This page chronicles the design and construction of DASL’s humanoid robot, which will be entered in the 2009 Robogames competition.
-
Control/Sensors
The IMU candidates for DASL’s Robogames humanoid are listed below. Detailed specifications of each IMU can be found in the comparison table.
Sparkfun 6DOFv4 control board
Pros
- Low cost ($385)
- Eliminates issues with control board and serial bottleneck with direct analog interface
- Acceleration, rotation rate, and magnetic field direction sensing
- Highest bandwidth sensed (150Hz-300Hz)
- Internal temperature compensation
- Adjustable accelerometer sensitivity
Cons
- Accuracies not rated or calculated. Specified accuracies in
- All calculations and calibration must be designed and implemented
- 6DOF v3 has failed in the past, producing noisy data and drift
Verdict: Fallback choice
Xbow IMU 440
Pros
- Shielded case reduces EMI
- Internal filtering & processing calculates orientation angles
- Digital output in RS232
Cons
- Large enclosure size (3”x3.75”x2.5”) and weight (1.3 lbs)
- Lack of magnetometers for on-the-fly recalibration
- High power requirements (350mA @9+ volts)
- Cost prohibitive
Verdict: Cost, Form factor, and power consumption all eliminate this IMU as a possible choice.
Pros
- Digital output in RS232, standard connector
- Highest rated accuracy and sensitivity of all IMUs found (see table)
- Magnetometers to recalibrate orientation
- Internal temperature compensation
- 16 bit analog to digital conversion
- Programmable bandwidth
- Small form factor (1.65 x 1.6 x 0.6″ and 25g without enclosure)
- Power draw 65mA @5V
- Formats orientation angles as Euler angles, rotation matrices, or quaternions
- Recommended by Dr. Hong and RoMeLa team, proven on DARwIN platform
Cons
- Costs twice as much ($1485) as the accelerometer-only model (3DM)
- Enclosure shape not ideal for tight confines of robot, lack of enclosure exposes circuits
Verdict: Optimal choice
The requirements for a good IMU for this project are:
- Measurement of body orientation to properly calculate the zero moment point location
- Measurement of torso rotation rate for angular momentum control
- Temperature compensation due to close confines with motors and batteries
- Noise rejection due to proximity of motors
- Light weight and small form factor to fit as close to center of mass as possible
Assuming that a $1500 IMU is affordable, the Microstrain 3DM GX-1 is the ideal candidate. It has been proven on the DARwIN robot already, and has all of the desired characteristics enumerated above. The plug-and-play nature of the device means that no additional compensations are necessary. Data is available in convenient representations with almost no additional programming.
Because of its sensor complement, the Sparkfun IMU theoretically could give the same data, albeit slightly less accurately. However, to actually achieve this would require programming and development to convert the raw sensor data into orientation and rotation rate. The burden of programming, troubleshooting and bug-chasing is a large risk given the short timeline of the project. It is therefore useful as a fallback option, only viable if money becomes the primary concern.
