AeroBot

Objective: 

To develop a predictive, robotic air quality monitoring and adaptive ventilation system + compare to conventional static sensor/systems to reduce spatial CO₂ concentration gradients, improve overall air quality metrics (PM2.5, TVOCs), and reduce HVAC energy consumption in an active classroom environment. 

Project Scope:

This project will involve the design, construction, and validation of AeroBot, an integrated system that leverages autonomous robotics, a wireless sensor network (WSN), and machine learning to dynamically manage classroom air quality. Moving beyond single-point sensors, AeroBot will generate real-time spatial heatmaps of pollutants, uses predictive algorithms to pre-emptively mitigate poor air quality, and optimizes ventilation strategies for occupant health and energy efficiency. This research aims to provide a scalable, data-driven model for improving indoor environmental quality in educational spaces.

Methodology:

A. The Sensing Layer (multi-modal data acquisition)

• Core Sensors: CO₂, PM2.5, TVOC, Temperature, Humidity.

• Precision Localization: Using ArUco markers placed on walls for drift-free, centimeter-accurate positioning, (wheel encoders could be unreliable). 

• Ultrasonic/LIDAR sensors for obstacle avoidance. 

• Autonomous docking? (for self-charging).

• Static Sensor Nodes: Low-cost ESP32-based sensors.

B. The Intelligence Layer (data fusion & prediction):

• Edge Processing: Raspberry Pi or Arduino handles real-time navigation and sensor data logging. Data is synced to a central database (via Wifi/Bluetooth?). 

• Spatial Heatmapping: Software will interpolate sensor data with location data to generate real-time and historical heatmaps of each pollutant.

• Predictive Algorithm (extension):

  • • Inputs: Historical pollutant levels, period of the day, class schedule, number of occupants, external weather data (viaAPI).

    • Output: A prediction of pollutant levels 10-15 minutes in the future.

    • The system can pre-emptively increase ventilation or activate air purifiers before air quality degrades. 

C. The Actuation Layer (Adaptive Response):

• Ventilation + Purification Control: use a smart thermostat and automated damper controllers to modulate fresh air intake? or buy a ventilator or something + link it up. 

• Human-in-the-Loop Alerts: A dashboard for teachers and facility managers displays air quality, system status, and simple recommendations (e.g., "Please open windows for 5 minutes").

5. Novelty & Expansion Ideas:

Differential Ventilation Strategy: Program the robot to identify specific pollutant zones. Instead of ventilating the whole room, the system could guide a portable air purifier to the hotspot or direct a smart fan to mix the air more effectively in that zone.

Occupant-Centric Metrics: Correlate air quality data with student cognitive performance. Administer short, standardized cognitive tests (e.g., Stroop test, digit span test) under different air quality conditions. This directly links your project to learning outcomes.

Pathogen Transmission Risk Index: Integrate a microphone to measure ambient noise levels as a proxy for speech activity. Combine this with CO₂ (ventilation) and humidity (which affects aerosol droplet lifetime) to create a simple, estimated "Relative Transmission Risk" index for respiratory viruses.

Energy Audit Module: Use a smart plug to measure the energy consumption of the HVAC system under AURA's control versus the standard schedule. Quantifying energy savings is a powerful argument for adoption by school administrations.

6. Validation Plan:

A/B Testing: Conduct experiments on Gold/Blue days, same class, same rooms. 

• Condition A (Control): Standard ventilation, static sensor.

• Condition B (Experimental): AeroBot, baby. 

Performance Metrics:

• Air Quality: Reduction in peak/average CO₂ (ppm), PM2.5 (μg/m³).

• Spatial Efficiency: Reduction/elimination of high-concentration zones (>1000 ppm CO₂).

• Temporal Efficiency: Reduction in the duration of "poor air quality" events.

• Energy Efficiency: Reduction in kWh used for ventilation during class hours.

• Cognitive Correlation: Positive trend between improved air quality and cognitive test scores.
  
  

7. Expected Outcomes & Significance:

My project will produce a functional prototype demonstrating a novel, scalable approach to indoor air quality management. The findings will provide tangible evidence on: 

The limitations of static sensor systems.

The efficacy of predictive, robotic monitoring.

The potential for simultaneous improvement of student health and energy conservation.

Materials:

• Mobile Base:

  • Option A: A pre-built robot chassis with motors and wheels. (e.g., a 2WD or 4WD kit from Amazon/SparkFun/Adafruit). ~$30-60.

  • Option B (a bit more custom): TT gear motors with wheels (x2), a caster wheel, and a motor driver (e.g., L298N or TB6612FNG). ~$25-40.

