D ShinFollow


Sensor positions within the indoor environment need to be selected to gather sufficient information to perform a mitigation action in response to a detected state. With current approaches there is no clear prior on where sensors are placed within a given indoor space, and therefore, actuation is based on assumptions, which can lead to a suboptimal control strategy. This disclosure describes techniques to estimate sensor topology within an indoor space using a custom large generative visual model, e.g., a diffusion model. Raw sensor signal streams are transformed into a matrix arrangement of spatiotemporal data to obtain a 2D matrix with entries representing confidence levels on occupancy. The matrix is used as a conditioner for a custom diffusion model. Diffusion steps are performed with an image-to-image generator to obtain as output a sensor topology heatmap that is descriptive of the sensor distribution within the indoor space.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.