PDA

View Full Version : Navigation Sensors

counter_struck
02-29-2008, 08:05 PM
Hi, I am an electronic engineer student. I have been given a project to design a mobile autonomous robot. My part in the project is the design of the navigation sensors and their software implementation. We are supplied with a toy car and must use this for our design.

I am having difficulty figuring how to estimate the position of our robot using ultrasonic and infrared sensors. We need to move our car from point A to B. However our problem is, when we aviod obstacles how do does the robot car get back to the course, and not wander around?

Any help will be appreciative :D

rao
02-29-2008, 10:21 PM
IT would be easy, if you can deduce a relation between two things :

Let the robot turn with 0 angle to the axis at a fixed velocity,same in magnitude,opposite in polarity on both wheels.

Now deduce a relation between the time taken for the robot rotating on a fixed axis (i.e. delay time to the motors ON command) and angle it rotates in that much time. So you can make a function called angle where when you give a certain angle, it is multiplied to a conversion factor and converted into a delay of milliseconds or something for which the robot turns to attain that angle.

Now, as soon as an obstacle is detected, define two arrays, angle and distance and sequentially store the values of angle turned and distance travelled in that angle to avoid the obstacle , after the object is out of range, track back on the array by inverting the angles (+ becomes -) and keeping distances the same. But to track back, follow the array from the last element to the first.

Try that out, if you face any problems, Let me know.

Rao

counter_struck
03-01-2008, 01:55 AM
Your approach to my problem sounds a lot like the process of dead reckoning. I also thought this would be a solution to my problem. However the course is undefined. The undefined course maybe a track from one part of our university corridor, to another. The doors along the corridor might be open, and in avoiding obstacles, my fear is that the car will go into a room than continue on the course.

I was thinking that if car is employed with some longitude or latitude device, having an algorithm with initial values of starting and stopping positions would ensure even if the car went off track, if would find its bearings to the stop position. GPS is out of our price range.

Camera's and image processing are expensive and will not be feasible given our short time period.

I don't understand how we will manoeuvre our car to get to the stop position without knowing our course before hand and making landmarks

rao
03-01-2008, 04:58 AM
OK..

I am sorry, I didnt know that you have already tried that out..would have saved me a lot of effort spent in writing the paragraph.. Dead reckoning...yup

You can try a compass...and retaining the actual direction at all times

If you want to make your own sensor let me know

Rao

mvharish
03-03-2008, 08:11 AM
I'm not an expert, and I've never tried this sort of thing, but here goes:
You get these things called digital compasses sort of ICs on a small board.. check them out.

GPS is useless for resoultions of < 20-30 feet. Anyway, have you checked these out:
http://www.sparkfun.com/commerce/categories.php?cPath=4_17
?
Rs 2400 - 3000 isn't a lot, and it looks like cutting edge stuff.

RoboticMax
04-23-2008, 04:21 PM
parallax.com offers a cheap GPS reciever for \$89. Even i'm building an autonomous gokart which is equipped with a GPS+IMU coupling. for collision avoidance, just vary your lateral offset from the reference trajectory by generating a set of parallel linear eqns. on a cartesian plane so dat u simply go past the obstacle without hitting it.

RoboticMax
04-23-2008, 04:27 PM
dear struck, i strongly recommend an inertial measurement unit so that u always have sense of vehicle orientation and if possible, try to make interpolations between IMU readings and GPS. btw, first construct an occupancy grid of your surroundings and employ simple cartesian-polar coordinate transformation eqns for collision avoidance. best of luck!!!(4 me too)