STMicro Unveils 6-axis IMU with Embedded Sensor Fusion and AI
Aimed at enabling low-power sensing applications such as AR/VR and wearables, ST's IMU serves up sensor fusion blocks and machine learning (ML) cores.
Among the most fundamental trends in the consumer electronics industry today is the proliferation of sensors. For applications such the Internet of Things (IoT), wearables, and AR/VR, sensing has been a core technology.
As devices become embedded with more and more sensors, system developers face about how to best handle and fuse that data into a useful format. To address this issue, STMicroelectronics (ST) has recently released a new 6-axis IMU which includes a number of integrated features such as sensor fusion blocks and Machine Learning (ML) cores.
The LSM6DSV16X is a 6-axis IMU that embeds sensor fusion capabilities. Image used courtesy of STMicroelectronics
In this article, we’ll discuss sensor fusion, why it’s a challenge, and how ST’s new product hopes to enable low-power sensing applications.
Today, many devices operate based on simultaneous input from a number of disparate sensors. For example, consider a health-tracking smartwatch which may include input from a pedometer, inertial-measurement unit (IMU), and optical heart sensors.
While having all of these different data sources allows for a highly detailed and comprehensive understanding of the environment, the challenge arises in how to actually handle and aggregate all of this data. To do this, these devices rely on sensor fusion.
A block diagram representing sensor fusion. Image used courtesy of Tzafestas and coauthors
Sensor fusion is the process of merging data from multiple sensors in order to produce a single stream of data which reduces uncertainty and improves reliability. While simple in concept, sensor fusion is particularly difficult to perform in practice.
Implementing sensor fusion is often a very compute-intensive application that relies on a number of complex algorithms. Some of the most notable algorithms used in sensor fusion include algorithms based on the central limit theorem (CLT), bayesian networks, the Kalman Filter, and convolutional neural networks (CNNs).
Owing to the complex and compute-intensive nature of the algorithms in sensor fusion, a major challenge is in how to perform sensor fusion in ultra-low-power applications.
ST’s LSM6DSV16X, a 6-axis IMU with Sensor Fusion
Recently, STMicroelectronics released a new product that they hope can enable more low-power sensing applications. Dubbed the LSM6DSV16X, the device is a 6-axis IMU that uniquely includes a number of integrated functions and hardware blocks.
Notably, one of these blocks is the sensor-fusion low power (SFLP) block, which is an embedded sensor fusion algorithm used to provide a 6-axis (accelerometer and gyroscope) game rotation vector represented as a quaternion. According to ST, the SFLP block enables gesture recognition or continuous tracking while drawing as little as 15 µA.
Block diagram of the filters in the LSM6DSV16X. Image used courtesy of STMicroelectronics. (Click image to enlarge)
Along with the SFLP block, the LSM6DSV16X includes a triple core for processing acceleration and angular rate data on three separate channels, each including its own dedicated configuration, processing, and filtering.
The IMU also comes embedded with a dedicated machine learning core (MLC) that works on data patterns coming from the accelerometer and gyro sensors, with the ability to also connect to external sensors. The goal of the MLC is to provide the system with flexibility by offloading some AI/ML from the application processor to the sensor itself.
Altogether, the LSM6DSV16X offers a power consumption as low as 0.65 mA in high-performance mode while also offering a compact footprint of 2.5- × 3- × 0.83-mm. More information is available in the LSM6DSV16X datasheet.
Enabling Low-power Sensing
As sensing becomes more important in modern electronic devices, it’s critical to find a way to perform sensor fusion at the lowest power expenditure possible. By embedding sensor fusion and other processing functionality into the LSM6DSV16X, ST seems to be paving an exciting path forward for low-power sensing applications.