Someone whose name I can not recall once said there was no point in simulating robotic behaviour unless you can simulate everything. That is to say, you are better off building one to see what problems are instead of trying to find problems by simulation. I saw the wisdom of this today, when I took photovore1 to play in the Sun.
Photovore1 is slightly asymmetrical, that is it tends to go to the left. An interesting thing happens when its left FLED is in the shadows - it follows the edge of the shadow! This behaviour comes about because the FLEDs are inhibited by light. So the right FLED in sun light is inhibited from firing, so the left FLED fires and pushes the robot right (the FLED trigger motors on their own side). However once the left FLED has the shadow, the aforementioned asymmetry comes into play and photovore1 veers to the left again! Effectively photovore1 follows any shadow on its left.
Neat huh? Completely unintended consequence that arises out of its constructions and its interactions with the environment. Something simulation would not have found :-D