Doctoral research · Drexel University

LuxREST

Personal light exposure and feedback

A circadian-light sensing project connecting near-eye exposure monitoring with personal feedback using LiDo, cellular IoT, MQTT, n8n, a custom GitHub-hosted dashboard and Telegram.

The actual LiDo sensor clipped beside the right lens of the glasses and connected to its companion.Glasses-mounted LiDo
  1. CaptureMeasure light at the eye.
  2. TransferSave locally and send in batches.
  3. ProcessRoute readings to the assigned stream.
  4. FeedbackView a daily profile and ask questions.
How a reading moves from the wearable to the participant. The phone view is schematic. Open the animated pathway or try the working demo.

Interactive tools

The dashboard and feedback tools share a synthetic light profile.

From measurement to feedback

The light reaching a person’s eyes changes as they move between daylight, electric lighting, and screens. LuxREST was built to show that exposure throughout the day and give the person wearing the sensor a way to ask about their own measurements.

The project builds on the TRIAD evidence map, which connects biological responses, lighting systems, and occupant engagement. LuxREST brings those connections into a working system, from the near-eye measurement to the information a participant can act on.

Follow the operating pathway
LiDoNear-cornealmEDICompanionUSB bridgeWi-Fi / MQTTReceiverLocal receiptRaspberry PiCloudBatch routingAssigned streamDashboardOpen / refreshChart + gaugeTelegramScheduledcommunicationConversationQuestion-ledcontext retrievalSENSINGPRESERVATIONINFORMATIONParticipant access → interpretation → possible action
Diagram and motion from the dissertation presentation.
Read the source scenes
  1. System overview

    One record connects near-eye sensing, local storage, cloud routing, and personal feedback.

  2. Capture a reading

    The LiDo Companion requests a reading over USB and sends it over Wi-Fi using MQTT.

  3. Keep a local record

    The Raspberry Pi preserves the payload and its receipt metadata before the cellular gateway transfers a reduced stream.

  4. Route to the participant

    Cloud workflows assign each batch to the correct stream. The dashboard reads it, scheduled messages reach the participant, and questions retrieve the relevant context.

  5. Display the record

    Transfer, dashboard refresh, and the three-hour gauge summary each have their own timing.

Data pathway and component responsibilities
Responsibilities along the verified data route
StageInput and output
LiDoTimestamped optical values, including mEDI
USB companionRequest readings and package wireless messages
Portable receiverMQTT receipt, payload and receipt metadata
Downstream serviceSelect alternating formatted records; scheduled handoff around 15 minutes
Cloud routingUnpack batches and map records to the assigned participant stream
InterfacesRetrieve stream data for a dashboard or communication context

Dissertation · October 5 release · Chapter 4 · system realization

Hardware and interfaces

Measure where light reaches the eye

Placement and direction affect a light reading. I selected the spectacle-mounted LiDo because it keeps the sensing plane near the eye and follows head orientation. Its spectral measurements support melanopic equivalent daylight illuminance (mEDI), expressed in lux. This is the illuminance of standard daylight that would produce the same melanopic irradiance as the measured light.

The LiDo sensor sits on the right arm of a spectacle frame and connects by USB to a separate companion.
The field setup places the sensor beside the eye and leaves the companion separate from the frame.

Preserve the reading at each handoff

With HSLU, I developed the companion hardware extension that connects LiDo’s USB interface to the wireless receiver. I developed the software route from acquisition through the participant interfaces. The receiver kept the payload and receipt metadata before a reduced stream was sent downstream. The cellular gateway attempted cloud batch uploads about every 15 minutes during the pilot.

The implementation used Pico W and Raspberry Pi for acquisition, MQTT for transport, Notecard and Notehub for cellular handoff, and n8n for automation.

Inspect the measurement-to-feedback evidence route

The receiver retained alternating successfully formatted records for downstream transfer. The cloud upload interval was configured separately from local measurement and receipt.

