I think you'll find the difficulty here is not with the thruster or logic restrictions, but the sensor restriction.
A pair of sensors have numerous capabilities that simply can't be replicated by a single sensor; without these abilities complex behaviors are very difficult to coordinate. This is essentially why one sees so few active one-eyed species in nature. With two directional sensors it becomes possible to detect distance and to establish closed or open-ended zones around the drone, generating a much larger possible variety of signals that can then be processed by a logic core. Three or more sensors can open up some additional possibilities, but most complex behaviors can be achieved by two sensors. This time, note the scarcity of three-eyed species. The increased performance of a second sensor will almost always outweigh its mass and part cost.
You will also benefit from making use of the tolerances on your sensors. While the 'Right' and 'Left' outputs can only be adjusted clumsily in non-90 degree increments (Devs, if you're reading this, round directional sensors, or the ability to turn the sensing portion within its block, please?), the tolerance can be easily fine-tuned, or even turned into a third output by using a pair of 'Not' gates on the sensor outputs and an 'And' gate to combine them.
Hopefully those points are of some assistance, though my ultimate conclusion is to agree with Garheardt. Under the proposed restrictions it's likely not currently possible to build a drone that simple that doesn't move erratically. Add a second sensor and make use of those tolerances though, and you shouldn't have too much difficulty.