⚡ electrical engineering

GEM: Gear-based Environment-Integrated Mobility for Adaptive Indoor Human Sensing

This paper presents GEM, a hybrid sensing system that integrates a matrix of gears into everyday surfaces to autonomously transport infrastructure-based sensors, thereby combining the unobtrusive nature of static monitoring with the adaptability of mobile devices while avoiding the burdens of wearables and observer effects.

Original authors: Shubham Rohal (University of California Merced), Dong Yoon Lee (University of California Merced), Phuc Nguyen (University of Massachusetts Amherst), Shijia Pan (University of California Merced)

Published 2026-04-14

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Shubham Rohal (University of California Merced), Dong Yoon Lee (University of California Merced), Phuc Nguyen (University of Massachusetts Amherst), Shijia Pan (University of California Merced)

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to watch a play in a theater, but you have a problem: the actors (the people you want to sense) keep moving around the stage, but your cameras (the sensors) are stuck in fixed seats.

If you put cameras everywhere to catch every move, it gets expensive, messy, and you end up with too much footage. If you give the cameras to a robot that chases the actors, the actors might feel watched and change their behavior, or the robot might run out of battery.

GEM (Gear-based Environment-Integrated Mobility) is a clever new idea to solve this. Think of it as turning the floor and walls of a room into a giant, invisible public transportation system for sensors.

Here is how it works, broken down with some everyday analogies:

1. The "Subway System" for Sensors

Instead of building a new road for every single car, cities build a subway system. GEM does the same for sensors.

The Tracks: Imagine the floor is covered in a grid of spinning gears, like a giant puzzle.
The Trains: The sensors (like Wi-Fi detectors or thermal cameras) are the "passengers." They sit on little rails attached to these gears.
The Movement: When the gears spin, they don't just spin in place; they act like a conveyor belt. Because the gears are meshed together (like teeth in a clock), when one turns, they all turn in sync. This allows a sensor to slide smoothly from one gear to its neighbor, moving across the room without needing a battery-powered motor on the sensor itself.

2. The "Dance Floor" Analogy

Imagine a dance floor where everyone is holding hands in a circle. If one person steps forward, the whole circle shifts slightly.

In GEM, the gears are the dancers.
The sensors are the people standing on the dancers' shoulders.
By rotating the "dancers" (gears) in a specific pattern, the "people" (sensors) can be moved from one side of the room to the other.
The Trick: The researchers figured out a mathematical rule (like a choreography guide) to make sure the dancers don't bump into each other. They proved that even with a crowded dance floor, you can always find a way to move a specific person to a new spot without causing a pile-up.

3. Why is this better than Drones?

No "Creepy" Factor: A drone flying around your living room feels intrusive. It buzzes and makes you feel watched. GEM sensors are hidden inside the floor or wall. They move silently and invisibly.
No Charging Hassles: Drones need to land to charge. Wearable devices need batteries. GEM sensors are part of the building's structure. They can be powered by the floor itself, so they never need to stop for a recharge.
Smart Relocation: If you are sitting in the corner of a room, a fixed sensor might not see you well. But a GEM sensor can "drive" itself over to the corner to get a better view, then drive back when you move.

4. The "Traffic Jam" Problem

The researchers had to solve a tricky math problem: What happens if two sensors want to go to the same spot at the same time?

They created a smart traffic controller (an algorithm).
This controller acts like a GPS for the gears. It calculates the perfect rotation sequence to move Sensor A to the left and Sensor B to the right without them crashing.
They proved mathematically that as long as you don't pack the gears too full, you can always move the sensors exactly where you need them to go.

5. The Real-World Test

The team built a small prototype (a 3x3 grid of gears) to prove it works.

They showed that a sensor could physically slide from one gear to another in less than a second.
They ran computer simulations with huge grids (up to 64x64 gears) to show that the system scales up well, meaning it could work in a whole house or office building, not just a tiny lab table.

The Bottom Line

GEM is like giving a building a nervous system. Instead of having static sensors that stare at the same spot, or mobile robots that get in the way, GEM turns the building itself into a smart, moving network. It allows sensors to "walk" around the room to get the best view, all while staying hidden, quiet, and powered by the floor beneath our feet.

1. Problem Statement

Current indoor human sensing systems face a trade-off between infrastructure-based and mobile-based approaches:

Infrastructure-based systems (e.g., Wi-Fi, thermal, vibration sensors) offer continuous, unobtrusive monitoring but suffer from static deployment. This leads to sensing inefficiencies, high maintenance costs, and massive data volumes when deployed densely to cover large areas.
Mobile-based systems (e.g., drones, robots, wearables) adapt well to human mobility but introduce significant drawbacks. They can induce "observer effects" (altering natural human behavior due to their physical presence), require users to wear devices, and impose operational burdens like frequent charging and noise.

The Core Question: How can infrastructure sensors be moved discretely and adaptively to monitor people without the drawbacks of autonomous mobile agents or the inefficiencies of static dense deployment?

