What electronic compass should I get for my Robotics Starter Kit? Is CMPS03 ok?

[Xandro]

Hi!

I´m working on my Final Project with the Labview Robotics Starter Kit (the 4-wheel version), and I need an Electronic Compass.

We had a CMPS09 in the lab, but it´s toasted 😞 So I did a little research and found out that CMPS03 has integrated modules in LV2010, which would help A LOT.

I´m asking for advice, Will CMPS03 do the job? It has a resolution of 0.1 degrees and a deviation of 3-4 degrees, which doesnt seem very much.

The objective is to develop a "go-to" algorithm in a previously know map, so it´s needed to track as accurately as possible the movements, position, etc, of the robot. It´s not complex at all.

[Brian_Powell]

---

@Xandro wrote:

The objective is to develop a "go-to" algorithm in a previously know map, so it´s needed to track as accurately as possible the movements, position, etc, of the robot. It´s not complex at all.

---

The CMPS03 is only going to give you a heading in a 2D plane.  Is that sufficient for you to figure out "movements, position, etc."?  It depends on your application, of course.

Will your robot ever be tilted, and do you need to know that?  If so, the CMPS03 by itself won't be able to measure that.  (Unlike the CMPS09, which can.)

Depending on the accuracy you need, the wheel encoders can give you an idea of the forward and backward motion.  Note that the 4-wheel DaNI is essentially a skid-steer model, and the amount of skidding on turns is hard to measure.  So, the accuracy of the distance measurement isn't great.

If you need more accuracy, you could add other sensors, and fuse the data with something like an Extended Kalman Filter (EKF).  You could add an IMU, or do optical velocimetry with a camera or LIDAR.  Lots of options for more accuracy, but they add to the cost.  You could also doctor up the robot to change the wheel mechanism so that it wasn't a skid steer model any more.  (You could make it look more like the 3-wheel DaNI, which doesn't skid.)

Good luck!

[Xandro]

Thank you for your reply.

The robot has to go to point A to B in a previously know map that doesn´t has ramps or anything like that. My algorithm flows like this:

1.- Apply the A* algorithm to get the path.in the map.

2.- Translates that path into an array of instructions (go X meters in straight line, turno 90º, go Y meters straight, turn -90º, etc)

3.- The robot follows the instructions one by one with a while loop, keeping track of the moviments with the wheel encoders (straight lines) and electronic compass (90º turns)

The robot will only turn when is stopped in one spot, it won´t turn while driving forward, and it will turn only 90º or -90º (left/right). I though on that to prevent big errors with the odometry.

I know it´s a bit rudimentary, but it´s the first project with the robot, and the goal is to have a good starting point (a little collection of control/teleop/movement algorithms) for future developments with LV Robotics.

So no, the robot shouldn´t tilt at all, and I already thought on the turning issues because I saw that turning was a pain in the ass while developing the odometry for the robot.

I worry for the precission of the wheel encoders/compass. Would it be enough to make the robot navigate successfully (asuming that the algorithm works flawlessly, of course)? I don´t have previous experience with robots, and I read things like the electronic compass go wild when they are near metal objects... and the robot is made of metal

[Brian_Powell]

Hard to give a good answer on the compass without trying it.  Electric motors can also cause problems, and DaNI has a couple of those, too. 🙂

On one of our robots, NIcholas (http://openmeas.blogspot.com/2009/02/short-robot-video.html ), we put the IMU/compass up on a mast, away from the electronics and motors.

The good news is that I have had better-than-expected success, but your mileage may vary.

[Xandro]

It´s hard to find any part in the Starter Kit to put a mast (My SK has the rear crowded with a router, the front has bumpers on it...) but I´ll think about it, in case the compass doesnt work well, it´s the only way. Thank you very much for the idea!

In the end, the university bought one CMPS03 and one CMPS09, so I´ll have room to experiment

Only one more question... with your experience, do you think the SK wheel encoders have enough precission for the job? Do I need to add more sensors? (I understand this is impossible to exactly know without actually testing it, but I appreciate any advice).