• Main Controller (The "Brain"):

  • Final version will probably use Raspberry Pi 4 (2GB or 4GB) or Raspberry Pi 3 B+. This is non-negotiable for the level of processing required (computer vision for ArUco, data fusion, WiFi communication). ~$35-75.

• Precision Localization (ArUco Markers):

  • A color printer and paper to print the markers. 

  • Camera: Raspberry Pi Camera Module (v2 or HQ).~$25-50.

• Obstacle Avoidance:

  • Ultrasonic Sensors: HC-SR04 (basic object detection) ~$2-5 each (get 2-3).

• Power System:

  • Battery: large capacity USB power bank (e.g., 10000-20000mAh) to power the Pi and sensors. ~$20-40.

    • OR: LiPo battery pack (e.g., 3S, 3000-5000mAh) with a UBEC (voltage regulator) for the motors and a separate 5V regulator for the Pi. More complex but better for autonomous docking if time. ~$30-50.

• Static Sensor Network

  • MHZ19B or other MHZ sensor for CO2. ~$20-30.

  • PM2.5: PMS5003 (Plantower) sensor. ~$20-30.

  • TVOC / eCO2: SGP30 from Sensirion (VOC) ~$15-20.

  • Temperature / Humidity: SHT31-D or BME280 ~$10-15.

• MISC:

  • Voltage Regulators: 5V and 3.3V regulators for battery. 

  • Breadboards, jumper wires, soldering stuff. 

  • Level Shifters to interface 5V sensors (MH-Z19B) with the 3.3V Raspberry Pi or ESP board (initial testing).

  • My amazing 3D printer, Robert. 

FINAL METHODOLOGY:

• Phase 1: Development of the Mobile Sensing Platform (AeroBot)

The core mobile robot will be constructed first (chassis, motors, motor driver). All environmental sensors (MH-Z19B, PMS5003, SGP30, BME280) will be connected to the Raspberry Pi via its UART and I2C interfaces (OR a separate Arduino board), with level shifters employed where necessary. The Pi camera will be mounted for forward-facing vision. A basic obstacle-avoidance routine using ultrasonic sensors will ensure autonomous navigation. Finally, a primary Python script will be developed to perform three key functions concurrently: (1) reading data from all sensors, (2) using OpenCV to process the camera feed, detect pre-placed ArUco markers on the walls, and calculate the robot's real-time XY-coordinate position within the classroom, and (3) publishing this fused sensor-and-location data to a central MQTT broker. 

• Phase 2: Static Sensor Network + AeroHub

3-4 static sensor nodes will be built using ESP32 microcontrollers and programmed via Arduino IDE to read their local SGP30, BME280, and PMS5003 sensors at fixed intervals and publish the data to the same MQTT broker. The AeroHub (a separate Raspberry Pi) will be configured with the Mosquitto MQTT broker, a time-series database (InfluxDB), and a dashboard visualization tool (Grafana). A custom Python script on the AeroHub will act as a subscriber and receive all data from all MQTT agents (from both the mobile and static nodes) and write it into InfluxDB. The Grafana dashboard will provide a real-time + historical view of all air quality metrics, accessible to teachers via a web browser. 

• Phase 3: Implementation of Intelligence and Adaptive Actuation

With data flowing into the database, an actuation logic script will be implemented on the AeroHub. This script will continuously query the latest air quality metrics. Based on predefined thresholds (e.g., CO₂ > 1000 ppm), it will publish a MQTT command to turn on a smart plug, which powers a portable air purifier or fan, thus creating a closed feedback loop. The spatial data from AeroBot will be processed to generate heatmaps within Grafana, identifying pollutant gradients within the room. As a possible project extension, machine learning models (e.g., Scikit-Learn) will be developed on collected historical data to predict air quality degradation and trigger pre-emptive actuation. 

• Phase 4: Validation and Analysis

A/B testing protocol will be employed for validation within the same 1-2 classrooms. The system will be evaluated under 2 conditions over multiple days: (A) Control: ventilation managed only by the existing classroom HVAC with static sensors logging data, and (B) Experimental: with AeroBot (and static sensor nodes) actively mapping the room and controlling the adaptive actuator. Performance will be measured across four key metrics: (1) Air Quality: reduction in average and peak CO₂ and PM2.5 levels, (2) Spatial Efficiency: elimination of high-concentration zones, (3) Energy Efficiency: reduction in kWh consumed by the actuator (measured by the smart plug) compared to the baseline HVAC energy use (check to see if this information can be obtained), and (4) Cognitive Correlation: (optional) by administering standardized cognitive tests to occupants under different air quality conditions to preliminarily assess impact on cognitive performance.  

Coming soon!

Coming soon!

Coming soon!

Coming soon!

Coming soon!

Coming soon!

Coming soon!

Coming soon!

Coming soon!