Sensors Are a Primary Source for Big Data



New IoT applications—from medical to smart energy to animal husbandry— drive the need for more layered intelligence to address security and privacy concerns, and to manage stunning volumes of data.

Consider a conference that is simultaneously transmitted to audiences in three cities. When a presenter asks the audience a survey question, they can respond by raising their hands to signify agreement to the proposition. As the vote is taking place, the total vote tallied across all three cities is displayed to the presenter and the audience in real time.

Actually, that futuristic set-up became reality recently at Freescale. We did that, in part, to illustrate the potential of the Internet of Things (IoT). Here is how it works. Each audience member is fitted with a wristband embedded with motion sensors. Sensor data from the wristband captures the movement of the audience member’s wrist. To minimize communications bandwidth consumed by this exercise and reduce the power consumption for wireless communications, contextually aware algorithms running on the wristband interpret sensor data and look for data patterns that suggest vertical displacements congruent to a user raising his hand. When such a signature movement is present, the wristband transmits its data to a wireless access point situated at that conference location.

NCS-11890_MPC8540_Block_v1
Figure 1. An example of IoT architecture Freescale used for the live vote tallying exercise

The wireless access points time-stamp the data they receive from the wristbands and then forward the information expeditiously to a cloud-based application. That application uses the results from the wristbands in all three conference locations to deduce when the presenter is taking a vote. Whereas the algorithms running in the wristband can recognize vertical motion, it is difficult for them to discern whether the sensor is moving vertically because a user is raising his hand, or simply because a listener is fidgeting or standing up. The intelligence in the cloud, however, can notice that during a narrow window of time the sensors carried by a larger portion of the audience are moving up simultaneously and deduce that it is a vote that it should tally.

Sensors at the Source

This example illustrates many of the architecture challenges for the IoT (see Figure 1). At the source of an IoT-connected device is often one or more sensors. Sensors convert signals from physical environments such as motion, magnetic field or ambient pressure into digital data. Because sensors provide data continuously and autonomously, sensor data can quickly exceed human-generated data in volume. To alleviate data congestion and associated transmission costs, smart sensors can make a real-time determination on the salience or relevance of the data and transmit them only when they are deemed potentially useful by upstream applications. For example, an algorithm on a motion sensor can determine that the sensor has been stationary and skip an update. A more sophisticated contextual algorithm may be able to differentiate between the wearer raising his hand and other actions such as standing up. Placing intelligence at the data source can reduce the communications bandwidth consumed by sensor data and prolong the battery lives of battery-powered wireless sensor nodes. However, computation capacity at the sensor node is more costly than cloud computing, and intelligent sensors designed for a specific application may be less effectively adapted to a different purpose.

Freescale Internet of Things Tree
Figure 2. Sensors are the root as a primary source of data for the Internet of Things, bringing layered intelligence to enrich human experiences

Intelligence at the data source is also critical where security is a concern. Different security and privacy protocols are being discussed that aim to require negotiation between cloud-based applications and data sources for permission to use their data, in part or as a whole. This is particularly sensitive in body-worn sensors that can record signals which may seem meaningless to the individual. When these signals are combined with other information in a data-mining algorithm, they can unintentionally breach consumer privacy.

At the gateway level, the demands generated by sensor data can also be acute. Sensor data is real-time data, so gateways may be required to help synchronize different sets of sensor data and control data latency. Sometimes further contextual processing may occur at the gateway to further reduce uplink bandwidth requirement. Within the cloud, the same set of sensor data would be made available to a large number of task workers from multiple servers and applications, so the same wristband that is tracking the votes of a conference could also be monitoring the wearer’s activity level, and helping to predict the daily traffic pattern. Invariably, one finds layered intelligence at the core of sensor-based IoT deployments.

IoT-connected devices, like the wristband given in our example, will dwarf all connectivity by 2020, including human-to-human, human-to-machine and machine-to-machine connections. This trend is fueled by four factors:

  • The decreasing cost of sensors and actuators, especially based on micro-electromechanical systems (MEMS) technology, make the vast deployment more feasible.
  • The decreasing cost of Wi-Fi routers makes massive connectivity more feasible.
  • The Internet Protocol Version 6 (IPv6) extends the number of unique Internet addresses to connect trillions of physical objects.
  • Ubiquitous smartphones and tablets demonstrate the process and results of unprecedented connectivity.

Industrial Applications Drive the Future

Although the example we used—and indeed, much of today’s attention on IoT has been associated with wearable devices—much greater business potential for IoT lies in industrial applications. When McKinsey & Company1, Cisco2 and GE3 all pointed to IoT making a multi-trillion dollar impact on our economy by 2025, they were looking at gains in healthcare and infrastructure deployments.

Wearable technologies today are not limited to satisfying consumers’ quantified lifestyles and providing fodder for social media. They are also being used to improve asset tracking for animal husbandry. For example, valuable race horses are wearing sensor patches and sensor-equipped horseshoes to help their trainers monitor their health, record their gait and upload the data so algorithms can monitor the behavior of the horse, diagnose illness and help promote the overall wellness of the animal. Simple motion sensors like the ones in an activity-tracking wristband are being used to detect and report tampering with smart meter installations to protect system integrity. Motion sensors along with pressure sensors are being used to monitor bedridden patient comfort, measuring respiratory and heart rates, and even alerting the nurse’s station to summon assistance when a patient is trying to get out of bed.

Today, we serve over 150 unique sensor applications per year. We see sensors integrating more intelligent functions and we see the need to more closely integrate our sensors with MCU and digital networking offerings as systems solutions. Our observations reflect the need for more layered intelligence across our product families to address power conservation, security and connectivity concerns. With the coming waves of IoT applications, we believe sensor systems to become more complex, more context and environmentally aware, and, fortunately for all of us who are working with them, more interesting.

Bibliography:

  • McKinsey & Company, Disruptive technologies: Advances that will transform life, business, and the global economy, May 2013.
  • John Chambers, The possibilities of Internet of Everything, February 2013.
  • GE 2011 Annual Report.

Ian-Chen_blue_HRIan Chen manages marketing, systems architecture, software and algorithm development for Freescale’s Sensor Solution Division. He held senior business, marketing and engineering leadership positions at Sensor Platforms, Mobius Microsystems, Analogix Semiconductor, Cypress Semiconductor, IC Works, National Semiconductor and Texas Instruments. Ian received bachelor and master’s degrees in electrical engineering and an MBA, all from the University of Illinois at Urbana-Champaign. He holds more than ten patents.

Share and Enjoy:
  • Digg
  • Sphinn
  • del.icio.us
  • Facebook
  • Mixx
  • Google

Tags: