rgordon
USA
Asked
— Edited
Here is a basic top view of my robot drive chassis. For navigation purposes, is this layout enough for autonomous navigation? Do I need to arrange them differently or use a different type. I need it to be able to get around reliably and get through doorways easily.
Do certain types of I.R. sensors work better for navigation?
Thanks for any input on this matter.....
Personally, ultrasonic doesn't work for me in my house, I believe it's due to the materials and colours of my furniture. For this reason I use IR more than ultrasonic. However IR doesn't like daylight.
Position wise, I plan to use IR sensors in positions 1 and 2 on your diagram and also a third IR above the front ultrasonic, plus the ultrasonic too. Noting around the back or sides (it can't move sideways so it shouldn't be a problem there), although a rear sensor could be useful when reversing.
It comes down to how many sensors you want and how many ports you have to play with. If you have 6 spare digital ports then 3 ultrasonic sensors could be useful. The same for IR. Just always take in to consideration the benchmark, mine can do around 25 reads per second
Sometimes the robot approaches a wall laterally as it moves forward because its trajectory is not parallel to the wall next to him. Side sensors are very useful for this case, as the front fixed sensor would not detect the sidewall.
I think that is a mainstay in robotic capacity accurately move our home robots and certainly hope that together we will succeed.
I made a modification of your drawing to show what would be my ideal platform for navigation (in theory, since I have not proven)
No what would my rate readings per second but I would like to try some seven IR sensors in this configuration:
IR would use two sensors on each side of the robot (4 total) for comparing the readings from both sensors on one side and correct the trajectory to be parallel to the side wall (such as in a hallway)
There was talk in other threads about sailing comparing readings to circulate side by always geometric center of the hall but slightly serious career as "S" or "Z".
With the four side sensors can navigate in the center of the corridor and parallel to the side walls.
I still like the idea of using sensors on the sides to help keep it centered when traveling down hallways and such. Would I.R. 's work better for this than Pings?
Thanks guys for replying with your ideas.
The 2 IR sensors on the front corners should help centre it on a hallway in theory, that's all I plan to do on Melvin but the proof will be in the testing I guess.
Then, basically...
If the ir sensor on the left detects something it'll turn right.
If the ir sensor on the right detects something it'll turn left.
If the ir or ping sensor in the middle detects something it checks left & right.
Much like my ping roam script but without sweeping.
If boxed in, or too close to an object the whole robot will turn rather than the sweep servo.
So far running on just the front IR he has avoided most things.
Also, with the 2 corner sensors, I'll probably have them centre up the robot in a hallway. If left ir is greater than right ir then the right hand wheels will slow down, causing him to veer off to the right slowly, until the two levels match. The same for the other side.
I could make the bumper so the corners could be adjusted, letting me experiment with the angle.
If you have them adjustable, you could always throw on a couple of mini servos and keep them adjustable, kind of like sweeping sensors but a smaller range.
Sensors Manufacturers
MOSTLY YOU PUT IR very low on your design to block sunlight
ON sonars to need to set them high or low ,but now where is will hit stuff like a sofa
ON POSITION to detect corners of wall you need it vertical so that the cone is very small.
A very common design is one set (sonar and ir) in back ,another set on the side ,1 or 2 sets facing forward on a robot and 2 at 45 deg to detect corners
MOST guys call me a sensor expert because i have and used and tested almost everyone made.
Even test some so secret i can say anything about them
How many Pings will you be using on Jarvis? Is there a limit on how many Pings you can use due to processor speed and performance issues? I am thinking Questor will end up with (8) Ping sensors.
On the front will be (3):
-one down low to the floor
-one midway up
-one on the top deck.
On each side will be (1): total of (2)
-just above the wheel section (for maneuvering in hallways and such)
On the rear will be (3):
-one down low to the floor
-one midway up
-one on the top deck.
Of course I am still planning on using bumper switches as well.
To all who are following this thread, I would interested in everyone's input on this matter.
Do you think I need more on the sides?
Do you think I need IR sensors also? Or are the Pings enough?
Second do need IR ,SINCE SONAR wont detect most stuff and same with IR,GLASS;SHINY STUFF,SUNLIGHT
I am looking at changing the IR frequency design to see if it will do better in sunlight
Real world example, mine...
15 reads per second. That would be just under 2 reads per second for 8 sensors, with nothing else happening.
Writing shows a lot more
Do a benchmark first and work out where the limits are. If you use multiple EZ-Bs you will probably find it's better to spread the sensors over them all rather than have one board for sensors and one for servos etc.
A very well made design.
just my 2 Cents.
@josh
What exactly do you mean "once I get the position where I want that works I will drill holes and poke the led sender and receiver through for a cleaner look." How do you do this?
I have experienced that my Ping units will bog down and slow down the robot if I use more than one. Maybe multi-EZBs.