#Runtime_Terrors| Sleep Shift Scheduling Tool
Awards & Nominations
#Runtime_Terrors has received the following awards and nominations. Way to go!
Sleep Shift Scheduling Tool
Cosmorepophilia - The cosmic sleep assistant for the cosmic travelers
Summary
Over the years of our advancement into space as a space-faring species, our understanding of biological reactions and impacts over the human body is steadily hiking. Several sleep suggestive mechanisms and hacks are suggested and developed. However, despite the technical assistance provided for the cosmo travellers, inducing sleep still remains a challenge. Our project aims at making a sleep schedule, nutrition controlling and also fitness recommendation system based on the stat of the art Deep Learning techniques. The recommendations issued by public health authorities are important for surveillance and help to inform the population of interventions, policies, healthy diet and healthy sleep
How We Addressed This Challenge
Several sleep suggestive mechanisms and hacks are suggested and developed. However, despite the technological assistance provided for the cosmo travellers, inducing sleep still remains a challenge. Technology has no means of interfering with the natural cycle of the human
Sleep suggestive and scheduling tools do help in customized suggestions for users to get a healthy and balanced sleep. However, considering the exclusive case of shift workers and astronauts, their nature of work is highly tentative. Over generations, our body got habituated to the classical 24 hr body cycle, where one reposes on night phase and works during the day. Despite the very fact of getting habituated, our body produces vital rejuvenating and life supporting hormones in this phase, like Melatonin. For a job like Astronaut and life saving roles like shift working nurses, attentiveness and scrutinization of the work is very mandatory. A minute of fatigue can turn out to be fatal.
In the medications and drug market, there are a huge range of sleep inducing and sleep preventing medications. The march towards technological advancement and the hesitant need to work over time for meeting the needs, pushed many of us to use these medications. However, are they absolutely safe for utilization in the long run? Statistics from google trends states that there is a steady hike in utilization of sleep inducing drugs among the public.
This case is even more concerning when it comes to astronauts. It is studied that astronauts use about 3 to 4 times more sleep inducing and preventing drugs. As stated by Medicinenet.com, a long term utilization of these drugs has profound and severe effects on the consumers.
Our project makes use of state of the art Deep Learning model to predict the amount of sleep and based on the amount of sleep, we also predict the food that the astronaut needs to consume and also the amount of exercise he needs to do to keep himself fit. We finally package it out into a very appealing GUI which also makes a timetable with a daily diet chart and a daily exercise schedule that is most optimum for the astronaut keeping in mind all the important factors including sleep, work, diet, exercise, medications, preflight and inflight sleep conditions and also the noise and slam shift.
To round it all up, we also provide hardware support in the form of an innovative and versatile sleep inducing mask - Cosmorepose
After extensive research and key engineering intuition applied on the non drug based search for solution, primarily focusing on accessibility and ease of setup and use, we embarked on a mission to develop a pcb infused sleep mask solution. Quality, feasibility and TRL were few of the primary scrutinization of our model design phase. Several studies done by prolific researchers around the world, indicated that light therapy was one of the leading solutions prevalent among the community for both sleep prevention and sleep induction. It was found that studies done on several overseas Olympics participants, experiencing jet lag, had a very positive output when subjected to light therapy. From a biological standpoint, exposing humans to certain lux(preferably high intensity of light), helps them adjust their circadian rhythms and induce better sleep. In space, astronauts are subjected to a lot of sleep preventing problems. Noise, light, Space day and night cycles and ambient temperature. Astronauts in ISS have a day and night cycle repeating every 90 minutes. Series of fluctuating lights, to which the astronauts are exposed, severely inhibits them in getting a good and fast sleep. Hence, this leads to hike in the sleep latency and thus lack of healthy sleep. Sleep also depends on nutrition intake, work burden and several other minor factors
How We Developed This Project
We know the fact that Neural Networks perform very well when it comes to classification problems. But instead of going for the orthodox regression solutions, we planned on using Neural Networks for regression to predict the amount of sleep. We preprocessed the data into a more comprehensible format and fed it into the model. Based on sparse data from NASA and some tedious research, we found the correlation between the parameters and made a dataset of 100,000 datapoints. Here is an example of the first 5 rows of our dataset:
We fed it into the Neural Network with a ‘normal’ kernel initializer and ‘relu’ as activation function for the hidden layers. We used ‘mean absolute error’ as our loss function and the output layer had a single linear node. We trained the model with a validation split of 0.4 and 10,000 epochs to make sure that the model is as accurate as possible. Here is the summary of our final model that predicted sleep:
This is the final graph for Prediction Error vs Count. As you can see, the graph is almost a Gaussian which proves the fact that our model has a very high accuracy coupled with a low loss.
Because of the current pandemic all our team members are located in different cities, which made it tough for us to collaborate. So we decided to break the work into modules and each of us worked on a particular module. With 2 of the members working on the GUI, 2 others working on the Deep Learning model and diet and fitness recommenders and one member working on hardware, electronics and the CAD designs we tried to optimize our workflow as efficiently as possible.
