Electronics > Electronics



Hello. I need some help regarding sensors theory. Can anyone help?

Which sensor?

What are trying to do?

possibly useful . . .

Thnx. I've visited the link.

This is what we (me and 3 colleagues of mine) have to do (this stage is only theoretical):

Let's assume we have a robot with the following sensors on it:
   - odometry,
   - sonar,
   - pan-tilt stereo camera,
   - gyroscope,
   - IR (infra red),
   - laser rangefinder,
   - pyroelectric,
   - GPS (Global Positioning System).

Ok. First of all I need to read something about this sensors.  :D (So far so good).
Next, we have to figure out a data structure for each sensors' retrieved information. (for instance: distance, angle and so on. Each sensor with it's specific return data).
Next, we have to use some algorithms to correlate the data from these different sensors
Next, we have to determine the robot's position
Next, we have to implement it in hardware

So, my request is the following:

Could you give me a link to a tutorial on sensors so that I can figure out what kind of data each sensor retrieves. (the odometry as far as I know works with data of type (x, y, angle) for example).

Important note: we are not concerned how the robot gets the info. We just assume it has it all there. All we have to do is "understand" and manipulate it.

Thanx, dan.

You should google each of those terms, there is a wealth of info online for each of those sensors. For example, for gps, type "robot gps."

Consider what environment your robot operates in. For example, each of those sensors will operate very differently (or not at all) depending on if your robot is in a house, out in a forest of trees, or underwater.

Consider if your sensors are scanning, or fixed with respect to the robot. For example, scanning sensor data for a fast moving robot will not be the same as a not moving robot.

Consider what you need to detect with your sensors and why. For example, sonar detects soft objects very differently than hard objects. The color of different objects will affect IR. Think about what data you can ignore, and what you must interpret.

Consider accuracy - high accuracy requires slowness for scanning and processing. Low accuracy means your robot can operate much faster, but there is error. You must decide how much error you are willing to accept. There will always be error.

Lastly is sensor cross-comparison. Each of your sensors will not tell you exactly the same thing. Which sensor should your robot trust the most? Your algorithm should add more weight to your more accurate sensors (or sensors that would be more accurate in a certain situation). For example, if your robot suddently finds itself inside a building full of pillows, your GPS will stop working entirely, your sonar will become somewhat inaccurate, but your laser rangefinder will still work perfectly. A forest of trees however would make a laser rangefinder almost useless.

In the end, your robot is trying to piece together a map using all of the sensor data. So that means the final interpretation of all data should be x/y/z coordinates for a map, and/or extra information to 'identify' the objects found in that map. And what will you do with the map? Navigate perhaps? That will decide on what kind of map you need, and what to do with it.

Hope this gets you thinking. Good luck!


[0] Message Index

Go to full version