PoCs of our technology in our homes over the past 4 months and are beginning beta tests at Interlake High School in April.

4.0 Impact, issues and risks

4.1 What impact do you feel your project could have? (100 words max)

Opex can have a massive impact curbing the spread of the pandemic. As previously mentioned, immediate contact tracing can prevent up to 80% of transmissions, Furthermore, universities and other organizations are struggling with enforcing room occupancy limits, our technology saves them time and forces students/employees to be more accountable. It also allows individuals to stay safe at work by avoiding areas with a high level of activity.

4.2 What do you think would make your project a success?

We are in a good place right now considering our partnership with Interlake and our discussions with other prestigious institutions. Unfortunately, due to our age the quality and applications of our technology are overlooked. Funding from a respected organization such as J.O.G.L. not only allows us to integrate our technology into more locations, but also establishes the credibility of our project.

4.3 Issues

The biggest issue related to our project is data privacy. To address this issue, we shifted our procedures to pseudonymize data collection and prioritize the development of Occupancy Detection software, which does not collect any PII (Personally Identifiable Data).

5.0 Originality

5.1 What other projects on JOGL are like yours? Search for them and Link them!

After doing some research, it seems like there are many amazing JOGL projects in different laterals with similar objectives, but none that aim to increase the efficiency of contact tracing in close contact environments.

5.2 Is this an innovative project? What makes this project different if it’s unique on JOGL?

This project is innovative to the extent that it leverages incredibly advanced Neural Networks to evaluate the location of a user. The technology is not unique to JOGL, and is currently employed by industry competitors. However, these companies apply a proximity-based approach rather than using receivers and beacons. To simplify, proximity-based approach does not identify the location of a person, rather if another

person breaches the 6ft barrier between the two. Unfortunately, this limits the applications of information gained from the technology as it isn’t possible to learn which rooms a Covid-exposed employee has been in.

5.3 Is there already an open-source version of this project?

There are no open-source versions of this project but have been many papers by respected institutions regarding the application of Bluetooth Technology in contact tracing.

6.0 Team experience

6.1 Please cite your team members and their roles in the project. 

Ishan Sinha – CEO, Student at Interlake High School interested in pursuing a career in engineering. Currently taking advanced college level statistics courses, applying math knowledge to regression modeling. Also conducting meetings with organizations to implement technology.

Jacob Frumkin – CIO/CTO, High School Student at Bellevue College earning an associate’s degree in computer science. Possesses 10 years of experience in programming, with deep knowledge in machine learning. Jacob is currently applying reduction factors to the occupancy detection solution.

Disha Singh – Software Engineer, Graduate Student at UMass Amherst focused on computer science. Also possesses significant experience in programming, specializing in deep learning and focused on training the Contact Tracing Neural Network.

7.0 Funding and Costs

7.1 Please provide a costing of your project be as detailed as you can.

The main reason we need funding is to purchase the infrastructure for Interlake High School Beta Tests and self-fund trials at universities/corporate offices. In addition, RTLS (Real Time Location Services) technology will increase Opex precision to within inches, maximizing the efficacy of our contact tracing solution.

7.2 How is your project being funded so far?

So far, our project has been self-funded. The primary investment we have made so far is hundreds of hours of our own time to train the models with only 3 Raspberry Pi’s and 2 beacons. Now, as we implement the technology in official settings there is a significantly larger need for infrastructure. Unfortunately, with institutions on spending freezes they are reluctant to fund projects, especially those led by high school students so self-funding will help convince them to let us conduct Beta tests.

Behind Our Build

• React/Java Script/CSS (UI)
• SQL (Database Queries)
• C# (Data Processing & API)
• Python (Machine Learning & Data Collection)

Please review our sites for more context: 

https://devpost.com/software/opex

https://www.opexservices.com/