2. Methodology: The GEM Framework

The authors propose GEM (Gear-based Environment-Integrated Mobility), a hybrid system that integrates a matrix of gears into everyday surfaces (floors, walls) to create a "public transportation" system for sensors.

A. Hardware Design: The Gear Matrix

Mechanism: GEM utilizes a matrix of interlocking gears of the same size. Unlike traditional rail systems that restrict motion to linear paths, GEM uses rotational motion.
Sensor Mounting: Each gear features $C$ channels (rails) evenly distributed across its surface. Sensors are mounted on these rails.
Transfer Mechanism:
- Neighboring gears are aligned such that when rotated to specific angles, their rails connect, allowing a sensor to transfer from one gear to another.
- The system uses a single actuator to drive the entire matrix. Due to the meshing nature of gears, adjacent gears rotate in opposite directions (synchronized rotation), enabling coordinated 2D movement across the surface.
Prototype: The authors fabricated a 3 × 3 gear matrix with 4 channels per gear. The system uses servo motors and rack-and-pinion mechanisms to convert circular gear rotation into linear lateral motion for sensor transfer.

B. Software: Constrained Path Planning

Moving multiple sensors simultaneously introduces the risk of collisions (two sensors attempting to occupy the same rail channel). GEM addresses this through a formalized path-planning algorithm.

Gear Channel Parity Constraint: The system enforces a rule where, for any two neighboring gears, there can be at most one sensor on the connecting rails to prevent collisions during rotation.
Formal Proof of Capacity: The authors provide a mathematical proof demonstrating that for an $n \times n$ gear matrix, the maximum number of sensors that can be moved to any target location collision-free is $\lceil \frac{n \times n}{2} \rceil \times C$ . This is based on a "checkerboard pattern" logic where sensors occupy alternating positions.
Algorithm:
- The system modifies the A* pathfinding algorithm.
- Cost Function: $f(n) = p(n) + h(n)$ , where $p(n)$ accounts for rotation costs and sensor transfer costs, and $h(n)$ uses Manhattan distance as a heuristic.
- Operations: The algorithm utilizes Transfer (moving a sensor to an adjacent gear) and In-Place Swap (swapping two nearby sensors to resolve blocking) operations.
- Dynamic Masking: A dynamic masking technique tracks connected channel pairs in real-time to filter out invalid paths during the search.

3. Key Contributions

Hybrid Design Concept: Introduction of GEM, a novel architecture that adds mobility to infrastructure-based sensing without requiring autonomous agents (drones/robots) or user-worn devices.
Hardware Implementation: Design and fabrication of a gear-based sensor movement platform that integrates rail and gear mechanisms for flexible 2D motion on ambient surfaces.
Theoretical Foundation: Development of a path-planning algorithm grounded in a formal proof of physical movement constraints (Gear Channel Parity), establishing the maximum sensor capacity for a given matrix size.
Validation: Demonstration of system feasibility via a real-world 3×3 prototype and scalability analysis through simulations up to a 64×64 matrix.

4. Results and Evaluation

Prototype Characterization:
- Rotation Time: Rotating a gear by 90° takes approximately 0.38 seconds (±0.025s).
- Transfer Time: Moving a sensor across connected channels takes approximately 0.56 seconds (±0.023s).
- The system successfully executed complex pathfinding tasks, rotating gears and transferring sensors between them as planned.
Scalability Simulation (Matrix Size):
- Simulations with 4 sensors on matrices ranging from 2×2 to 64×64 showed that the average number of steps required to reach a target increases linearly with matrix size.
- This indicates the pathfinding algorithm finds near-optimal solutions close to the theoretical Manhattan distance.
Scalability Simulation (Sensor Density):
- On a 3×3 matrix, as the number of sensors increased (up to the theoretical max of 20), the average steps required increased.
- This is attributed to reduced available channel space, forcing sensors to take longer, more circuitous routes to avoid collisions. Standard deviation in steps also increased with density.

5. Significance and Future Work

Significance:
GEM represents a paradigm shift in indoor sensing. It bridges the gap between static infrastructure and mobile agents by embedding mobility directly into the environment. This allows for:

Unobtrusive Monitoring: Sensors move without the noise or physical presence of robots/drones.
Adaptive Coverage: Sensors can dynamically relocate to areas of interest, improving data fidelity without dense, static deployment.
Energy Efficiency: The system leverages the environment's structure, potentially reducing the need for individual sensor batteries (though current prototypes use wired power).

Future Directions:

Path Optimization: Moving from greedy algorithms to optimized multi-agent path planning to reduce bottlenecks.
Power Integration: Developing embedded power circuits within the gear rails to eliminate tangled wires and enable wireless power transfer.
Sensing Quality Optimization: Integrating a layer that dynamically predicts optimal sensing locations based on real-time data quality rather than pre-defined targets.

In conclusion, GEM demonstrates that infrastructure-based sensing can be made mobile and adaptive through mechanical integration, offering a robust solution for next-generation indoor human monitoring.