Typically associated with cars, aircraft and ships, precision navigation has become widely used within the industrial and medical segments for applications ranging from factory machinery and surgical robots to first-responder tracking.
There are many existing approaches to derive location, direction and movement as they relate to pointing, steering and guiding equipment. In fact, it has become common for many applications to rely on GPS. But when it comes to navigating indoors and addressing more complex and environmentally challenging scenarios, GPS alone is insufficient.
For such applications, you can deploy various sensor types to improve a system’s ability to determine actual from anomalous motion. The ability of a given sensor to address a particular navigation problem isn’t dependent only on the performance level of the sensor, but also on the unique dynamics of the application.
As with any complex design problem, the starting point is to understand the end application objectives and limitations. From there, rank the critical performance parameters to arrive at a rough understanding of the required sensors; then optimize the design through careful sensor conditioning, integration and processing.
The navigation problem
Let’s begin with an analogy: Say you’re at work and want a cup of coffee, so you head for the break room. If you’ve been to the break room before, you likely have a route in mind, but along the way you will rely on various senses—optical, audio, balance and perhaps even touch—to help get you there. Your own “personal processor” combines the inputs from the various “sensors” and applies some embedded pattern recognition. If it’s been a rough day, you may need to obtain external input (get directions). Throughout this process, your personal sensors must be individually precise but must also work well together to filter out and reject misleading information, such as the smell of coffee from your neighbor’s cubicle.
In other words, to reach the break room, you employ the same techniques used by designers of navigation systems for vehicles, surgical instruments and robotic machinery.
The industrial corollary to this example consists of various sensing techniques, none of which singlehandedly addresses the requirements of most applications. GPS is prone to errors due to obstacles that block satellite reception. Another common navigational aid, the magnetometer, requires clear access to the Earth’s magnetic field; there are many field interferences within industrial environments that make a magnetometer’s reliability intermittent at best. Optical sensors are subject to line-of-sight obstructions, while inertial sensors are generally free of these interferences but have some limitations of their own. For example, they lack an absolute reference (where is north?).
Except for the simplest of problems, most solutions rely on multiple sensor types to deliver the required accuracy and performance under all conditions. Inertial sensors, such as microelectromechanical systems (MEMS)-based accelerometers and gyros, can potentially fully compensate for the shortcomings of other sensor types because they are free from many of the same interferences and do not require external infrastructure—no satellite, no magnetic field, no camera, just inertia.
With a 20-year track-record in the automotive industry, MEMS inertial sensors are highly reliable and commercially attractive, as has been demonstrated by their successful application in mobile phones and video games, for which the sensors’ low power consumption, size and cost are favorable factors. There is a large variation in available performance levels, however, and devices suitable for gaming are not capable of addressing high-performance navigation problems. Precision industrial and medical navigation, for example, typically require performance levels an order of magnitude higher than is available from MEMS sensors used in consumer devices.
In most cases, a device’s motion is relatively complex (more than one axis), which drives the need for full inertial measurement units (IMUs), which may integrate up to six degrees of freedom of inertial movement (three linear and three rotational).
For example, Analog Devices Inc.’s ADIS16334 iSensor IMU is amenable to many industrial instruments and vehicles. In many cases, you can integrate four or more additional degrees of freedom, including three axes of magnetic sensing and one axis of pressure (altitude) sensing.
An inertial measurement unit outputs highly stable linear and rotational sensor values that must compensate for the following influences:
• temperature and voltage drift;
• bias, sensitivity, and non-linearity;
• vibration; and
• x,y,z axis misalignment.
Depending on their quality, inertial sensors encompass varying degrees of drift. Designers can occasionally correct for this by employing GPS or a magnetometer.
A central challenge in navigation, beyond good sensor design, is determining which sensors to rely on and when. Inertial MEMS accelerometers and gyros have proved that they are a good complement to help designers craft a fully functioning sensing system.
In an indoor industrial or medical setting wherethe GPS signal is denied and where machinery and electronics introducemagnetic interference, designers must establish less traditionalapproaches to machine guidance. Many emerging applications, such assurgical tool navigation, also require significantly higher levels ofprecision than, say, automobile navigation. In all of these cases,inertial sensors are an option for providing the dead-reckoning guidancerequired to maintain accuracy during line-of-sight blockage or otherinterference sources detrimental to noninertial sensors.
The accompanying figure depicts a generic inertialnavigation system (INS) for navigating anything from a surgeon’s toolto a vehicle or an aircraft. The INS model incorporates a Kalman filter.First used on the Apollo moon missions, these filters are pervasivetoday in phased-locked loops within mobile communications to provide amechanism for merging multiple good but imperfect sensors and therebyobtaining the best estimate of location, direction and overall motiondynamics.
When applied to surgical applications, the INScould be used as a navigational aid for aligning artificial joints, suchas knees or hips, according to a patient’s unique physicalcharacteristics. Besides enabling better alignment (for improvedcomfort) and faster, less invasive surgery, use of the right sensors canhelp counter hand tremor and fatigue.
Purely mechanical alignment has been supplementedby optical alignment in recent years, but just as there are GPS signalblockages that can impede vehicle navigation, there are potentialline-of-sight blockages in the operating room that limit optical sensoraccuracy. An inertial-guided surgical alignment tool can supplement (oreven replace) optical guidance, with no line-of-sight issues, and alsooffer potential advantages in size, cost and automation.
Though there is consistency across applications inthe basics of solving a navigational problem, the end-system specificsmust be well understood. Those considerations will ultimately guide theselection of appropriate sensor types, which in turn affects overallend-system performance.
Thus, in parallel with the strong push for small,low-power and multi-axis inertial sensors for consumer applications,some sensor developers are equally focused on turning out compact,high-accuracy and low-power high-performance sensors.
These environmentally robust sensor developmentsare driving a surge in the adoption of MEMS inertial sensors within theindustrial, instrumentation and medical markets.
|Inertial navigation system merging multiple sensor types with the aid of Kalman filtering.|
About the author
Bob Scannell is a business development manager for Analog Devices Inc.’s InertialMEMS products. He holds a BSEE from the University of California, LosAngeles and an MS in computer engineering from the University ofSouthern California.