IR doesn't work outdoors in daylight because that huge yellow IR emitter in the sky kind of blows everything else away. Sonar sensors, like the ones used on the bottom of the Parrot AR Drone work better. There are sonar sensor kits for Arduino. Of course, any quad would have to fly very slowly for a controller to find its way around obstacles even with good sensors.
And actually, the whole flying and 'quad' part comes last anyway. First you need to setup a testing prototype to determine the best placement for sensors, collect returned sensor data and figure out how to have that drive the controller. That would be best achieved with a board atop a stand on a rolling platform with obstacles placed at known distances and video taken of the transit that can be studied with the sensor returns overlaid on it. Then you determine the necessary parameters, triggers and logic. You'll probably start with three sets of sensors, left at 45deg, straight ahead and right at 45deg -- with enough axis delta to stop the returns from one being intercepted by its neighbor. Then it's just a matter of running measurements and decision logic through a loop. The tricky part is converting the output to simulated joystick inputs! Only when you're ready to start testing that conversion and control process do you need to introduce the rotor platform.
I like the idea of an APM controller. You're on the right track there.
Oh, and a little trivia: Mauritius is exactly on the opposite side of the earth from Los Angeles.