See the animated data handoff

Give participants a view of their day and a way to ask questions

I built the GitHub-hosted dashboard to load the assigned stream when participants opened or refreshed it. Date and resolution controls let participants inspect daily patterns against fixed clock-time reference levels. The gauges summarized readings from the preceding three hours.

The Telegram workflow supported scheduled messages and an integrated AI assistant chatbot that retrieved the participant’s context when answering questions about their light exposure.

The interactive dashboard below is the later Observatory design. It uses synthetic readings, while the 2025 field dashboard displayed the assigned participant stream.

Dashboard and messaging routes
The original workflow separates a dashboard fetch, a scheduled message and a question-and-response exchange. Each provides participant-facing information and leaves a different evidence record.

Dissertation · October 5 release · Figure 4.3

Try the Observatory dashboard and feedback

The Observatory is a later redesign that uses a synthetic light profile. Edit the profile in the dashboard, then ask the feedback tool about the same day.

Interactive tool · connected system

Explore LuxREST in action.

Shape a day of light exposure, then follow that same profile into a dashboard and a question-led feedback workflow.

Light dashboard
Feedback workflow

Editable demonstration using a synthetic profile and source-based messages. Replies are calculated in your browser.

How the field pilot worked

In the Philadelphia pilot, participants wore the sensor in daily life from June through September 2025. Each participant first wore the sensor during passive monitoring, then gained access to the feedback system. Morning and evening surveys complemented the light record, and a post-study survey asked about the participant experience.

  1. BaselineWear the sensor and report daily routines.
  2. ActivationIntroduce the participant’s dashboard and feedback.
  3. FeedbackContinue wearing the sensor with the information available.
Participants moved through these phases on their own schedules. The principal exposure comparisons excluded windows assigned to the activation date.

Compare exposure in windows tied to behavior

Daytime exposure was aligned to waking, and evening exposure to the three hours before attempted bedtime. For each window, threshold duration was divided by valid observed time and expressed as minutes per observed hour. Phase comparisons gave each paired participant equal weight.

Study timeline, eligibility rules, and sensitivity checks
Each row is one completer’s official timeline. Blue is baseline, orange is activation and teal is feedback. The horizontal axis gives Philadelphia local calendar dates.

Dissertation · October 5 release · Figure 5.1

Principal exposure-window definitions
RuleDaytime H1Pre-bedtime H2
Behavioral windowWake to 3 h before next attempted bedtimeExactly 3 h before attempted bedtime
Exposure countedmEDI ≥250 luxmEDI >10 lux
Observed-time eligibility≥25% coverage and ≥3 observed hours≥60 observed minutes
Maximum contributing gap120 seconds120 seconds
Result unitMinutes per observed hourMinutes per observed hour
Phase assignmentWaking date; activation excludedActual bedtime date; activation excluded

Dissertation · October 5 release · Section 3.7 · core personal-light outcomes

Eligible days or nights were first averaged within participant and phase. I then calculated each participant’s feedback-minus-baseline contrast and averaged those paired contrasts. This kept one participant with many valid occasions from dominating another with fewer occasions.

A 10,000-replicate participant bootstrap supplied 95% confidence intervals. Exact sign-permutation tests and Holm adjustment across the two principal outcomes accompanied the estimates. I also varied gap caps, coverage, minimum observed duration, anchors and phase boundaries. The robustness figure carries those alternatives through the same outcome definitions, making the effect of analytical choices visible.

Points are equal-participant feedback-minus-baseline differences in minutes per observed hour; bars show 95% participant-bootstrap intervals. Each row changes an identified analytical rule.

Dissertation · October 5 release · Figure 5.3

What the field pilot showed

System operation

Fourteen of 18 operational records corresponded to study completers. The analysis retained 218,708 of 234,614 light observations from completers after time and quality screening.

Participant experience

Eleven of 14 rated feedback very or extremely helpful, and 9 selected greater awareness of their light exposure as the most beneficial aspect of participation.

Measured light exposure

Each estimate compares feedback with baseline in minutes per observed hour. Both confidence intervals include zero, so the pilot did not establish a change in either exposure measure.

