The Next Wave in Activity Tracking

Tracking human activities and monitoring object statuses in smart buildings is essential for a variety of reasons, primarily related to improving efficiency, safety, and comfort. By collecting data on how occupants interact with the building and its resources, facility managers can optimize energy usage, space utilization, and overall building performance. This real-time tracking provides a comprehensive understanding of how spaces are used, allowing for intelligent adjustments to heating, cooling, lighting, and other environmental controls to maximize energy savings.

While most conventional solutions involve installing sensors on each user and object of interest to capture relevant data, this approach has its own set of challenges. Maintaining and managing a large number of sensors can quickly become a logistical nightmare, especially when dealing with a large number of objects. Ensuring that each sensor is operational and its batteries are charged is a demanding and time-consuming task. As a result, the practicality of this approach diminishes significantly as the size of the building and the number of objects to be tracked increases.

Alternative methods that leverage WiFi, sound, light, or other signals to infer activities have emerged to address these scalability issues. However, these approaches often require computationally expensive algorithms to operate effectively. The processing power needed for these algorithms can be a limiting factor, especially in scenarios where real-time data processing and low energy consumption are essential.

While it may not be pretty, researchers at the University of California, Los Angeles have come up with an interesting solution that sits somewhere in-between the two present predominant methods. Their system, called CubeSense++, requires the installation of a device on each object to be tracked, however, these devices are inexpensive and battery-free, meaning that they require little to no maintenance. Mechanical energy harvested from normal interactions with the objects to be tracked causes the trackers to spin in unique ways, which can be sensed by their interactions with millimeter wave radar signals.

The electronic-free CubeSense++ sensors are 3D printed, and are designed to translate various motion types (e.g. sliding, rotational) into a distinct spinning motion of an attached reflector that reflects millimeter wave radar signals. To give each sensor a unique radar signature, a customized set of gears is produced to control the speed of the spinning reflector. The design of these reflectors was facilitated by the use of a genetic algorithm.

These reflectors are monitored by a single Texas Instruments AWR1843 radar unit operating at 77 to 81 GHz. Three transmitters emit radar waves, which then bounce off of the reflectors, and are modulated in the process, before they are measured by a set of four receivers. A time sequence of received radar signals is analyzed by a custom, lightweight algorithm that was designed to detect the presence of an event, as well as the direction, speed, angle, and uses of any target objects in the area.

A set of fourteen sensors were placed in both indoor and outdoor environments to test the utility of the system. After conducting 840 trials, the team found that their system exhibited a true positive rate of 98.25% for use detection.

The researchers identified some areas where improvement is needed in the CubeSense++ system. At present, for example, the system can only recognize activities within the radar’s line of sight. There are also some difficulties associated with the initial calibration of a new setup, and the design of the gear systems for new reflector units. If these limitations can be overcome, a version of CubeSense++ may one day power all manner of future IoT devices.