A platform for environmental health research
Luminate is designed to passively organise mobility, activity, place and modelled air pollution exposure into a continuous record — creating a foundation for research at a scale dedicated sensors rarely reach alone.
THE RESEARCH CASE
Wearable air-quality sensors can provide valuable direct measurements, but cost, charging, carrying and sustained adherence constrain very large or long studies. A well-designed phone app can complement sensor campaigns with continuous, lower-friction context: where people went, how they travelled, where they stayed and the modelled pollution fields they encountered.
Directly observe pollutant concentrations close to the participant and are essential for validation and high-resolution measurement studies.
Uses a device participants already carry to build exposure and mobility context across days, months or potentially years.
THE PIPELINE
Luminate is not a GPS logger with a chart attached. It is a layered system that turns noisy, asynchronous signals into a reviewable timeline and then into exposure and behaviour summaries.
Background location, device motion and temporal context provide the raw stream.
Movement is partitioned into journeys and stationary visits with noise handling.
On-device classifiers and confirmed recurring locations add semantic context.
Timed coordinates and mode are matched to air-quality fields for each closed interval.
Daily records become comparable periods and structured natural-language answers.
CURRENT TECHNICAL CAPABILITIES
The hardest research variable is often the missing day. Luminate's engineering focuses on background continuity, participant-correctable interpretation and bounded resource use.
Recording cadence responds to activity, battery level, Low Power Mode and thermal state. More evidence can be collected when movement needs resolving; less work is done when the participant is settled.
Confirmed recurring places accumulate visit statistics and time-of-departure histograms. Those histories help the recorder balance sleep duration against the probability that a participant is about to leave.
LocoKit combines motion and trajectory evidence to classify movement. Confidence is visible, corrections remain possible, and model maintenance is scheduled for charging periods instead of competing with foreground use.
Background time on battery prioritises recording. Foreground time handles user-visible classification and exposure. Charging, background time is reserved for bounded model and place-statistics maintenance.
Confirmed, ended journeys use timed route coordinates and activity mode. Stops are also scored; long stationary periods use a compact series of centroid samples at the start, hourly boundaries and end.
Place search and recurrence can suggest a name, but only an explicit participant action confirms a stop. This separates inferred identity from participant-validated state.
Ask Luminate converts questions into deterministic queries over local timelines, places and exposure results. Answers can preserve dates, sample coverage and the episodes from which an aggregate was derived.
RESEARCH OPPORTUNITIES
With an ethically governed research layer, Luminate could support observational cohorts, intervention studies and mixed-methods work that links objective mobility patterns with participant-reported experience.
Decompose exposure by hour, microenvironment, recurring place, journey, mode and day type — while retaining the episode behind each aggregate.
Characterise mode share, journey duration, route choice, active travel and time spent at recurring places across seasons or life changes.
Test whether personalised feedback is followed by cleaner route selection, mode substitution, changed departure time or greater physical activity.
Randomise message framing, insight timing or suggested actions, then compare within-person and between-group changes using pre-specified outcomes.
Follow trajectories over months or years rather than relying only on short campaigns, and detect adoption, decay, rebound and seasonal effects.
Deliver event-triggered or scheduled surveys and combine exposure context with reasons, perceptions, constraints and self-reported health or wellbeing.
EXPANSION ROADMAP
These are research directions, not claims about the current production app. Each would require protocol design, participant consent, validation and appropriate platform permissions.
Link heart rate, activity, sleep or other consented wearable measures to exposure episodes and travel context; investigate acute response and accumulated dose.
Attach additional environmental surfaces or device observations to the same timed mobility record for combined-exposure analyses.
Schedule ecological momentary assessments or trigger short questions after journeys, exposure events or insight views.
Support A/B or factorial studies of messages, route suggestions and feedback timing, with transparent assignment and pre-registered outcome definitions.
Pair a subset of participants with personal monitors to quantify bias, calibrate modelled estimates and identify contexts where model performance differs.
Add consented pseudonymous export, cohort management, retention policy and reproducible researcher datasets without making the consumer app a silent surveillance tool.
SCIENTIFIC AND ETHICAL GUARDRAILS
A research deployment should preserve the distinction between observed, inferred, modelled and participant-confirmed data — and make missingness, coverage and uncertainty visible rather than smoothing them away.
The current app is local-first. Upload, linkage, survey delivery and study analytics would require a separate, clearly consented research protocol.
Retain whether values came from sensors, models, classifiers, place suggestions or participant confirmation, with coverage attached to analyses.
Collect only what a study needs; define retention, access, pseudonymisation and withdrawal before recruitment.
Assess exposure and mode performance across transport environments, geographies, devices and participant groups before generalising findings.
Collaborate with us
We are interested in collaborations spanning air pollution, exposure science, mobility, physical activity, digital interventions, epidemiology, human–computer interaction and responsible health data research.
hello@ecoquity.tech
Luminate Research (V1 archive)