Daytime · At or above 250 lux mEDI
+0.56 min per observed hour 95% CI −3.68 to +5.43 · 12 paired participantsBaseline 12.14 · feedback 12.69 min per observed hour
Three hours before bed · Above 10 lux mEDI
−1.61 min per observed hour 95% CI −11.35 to +9.23 · 11 paired participantsBaseline 32.27 · feedback 30.66 min per observed hour
Feedback minus baseline. Dots are equal-participant estimates and bars are 95% participant-bootstrap confidence intervals. Dissertation §§5.7.1–5.7.2.
Field counts, interface use, and survey responses
Field record and interface evidence
MeasureRecorded result
Study completion14 / 18 operational records
Retained light observations218,708 / 234,614 completer source observations (93.2%)
Official participant-days119 baseline · 14 activation · 122 feedback
Observed baseline / feedback time522.5 / 512.7 hours
Morning / evening surveys204 / 255 days · 172 / 255 days
Post-study evaluation14 / 14 completers
Dashboard / Telegram evidence13 / 14 · 14 / 14 completers

Dissertation · October 5 release · Tables 5.1–5.2

Recorded interface use during the official intervals
Interface measureRecorded result
Attributable dashboard views156 views · 13 tracked completers
Dashboard views per tracked participantMedian 10 · range 2–24
Recorded dashboard-day cells with a view62 / 103
Participant-originated Telegram entries98 entries
System / researcher Telegram entries492 entries

Dissertation · October 5 release · Section 5.4 · intervention fidelity and observable engagement

Post-study experience, n=14 completers
ItemResponsesWilson 95% CI
Feedback very / extremely helpful11 / 14 (78.6%)52.4–92.4%
Satisfied / very satisfied with participation12 / 14 (85.7%)60.1–96.0%
Probably / definitely participate again12 / 14 (85.7%)60.1–96.0%
Prompted changes occasionally / frequently11 / 14 (78.6%)52.4–92.4%
Awareness selected as most beneficial9 / 14 (64.3%)38.8–83.7%

Dissertation · October 5 release · Table 5.3

Participant-reported experience after the study. Points show each response proportion; horizontal bars show Wilson 95% confidence intervals. All five items use the same denominator of 14 completers.

Dissertation · October 5 release · Table 5.3 · plotted from the reported counts

Paired participant results and exposure profiles
Each line is one paired participant. Daytime H1 counts time at or above 250 lux mEDI; evening H2 counts time above 10 lux. Both axes use minutes per observed hour.

Dissertation · October 5 release · Figure 5.2

Panel A uses participant-specific 15-minute clock-time medians; Panel B aligns to attempted bedtime. Lines show medians and bands show interquartile ranges. The mEDI axis is symmetric-log; contributors vary by bin and unobserved periods remain missing.

Dissertation · October 5 release · Figure 5.4

Presentation and resources

Wearing instructions from the field pilot

Connect the sensor to its companion and attach it to the glasses with rubber bands. Adjust the position, switch on the data broker, then switch on the companion. The kit is ready when the indicated light is solid blue and the green light blinks every five seconds. Keep the broker nearby and carry the companion in a pocket, bag, or attached to you. Wear the sensor during normal activities and, when practical, until bedtime. Then remove the glasses and switch off both devices.

Dissertation presentation 41 scenes · 26 backups

The dissertation presentation

LuxREST

Scroll to animate the presentation here, or open its full-screen page.

Full screen

41 main scenes · 26 technical and source backups · Use the presentation’s slide selector or scroll within the frame.

Archived dissertation presentation, October 2, 2026. Its TRIAD figures reflect that release. See the current TRIAD manuscript results: 728 publications and 646 study families.

Contact

Do you have a question or want to chat?

Analytics preferences

Optional Google Analytics uses cookies to measure page visits and engagement. You can browse the entire portfolio with analytics off.

Analytics is off in this browser.

Privacy details. Change your choice here at any time. Your preference is remembered for up to 180 days.