Hummingbird 1.0


Inertial Sensing for Things That Think

Chris Verplaetse

MIT Media Lab

Physics and Media Group


Introduction
In the push towards Things That Think, computation, communication, and I/O are migrating away from the desktop computer and into everyday objects. The premise is that things should be able to sense and respond to other things, to their surrounding environment, and to their own state. One important aspect of a thing's own state is its motion - a time series state vector consisting of position and orientation (X, Y, Z, roll, pitch, yaw) with respect to some global coordinate system.

Motion sensing of objects can be performed with a number of technologies including radar, RF, magnetic, electric fields, GPS, and inertial. All of the above sensory technologies, with the exception of inertial, rely on external references and therefore are limited in their modes of sensing. A quick comparison follows:

Sensory Type Reference Example Apps. Limitations
radar external body in motion - air traffic tracking
- 10/10 performance space		 in motion
electric field sensing (fish) external body with electric charge - 3D mouse
- cello bow		 limited range
magnetic transmitter - polhemus
- FOB		 limited range
GPS Earth orbitting sattelite network navigation, more - outdoors
- Earth
Inertial self - ICBM navigation systems
- Hummingbird		 errors increase with time


Inertial Sensors
Inertial sensing is performed with two types of sensors: accelerometers which sense translational acceleration, and gyroscopes which sense rotational rate. Inertial sensing's big win is that it allows for fully autonomous self-motion sensing. Inertial navigation systems are of the dead-reckoning type, however, in which an initial estimate of position is required and current position is determined reletive to the original. Errors inevitibly accumulate in such systems. These systems are best suited for reletive motion sensing applications (not absolute position).

Accelerometers
Accelerometers behave as 2nd order mechanical systems (ie. damped mass-spring system under an applied force) and sense translational acceleration. In its simplest sense position is found by double integrating.

The accelerometers used in Hummingbird are Analog Devices ADXL05 (noted in the table below). Their operation is explaned here

Gyroscopes
Gyroscopes sense rotational rate. Angular position for a given axis is attained by integrating the corresponsing gyro's output. Gyros originally operated via the principles of conservation of rotational energy. Most modern electromechanical gyros use the principles of Coriolis forces to sense rotation. The Hummingbird uses Murata's Gyrostar which is mentioned below.

The Humminbird Board(s)
The Hummingbird (HB) "board" consists of two modules containing a total of three seperate circuit boards. One module, the "Birdboard", houses HB's six inertial sensors and one temperature sensor. The other module, the "Hummingboard", is a general purpose multiplexing a/d converter with both tty and rs-232 serial interfaces. Each Hummingbird module is discussed below.

The Hummingbird's default serial connection is: 9600 baud, 8 data bits, 1 stop bit, no parity. This can be changed.

Hummingboard
The Hummingboard (pictured below) is a general purpose multiplexing a/d converter circuit with both TTY and RS232 I/O lines and a user-specified configuration switch. It has the following specs and functionalities:


Birdboard
The Birdboard module houses the six inertial motion sensors and a temperature sensor (to account for the accelerometers' and gyroscopes' temperature dependencies.) The default sensor settings are optimized for the motions of a handheld video camera (under human control) - that is: translation 0 to 3 g, rotation 0 to 60 deg/sec, and bandwidth of dc to 12 Hz. These settings can be manually adjusted with the swtches and potentiometers on the Birdboard (shown below).



Birdboard "A"



Recovering Orientation / DofBox Code


Motion cognition at the Media Lab
Lastly, let's talk about some of the applications that are driving our needs for motion sensors.









On the web

Inertial proprioceptive devices: Self-motion-sensing toys and tools from the IBM Systems Journal.

Can A Pen Remember What It Has Written Using Inertial Navigation ?: An Evaluation Of Current Accelerometer Technology From Gershenfeld's Physics of Information technology

Thanks

I would like to note the aid and efforts of a number of people:
verp@media.mit.edu | verp | Media Lab | Physics and Media