Society of Robots - Robot Forum

Mechanics and Construction => Mechanics and Construction => Topic started by: sinista on October 08, 2010, 02:54:53 AM

Title: question about motors
Post by: sinista on October 08, 2010, 02:54:53 AM
Hey,

Im at uni making my first robot as part of my final year project, (studying software not robotics)

for the project i am to make a robot car (more than likely a modified Heng Long tank) that travels arround, as it goes it will take distance measurements of objects and send them to the host PC , when there i will use the distance to draw an occupancy grid (map of the robots surrounding enviroment)   

my question is how do i actually know the robot is moving without having to look at it? Im obviously going to need to know its speed and direction to plot on the grid,

for going forwards and backwards i can use a distance sensor, but what about left and right turns,

my project supervisor kinda hinted that i could attach a webcam find a blob in the image and track it, if it moves this will notify the program that the robot is turning and should also give me the speed and direction (which sounds ok i guess, but webcams are unreliable, because of the changing light)

I was just wondering what you guys thaought, maybe there would be a better, easyier or more reliable way to do it

Thanks in advance :)
Title: Re: question about motors
Post by: knossos on October 08, 2010, 04:42:58 AM
Would encoders on the wheels give you sufficient accuracy for your application?
Title: Re: question about motors
Post by: sinista on October 08, 2010, 04:59:24 AM
cheers I didnt even think of that,  are there any other ways to solve my problem ?
Title: Re: question about motors
Post by: knossos on October 08, 2010, 08:33:02 AM
Well if you know objects on the course are stationary, you could mount the sensor on a servo and then calculate how much your angle has changed by the amount your servo has to adjust to track it.  Similar to the webcam only done with a sensor.  Problem is this fails with a mobile reference point.

Multiple sensors (or even multiple readings by the same sensor) could also use multiple reference points to triangulate location.  That way even if you choose a moving reference, the other references would help to reduce error.

My suggestion is to use a combination of methods to increase your accuracy.  I would probably go with encoders + servo mounted sensors.
Title: Re: question about motors
Post by: ballbreaker on October 13, 2010, 10:51:17 AM
how about gps
Title: Re: question about motors
Post by: knossos on October 13, 2010, 11:10:41 AM
GPS is a great tool, especially for traveling long distances and faster speeds, but it also has a significant amount of inaccuracy when dealing with shorter distances and slower speeds.  I would still recommend having a variety of other sensors for autonomy and (if you choose to include it) reference the GPS for navigation (for example waypoint routing).