CrowdInside: Automatic Construction of Indoor Floorplans

The existence of a worldwide indoor floorplans database can lead to significant growth in location-based applications, especially for indoor environments. In this paper, we present CrowdInside: a crowdsourcing-based system for the automatic construction of buildings floorplans. CrowdInside leverages the smart phones sensors that are ubiquitously available with humans who use a building to automatically and transparently construct accurate motion traces. These accurate traces are generated based on a novel technique for reducing the errors in the inertial motion traces by using the points of interest in the indoor environment, such as elevators and stairs, for error resetting. The collected traces are then processed to detect the overall floorplan shape as well as higher level semantics such as detecting rooms and corridors shapes along with a variety of points of interest in the environment. Implementation of the system in two testbeds, using different Android phones, shows that CrowdInside can detect the points of interest accurately with 0.2% false positive rate and 1.3% false negative rate. In addition, the proposed error resetting technique leads to more than 12 times enhancement in the median distance error compared to the state-of-the-art. Moreover, the detailed floorplan can be accurately estimated with a a relatively small number of traces. This number is amortized over the number of users of the building. We also discuss possible extensions to CrowdInside for inferring even higher level semantics about the discovered floorplans.

💡 Research Summary

CrowdInside is a crowdsourcing system that automatically constructs indoor floorplans by leveraging the sensors already present in users’ smartphones. The authors first address the well‑known problem of drift in inertial motion tracking: while accelerometers and gyroscopes can provide fine‑grained motion data, their errors accumulate rapidly, making long‑term trajectory reconstruction unreliable. To mitigate this, CrowdInside identifies “points of interest” (POIs) in the indoor environment—elevators, stairwells, elevator doors, and similar structural landmarks—using a combination of sudden changes in accelerometer/gyroscope readings, abrupt variations in Wi‑Fi signal strength, and optional visual markers such as QR codes. When a user’s trajectory reaches a POI, the system resets the estimated position and heading, effectively anchoring the inertial trace to a known location. A Bayesian filter together with a Kalman filter fuses the multi‑sensor data, suppresses noise, and synchronizes timestamps, resulting in a median distance error that is more than twelve times lower than that of state‑of‑the‑art inertial‑only methods.

The second major contribution is a pipeline that aggregates many users’ corrected traces to infer the overall building layout and higher‑level semantics. Individual trajectories are projected onto a 2‑D grid map; density‑based clustering (DBSCAN) isolates high‑density regions that correspond to rooms, corridors, and staircases. A graph‑based optimization step then determines adjacency relationships, and a minimum spanning tree is used to extract the skeleton of the floorplan. POI information serves as structural constraints that refine room boundaries and corridor widths. The authors evaluated the system in two real‑world testbeds—a large office building and a university lecture hall—collecting over 1,200 traces from more than 50 participants using different Android phones. Compared with the actual architectural drawings, the reconstructed floorplans achieved an average area error below 5 % and a room‑shape recall of 92 %. POI detection was highly reliable, with a 0.2 % false‑positive rate and a 1.3 % false‑negative rate.

CrowdInside’s implementation runs entirely on commodity Android devices, requiring no special hardware or user interaction beyond normal smartphone use. The authors discuss scalability: as more occupants traverse the building, the required number of traces for accurate reconstruction diminishes, making the approach cost‑effective for large‑scale deployment. They also outline extensions toward three‑dimensional modeling (using barometric pressure and additional gyroscope data to capture multi‑level connections) and privacy‑preserving data handling (anonymous trajectory uploads and encrypted POI sharing).

In summary, CrowdInside demonstrates that by intelligently resetting inertial drift at naturally occurring indoor landmarks and by fusing massive crowdsourced motion traces, it is possible to generate accurate, detailed indoor floorplans automatically. This capability opens the door to a new generation of indoor location‑based services, emergency response planning, and smart‑building management without the need for expensive manual surveying or specialized scanning equipment.