Occupancy Aware Smart Lighting

Multi-Service Smart Lighting Occupancy Sensors

Pervasive building occupancy sensors for energy efficient lighting control are already a given as mandated in ASHRAE 90.1 and IECC codes.  However, the opportunity exists for innovative smart building system OEMs to vastly improve  performance, occupant satisfaction, commissioning and on-going calibration burdens, and over-the-top building services that can be provided by use of smart lighting sensors powered by SensiML algorithms.

Today's passive infrared (PIR) occupancy sensor leaves much room for improvement with the application of smart sensing.  Current PIR devices are difficult to calibrate well and often lead to frustration as conference rooms routinely go dark in the middle of meetings and fail to provide the level of insight needed to delight users and building owners while maximizing the true potential for energy savings.

Using digitally processed PIR pattern recognition algorithms, SensiML has demonstrated the next generation of building occupancy sensors that:

  • Are more reliable in determining actual space occupancy
  • Can delineate types of movement for better contextual information
  • Enable real-time room occupancy counts (ex. meeting rooms, auditoriums, restrooms)
  • Can provide traffic profiles for over-the-top applications like footfall traffic heatmaps in retail settings
  • Can adapt and learn building usage profiles over time and adapt automatically without continuous intervention


Building sensor and system OEMs seeking to provide smart services can rapidly integrate such capability into their existing PIR sensor endpoints to deliver truly intelligent lighting and occupant sensing.

Smart City Infrastructure

Distributed Smart City IoT Sensors

Smart city initiatives are sprouting up across the globe with great promise of harnessing the power of cloud computing and analytics to solve some of the biggest urban growth challenges.  A variety of applications and approaches have been envisioned and are beginning to be piloted and tested.  Regardless of approach and scope, a common issue with all such systems is how to deal with the sheer volume of data that will be generated and analyzed.  When vision meets such reality, two outcomes arise:

  • Reduction and simplification of scope to ensure that the level of data generated can be coped with by cloud analytics and networks that feed them
  • Distributed analytics where localized raw sensor processing can be handled at the edge (or as with SensiML even at the sensor endpoint itself), thus providing rich sensing with the network impact of traditional simple sensors

With tools allowing sensor device developers to quickly integrate smart sensor algorithms directly into their devices, SensiML can support distributed processing smart city use cases that are difficult or impossible to deliver through purely cloud-based analytics.  SensiML's sophisticated machine learning algorithms are optimized to run locally on devices ranging from low-cost/low-power microcontrollers found in sensors themselves up to more powerful processors found in edge hubs, servers, and gateways. SensiML can provide smart city sensor OEMs and system integrators the ability to offer meaningful insights such as:

  • Audio profile-based alerts - Sensors programmed to recognize and report gunshots, spoken keywords, breaking glass, vehicle accidents, and other sonic events
  • Visual cues - Using PIR or optical sensors to provide motion triggers and activity type/volume data
  • Seismic vibration events - Building, roadway, and infrastructure sensors useful for vehicular traffic monitoring and classification events (light trucks/cars versus heavy trucks), emergency / earthquake response
Indoor Location and Tracking

Contextual Sensing Based Location and Tracking

A variety of applications demand a need for accurate position and movement tracking indoors.  The challenge of course is that satellite based GPS is general not viable indoors due to signal attenuation through walls, ceilings, and floors.  Various proprietary systems do exist to address the need for indoor location but such system can often involve expensive RF/IR and other infrastructure overlays.  SensiML can enable smart sensors and tags to have location awareness utilizing contextual sensing of existing building and RF networks already in place.  Using pattern recognition technology that can be implemented in low-cost microcontroller SoCs, SensiML can provide algorithms that have been demonstrated to 1-3m accuracy levels using only RF signal strength detection of existing network access points and routers along with magnetometer data found in common smartphone sensor ICs.  The combination can be used to rapidly train a model for a given environment with a resulting algorithm that can be reduced to executable code that fits in a very low-power sensor.