[youtube=http://www.youtube.com/watch?v=–YZHMqHzPw]
[Vincent] started building this tank (translation) with a regular hobby model: the Heng Long Tiger 1. However, after considering some goals for the project, he decided to nearly gut the tank and redesign it, basing it on the Arduino and a standard Motor Shield. The possibilities with this setup are nearly endless. In its current form, the ArduTiger detects obstacles in front of it by way of three servo-mounted infrared rangefinders. The tank’s trajectory can be adjusted automatically based on feedback from the servo positions. Two additional short-range rangefinders detect if there is ground for the tank to roll over, keeping it safe from cliffs and black holes. [Vincent] plans on updating this beast by adding a Raspberry Pi for live video and advanced control… and maybe even adding a Geiger counter!
Nice, but I’m still waiting for somebody to add IR targets finding capabilities to their tank so that we can have an autonomous fight:)
http://trandi.wordpress.com/2011/05/14/tiger-1-bb-airsoft-rc-tank-%E2%80%93-v3/
dan
Those IR sensors will give inaccurate readings while moving. Probably better not to sweep them, and just to mount them fixed.
How come that they get inaccurate when moving? The few micometers displacement or small fraction of a degree in angular movement can surely not affect the measurement?
Ultrasonic sensors might be a different story though…
Actually the biggest problem I had was the delay between the IR measurement and the reading of the direction of the servo (as I use it to detect the direction of the obstacle). Obstacle detection was off by approximately 15°. With some trim it is quite accurate. I have to sweep the detectors as their field of view is very narrow and they can miss an obstacle if it is not right in front of them. The sweep is calculated to have a very slight overlapping of the detection fields of each sensor.
He should really buffer the input of the sensor so when it is time to turn it already knows if something is in the direction of the turn and then it can smoothly use a larger turn or turn the other way, or in other words make it more ‘intelligent’ by anticipating. Perhaps also make it focus its ‘stare’ more in the direction of a turn at the start of a turn, since the direction is decided already then and it can look more in that direction to see how much it needs to turn.
It’s a general thing with such sensing robots that they don’t seem to do that though, it’s not just in this case, in fact even NASA projects often don’t include such anticipation coded in.
Obstacle-avoiding? It should be blasting those obstacles to smithereens ;)
However… get project :)