Considering the current advancements in the sleep technology and realizing the need to develop an non drug based sleep inducing solution, pur design incorporates a fusion of sleep inducing high intensity led and vital measurement sensors to aid the astronauts to track and monitor their sleep. Our design houses a PCB based control, monitoring the LED, the pulse oximeter, the accelerometer and the transmission module. Sleep study on light therapy performed on several human subjects indicates that red wavelength of light helps in inducing a faster sleep. The actigraphy inspired sensor components collect real time data from the user about his sleep. Human sleep is divided into Normal, Deep and REM cycles. The following criteria were set to determine the phase of sleep.
- if movement is more and pulse is normal, then high intensity of light - Normal phase
- if movement is less and pulse is normal, then medium intensity of light - Deep phase
- if movement is less and pulse is high, then the led goes off - REM phase
- Pulse is normal - 60 to 100 beats per minute
- Pulse is abnormal - + or - 15 beats per minute
- Temperature normal - 37 C
- Temperature upper limit - 41 C
- Movement - Boolean
Considering this criteria, a conceptual arduino sketch and circuitry connections were done, to support our design ideation.
Working mechanism:
Working for NASA means thinking like NASA. When it comes to control and data acquisition, adding redundant checks for the activation of the model, helps in accurate utilization of the design.
- The user turns on the slide switch. This is checked against temperature sensor data, if the slide switch is turned on and if the temperature is within the limits of 33 C and 42 C, then the data acquisition initiates. This redundancy is ensured to start data acquisition only if the switch is turned on and when the person is wearing it(by detecting his body temperature)
- Post initiation, data from the potentiometer(mimicking pulse oximeter) and tilt sensor is fused to control the lighting of LED. **It is to be noted that the data from the pulse oximeter and accelerometer is constantly being transmitted via the transmission module incorporated in the PCB** As specified above, depending on the criteria, the appropriate sleep phase is predicted and LED light intensity is controlled.
- The data transmitted from the mask to the control unit, will later help in suggesting appropriate sleep schedules for the user - a future prospect, expected to be realized via Reinforcement learning
Apart from the technical consideration, this wearable sleep inducing mask is very flexible and comfortable to wear. It is aesthetically and ergonomically seamless, aiding the user to sleep with utmost comfort whilst satisfying the purpose of suggestive sleep data collection.
How We Used Space Agency Data in This Project
As we stride towards technological advancement, the work input from the human resources is going to steadily hike. Trends from Google is a very good support for this claim.
The hike in the shift workers over the past 5 years in the United States alone is hiking year by year. This trend is expected to reach heights in a few years. We can’t compromise human health for technology, however that doesn’t mean that we can compromise advancement as well. Seeking for a perfect balance, and we as students being a part of the problem, very well understand the situation and set forth on designing this product.
Since we had a very sparse amount of data from NASA and also we emailed them regarding the same, we couldn’t completely make a dataset based on such a small amount of data so keeping it in mind we decided to come up with researched datasets for sleep prediction from kaggle. And also based on the data given to us by NASA we researched to find more data pertaining to the parameters that affect sleep. We compiled all the data to make a huge dataset of 100,000 training points. We have also decided to make the dataset public so that others can use the data.
Considering the hardware design, due to the relatively ease of use and manufacture with very high TRL, utilization of this sleep inducing mask is expected to be a significant change of track in the existing sleep inducing techniques. Our product is technological and economical friendly, ensuring quick adoption of this technology by the greater community. We as engineers, always have eyes over feasibility alongside ideation. Product was chiseled down to the current stage after extensive study of user cases for light therapy and its success rate. (Kindly glance at references for validation of our study).
When our product is coupled with the power of Machine Learning and Reinforcement Learning, it will revolutionize the existing technologies for sleep and will have a profound effect on the greater good.
Project Demo
https://youtu.be/HLhLESqKzyI - 30 seconds video
https://youtu.be/D5h9DFsRxlU - Local Judging video
Project Code
Data & Resources
- https://www.nasa.gov/feature/ames/nasa-research-reveals-biological-clock-misalignment-effects-on-sleep-for-astronauts
- https://docs.google.com/document/d/1SZv8_wEfDJ5Yoi1tZoWEUfSIJkrr9XXgCd0rMhWPnHI/edit
- https://www.nasa.gov/mission_pages/station/research/astronauts_improve_sleep
- https://www.medicinenet.com/hypnotics_drug_class_side_effects/article.htm#natural_herbal_hypnotics_and_hypnotic_supplements
- https://www.cochrane.org/CD003218/NEUROMUSC_drugs-that-increase-alertness-psychostimulants-for-excessive-daytime-sleepiness-hypersomnia-in-myotonic-dystrophy
- 1991. Research on sleep, circadian rhythms and aging_Applications to manned spaceflight
- 2016.Circadian misalignment affects sleep and medication use before and during spaceflight