Sensor Fusion: Combination of Multiple Sensors for Enhanced Measurement Estimations

 

Sensor Fusion, also known as multi-sensor, is the art of integrating multiple physical sensors to enhance the application and system performance. Sensors are now deployed in a broad spectrum of applications, such as smart mobile devices, automotive systems, healthcare, oil exploration, climate monitoring, and industrial controls. Sensors are used everywhere, and now the advancement in sensor technology is enabling them to mimic human beings. This is possible now because of sensor fusion technology. Sensor fusion leverages a microcontroller to combine the individual data collected from multiple sensors to get a more reliable and precise overview of the data. Sensor fusion enables context awareness, which, in turn, can extensively benefit the Internet of Things (IoT).

Sensor fusion is the ability to bring together input from multiple radars, lidars, and cameras to form a single model or a picture of the vehicle’s surrounding. The resulting model is an accurate representation as it balances the strengths of different sensors. Sensor fusion dynamically improves the lane detection performance as more sensors are used to boost the perception capability. Sensor fusion collects data from each sensor used through software algorithms to offer the most comprehensive and accurate environmental model. The escalating use of sensor fusion in consumer electronics and automotive applications is propelling the sensor fusion market’s growth. According to Emergen Research, the Global Sensor Fusion Market is forecast to attain a market valuation of USD 16.72 Billion by 2027, registering a remarkable CAGR of 19.6%.

Importance and Working of Sensor Fusion Algorithms

Sensor fusion algorithms integrate real-time sensor data that assists in reducing doubts in the object’s location and position. They combine data from several sensors to estimate the correct positions of the objects. Sensor fusions mostly rely on data from numerous same types of sensors, also known as “competitive configuration”. However, when merging data from different types of sensors, such as amalgamation of object proximity data and speedometer data, more often results in an in-depth interpretation of the object examined.

For instance, during a foggy climate, a radar sensor offers more precise data than a LiDAR sensor would. Whereas, in clear weather, LiDAR sensors’ spatial resolution is much more reliable than radar sensor. Every sensor has its own pros and cons, which is why sensor fusion algorithms take into consideration multiple types of sensors. The data from these various sensors are complementary, and the setup is typically called a complementary configuration.

Due to each sensor’s strengths and weaknesses, a robust algorithm gives preference to some critical data points over others. For instance, speed sensors are more precise than parking sensors, so those sensors are mostly preferred. The importance varies, and they generally depend on the specific use. 

Sensor fusion algorithms examine all the input data and generate output with enhanced accuracy and reliability. To measure the kinematic state of a vehicle, two equations and models are applied ‒ namely predict equation that used motion model and update equation that uses the measurement model. The motion model gives data about an object in periodic intervals, and the measurement type is more involved with the vehicle sensors’ dynamics. One of the essential foundations of these algorithms is the Kalman filter.

·       Overview of Kalman Filter

A Kalman filter refers to an equation that obtains data inputs from numerous sources and predicts unknown values, even in the case of augmented signal noise. Majorly applicable in control and navigation tech, Kalman filters confer the prediction of uncertain values more precisely than specific predictions by single measurement method.

As these algorithms are the most commonly used sensor fusion application and provide a resilient foundation for the concept, sensor fusion and Kalman filters are generally considered synonymous. Kalman filters are one of the most popular algorithms in sensor fusion and were invented by Rudolph Kalman in 1960. Now the algorithm is widely deployed in smartphones and satellites from navigation and tracking.

Factors Influencing the Growth of Sensor Fusion Market

The escalating trend of autonomous vehicle and advanced driver assistance systems (ADAS) are further augmenting the need for new radar, GNSS, lidar, and camera sensors in the vehicles. The increasing penetration of smartphones globally and the augmenting trend of micro-sized electronics are anticipated to add to traction to the market’s growth. Moreover, the growing application of fusion technologies to survey and predict environmental conditions, such as temperature, pressure, and humidity, have created an added demand for sensor fusion technologies.

The augmenting need for application-based location detection is attracting consumers and investors to invest in the technology and deploy them across the globe, thereby contributing to the market growth. The increasing advancement in ADAS and the surging use of the GPS-Inertial Measurement Unit fusion is highly beneficial for solving mounting errors. For instance, Tesla’s Autopilot automated driving feature, which is an example of ADAS, has the ability to carry out operations like maintaining the lane of the vehicle on a highway by predicting the accurate location of the vehicle from the data obtained from a forward-facing camera and steering control.

Collaborations and agreements are a common occurrence in the market and are primarily contributing to the market’s expansion. For instance, in April 2020, Foresight Autonomous Holdings Ltd., an Israel-based leader in automotive vision systems, joined forces with FLIR Systems Inc., to develop and commercialize the former’s QuadSight Vision System, which is based on sensor fusion, combined with infrared cameras of FLIR Systems to a broad spectrum of customers. In 2019, Ceva Inc. gained a controlling stake in the smart sensor technology business of Hillcrest Lab. Under the acquisition, CEVA’s intelligent sensing technology’s portfolio has the chance to expand to include CPU vision and AI processing for cameras and sound processing for microphones.

In December 2020, Himax Technologies Inc., joined forces with Edge Impulse to enable the accelerated development and deployment of the machine learning model. The AI vision and sensor fusion solution are designed for predictive maintenance, asset tracking, occupancy detection, and condition monitoring.

STMicroelectronics and Mobileye have formulated a strategy to develop EyeQ 5 system on chip in association with each other, to be deployed in 2020, to perform the function of central computer control and sensor fusion in the fully autonomous driving vehicles. Companies actively focused on developing technologies linked with IoT and smart driving are anticipated to create lucrative future opportunities. The increasing demand for monitoring and controlling utilities in home automation and the escalating development in robotics is forecast to influence the market growth due to sensor fusion technology’s wide application in mobile robot navigation.