View Full Version : Mech Warfare Tilt/Accerometer

10-04-2011, 03:13 PM
For a mech warfare biped, whats the norm for tilt/accerometers?
Price-wise and Spec-wise.

Edit: So far the only differences I see are the G force sensitivities, and axis but I'm assuming I'll be needing 3 axis.
So...how many G's do I really need? 1.5? 3? 100000000!?

11-22-2011, 08:31 AM
I have the same question as Fanatic, not sure if no one knows the answer or if no one responded? Here is a link to sparkfun page that discusses the basic's of Accelerometers and what differenciates them.


If I make any progress I will post it here.


Did some napkin math, estimated most motion seems to be over very short distances, so I used 0.1m (for falling/large body motions) and 0.05m (for small body motions) and I figured the time frame is quite small, so 0.1 s for small motions and 0.2 for larger motions, again like falling down. I get 2.5 m/s^2 (0.25 g) and 5 m/s^2 (0.5 g). Does this math make sense to you guys? Seems like most acceleromters are way outside this range, but I can't find any good information on the CM line of controlers and their corisponding hardware...

11-22-2011, 10:24 AM
I think the reason nobody has answered the question is: there is no "norm" for this. I'm not sure if any biped other than Giger has used an IMU during Mech Warfare.


11-22-2011, 10:45 AM
In common biped usage accelerometers are used for tilt angle detection, and determine if the robot is fallen and in which direction. If this is your intention then +/- 1G is sufficient, though +/- 1.5 to 2G allows for error. 3 axis are cheap enough and may simplify maths and improve accuracy. This does not really require high accuracy, so pretty much an device will do
Accelerometers in conjunction with Gyro in an IMU can help to compensate for gyro drift, but this again does not need range greater than +/- 1G. Here I would avoid too high a G range, because you want good angular accuracy

The higher G accelerometers are usually used for bump detection on laptop disk drives. They are not much value in bipeds unless you want to grunt as you hit the deck

11-22-2011, 11:19 AM
Firstly, thank you for the quick replies! I understand 'normal' is some what a misnomer in this case, hopefully we can shead some light on the subject so that we can use 'normal' in the future :*)

What I am looking to do (not sure about Fanatic) is add in an IMU to help with walking and balance. I thought that this was fairly common as the Bioloid systems have IMU's built in via the CM controllers, thus I assumed most everyone with a bi-ped would be using one? Is this not the case? I wouldn't complain about using it to determine falls, but what I really want is something sensitive and quick enough to provide some feed back on motion while walking. I plan on using an Arbotix as that is what I have, so something that can be directly connected to it would be a major plus.

@ibot- You mention 'common biped usage,' is this first hand knowledge or can you point us in the direction of some robots that use this type of sensor?

Thanks again for all your help!


11-22-2011, 02:31 PM
Most accelerometers and rate sensors have selectable output ranges. The LIS331DLH used in the CM-730 on the DARwIn-OP has ranges of +/-2g, 4g, or 8g, and the L3G4200D rate sensor has ranges of +/- 250, 500, or 2500 degrees per second. There have not really been many MEMS inertial sensors released by the big manufacturers in the last couple of years that have not had multiple output ranges available.

Only the CM-730 has a built-in IMU. The CM-510 and CM-700 can use analog gyroscopes, but they are external devices. There was an AX-S20 which used a magnetometer/accelerometer sensor IC, but the company that made the sensor (Amosense - AMS0805WAH - intended for cell phones) either folded or just stopped production. Robotis has not bothered to create a replacement, although the LSM303DLH is a comparable replacement (you would have to create your own dynamixel bus device to collect the values and make them available to the CM-5/510/700 over the bus). If you use the arbotix, you should be able to directly interface it (if there is a free I2C interface - should be) to the LSM303DLH breakout board from Sparkfun. Of course there are plenty of other boards/sensors available from many sources, and it would likely be best to get one that includes a gyroscope and an accelerometer instead of one with just an accerometer and magnetometer.

11-22-2011, 03:35 PM
The "common bipeds" I have used with accelerometers in this way are Robonova and Robobuilder. The accelerometers are read into Robobasic or Robobuilder Action and determine the next motion. This happens fairly slowly (seconds usually), but allows the robot to detect it has fallen and call the appropriate stand up routine.
For walking and balance it is usual to develop a gait which is stable without gyro. The gyro or IMU is then used to allow the walk to manage uneven surfaces or outside forces, rather than to manage an unstable system. Gyro only stablisation was first started with Robonova and Kondo robots. Unlike the tilt sensing, the gyro outputs are usually fed direct into the motion engine with variable gain and filter parameters. They are working more in the 100ms or less timescale. I have not used gyro on the bioloid, but it is interesting in that it seems to read the Gyro and modify motions in Task rather than direct to the motion engine.
The Robotis DARwin OP-has both Gyro and accelerometer, though they are not combined in current software release. The DARwin-OP uses gyro to modify specific servo angles in the motion engine similar to the Robonova and Kondo.
Full IMU are used in more demanding applications like quadrocopters, they mix the Accelerometer and gyro using Kalman or complimentary filters to give fast and accurate, angle and angular velocity information. Maybe as Biped gaits move from static to dynamic walking, this will become more common in humanoids too.
The other selection criteria for accelerometers and gyro is the interface. Early devices were analog and required A2D inputs. Later devices are SPI or I2C. Get this later type, they are much more accurate.

Of course the above is current practice and convention. Use of gyro, accelerometers and IMU is still in very early stages.

11-23-2011, 09:42 PM
Thank you both for the information, very much appreciated!