RadarFloor

Inspiration

RadarFloor started with a gap raised during our interview with an occupational therapist who conducts home visits.

During these visits, healthcare professionals may identify older adults who need fall monitoring, especially in bathrooms. However, the therapist explained that existing solutions can be too expensive for practical household use. Even with broader government efforts to support ageing in place, therapists may still have no affordable device to offer when they identify this need.

The urgency became clearer after we read about an 85-year-old woman who fell in her shower at a community care apartment in Bukit Batok. She pressed an emergency call button, but an ambulance was only called more than two hours later, according to her daughter.

Commercial fall-detection systems already exist. For example, Canon’s AI-powered fall-detection system combines video analytics, LiDAR and sound detection. These systems demonstrate what the technology can achieve, but our interview highlighted a need for a simpler and more affordable option for individual homes.

What RadarFloor does

RadarFloor is a camera-free and wearable-free bathroom fall-detection concept.

Low-mounted sensor units form a radar sensing grid across the bathroom floor. The system measures how many sensing paths a person interrupts and how long the interruption lasts.

Normal activities should interrupt fewer paths briefly. A person lying across the floor should interrupt a wider area for longer.

Our current prototype hypothesis identifies a possible fall when:

$$ B \geq 3 $$

where (B) represents the number of sensing paths blocked simultaneously.

The system checks the pattern across multiple readings before sending an emergency alert to a caregiver or response team.

RadarFloor does not record photographs, video or conversations. It processes only the sensing pattern required to detect a possible fall.

How we built it

We designed RadarFloor as a six-step process covering bathroom measurement, sensor placement and monitoring.

1. Scan the bathroom

The installer conducts a phone LiDAR walkthrough using a room-scanning tool such as RoomPlan. The scan records the walls, doors and major bathroom fixtures.

2. Calculate sensor coverage

A deterministic coverage engine uses the room measurements to test possible sensor positions. It calculates line of sight, coverage and blind spots for each layout.

3. Compare possible layouts

An AI layer proposes and compares candidate configurations. It calls the coverage engine to evaluate each option.

The AI helps explore possible arrangements, while the deterministic engine performs the geometry calculations.

4. Produce an installation plan

The system recommends sensor positions and generates a coverage report. The installer can use this plan to understand where the sensors should be mounted and where blind spots may remain.

5. Mount the sensors

Low-mounted units create crossing sensing paths near floor level. The exact number, spacing and height depend on the bathroom layout and require physical calibration.

6. Monitor for falls

The embedded controller continuously reads the sensing grid. It checks the number and duration of blocked paths. When the pattern meets the tested fall criteria, the system sends an alert.

What we learned

Our interview taught us that identifying fall risk does not automatically lead to an intervention. Healthcare professionals need an option that is affordable, simple to install and appropriate for private spaces.

We also learned that sensor placement cannot follow one fixed arrangement. Every bathroom has a different shape, and fixtures may block sensing paths. This led us to include the room-scanning and coverage-planning process.

The Bukit Batok incident highlighted another issue: detecting an emergency is only the first step. A complete system needs alert acknowledgement and escalation. If the first caregiver does not respond, the alert should pass to another contact or response service.

Challenges we faced

Distinguishing falls from normal activity

Crouching, cleaning or assisting another person may interrupt several paths. We plan to combine the number of blocked paths with the duration of the obstruction and confirmation across multiple readings.

Different bathroom layouts

Walls, doors and fixtures may create blind spots. The coverage engine helps predict these gaps, but physical tests must confirm the results.

Wet and reflective conditions

Water, steam, flooring materials and reflective surfaces may affect sensor performance. The hardware will also require suitable protection for bathroom use.

Sensor placement

Sensors mounted too high may react frequently to normal leg movement. Sensors mounted too low may fail to detect parts of a fallen body. The mounting height and three-path threshold remain prototype assumptions.

Keeping the system affordable

RadarFloor aims to reduce hardware and processing requirements through simple geometric logic. We still need to calculate the full cost of the sensors, controller, waterproofing, communications and installation before making a confirmed affordability claim.

What we are proud of

RadarFloor protects privacy by design. It does not depend on a bathroom camera or a wearable that the user may forget to charge.

The project also connects fall detection with installation planning. Instead of placing sensors through guesswork, the workflow scans the bathroom, calculates coverage and produces a recommended layout.

Most importantly, we based the project on a need raised by a healthcare professional who works directly with older adults in their homes.

What’s next

Our next steps include:

• Building and testing the physical sensing matrix
• Recording patterns for walking, crouching and different lying positions
• Measuring false alerts and missed detections
• Comparing predicted coverage with physical results
• Calibrating the sensor height and detection threshold
• Developing alert acknowledgement and escalation
• Calculating the total hardware and installation cost
• Conducting further interviews with therapists, caregivers and older adults

Our goal is to give healthcare professionals a practical fall-monitoring option that they can recommend when they identify an at-risk senior during a home visit.

Built With

• python