There was a recent accident where SpaceShip2 broke apart in mid-flight, killing one and injuring another.
Given that we can remotely control drones from the other side of the world, why couldn't we do the same with this test airplane?
Considering that it's all launched locally, it would not even have the lag that drones have.
Today's highly automated airplanes should be not too hard to control remotely - after all, all the cockpit control inputs become electrical signals, and at that point a multichannel remote could be plugged in. The Russians even flew their Buran version of the Space Shuttle autonomously, using a neural network to command the craft on its single flight in 1988.
However, a test pilot flies an aircraft with all his senses. Without the acceleration, a full 3D-view and the force feedback from the controls, he doesn't get the full picture when sitting in a control room and looking at screens full of data. That is done by test engineers anyway, and the pilot's job is to complete the observation of a new airplane's behavior by exposing himself to it.
You might argue that the 3D-view is just a matter of several cameras and displays, but the needed telemetry bandwidth for this would be immense. In a normal test flight, already hundreds of parameters are measured many times a second and transmitted to a ground station, where specialists for each system watch for anomalies in the data. It is much better to spend the telemetry budget for this than to send pictures down. The lack of acceleration and vibration alone would deprive the test pilot of an essential part of the picture, because it simply cannot be realistically recreated in a ground station.
Also, the inevitable delays incurred by radio transmission and processing for display will quickly put a pilot out of a loop which he could easily control if he were sitting in the airplane. The range of controllable eigenfrequencies would be reduced to phenomena happening at maybe 0.2 Hertz or less, whereas a pilot in the cockpit can comfortably stay ahead of oscillations with 0.5 Hertz or less.
However, it must be said that sometimes the test pilot does turn out to be a burden for the progress of the flight. On the second flight of the British TSR-2 prototype, one fuel pump began to develop an imbalance, oscillating just at the eigenfrequency of the pilot's eyeballs. He had to throttle back one of the engines to regain his vision.
Taking a plane that is designed to be operated by humans and retrofitting a system for remote control would be a big task. They do it with old fighter jets for use as target drones. But a target drone's task is pretty simple; fly like a fighter jet until it gets shot down.
At the point where SpaceShipTwo is, a bunch of testing has already been done. This is supposed to show that the vehicle is safe enough to move on to piloted testing. But testing aircraft and spacecraft is a dangerous job and things happen.
Creating a remote control system for an spacecraft like SpaceShipTwo would be much more of an issue. First of all, there will most likely be physical changes required, whether this is to install hardware to manipulate controls, or just electronics to remotely operate the flight computers. These physical changes will be extra work and will affect the tests, since when the plane is actually operating, those things won't be in place. And if those remote control systems fail, you lose the aircraft. Sure, you don't lose pilots, but why lose a plane just because a remote control system (which won't be used on production) fails? Should you spend time and money developing a complicated remote control system, or developing the actual vehicle?
Also, many tests require the pilots to be on board. They are not just testing the systems, but how the pilots interface with them. Having real pilots actually flying the vehicle is the whole point of this phase of testing.
See this related question as well: Why do the engineers need to be on board during testing?
Remote control systems have bugs and problems the same as any other system.
If you install a remote control system, you now have two systems that need to be tested, checked, maintained and working properly.
If there's a problem, and the vehicle crashes, you're immediately left with two possibilities: maybe the aircraft had a problem, or maybe the remote control system had a problem.
If the remote control system had a problem, maybe your airplane was fine, and you just lost a perfectly good plane due to a bug in your test system!
Barnes Wallis asked this question in the 1940s.
The loss of life in the Dams Raid deeply affected him and he was adamant that he would not risk human life on his more experimental post-war designs until their basic concepts were proven through unmanned flight. (Source for this and details below)
These included the revolutionary Wild Goose and later Swallow designs, where variable geometry wings were used, not just to improve wing efficiency across a range of flight conditions (as in the much later F-111 and Tornado) but to eliminate the tail and its associated weight and drag penalties altogether. Wild Goose flew (remotely controlled) in January 1950.
However testing seems to have been slow and remote control techniques of the time didn't help. Without onboard cameras, at least one model was crashed by a pilot misjudging its position from a distance.
Swallow (in the form of a rocket propelled model) is reported to have flown stably at Mach 2.5 in 1955.
Neither progressed to manned prototypes, thanks in part to financial difficulties in post-war Britain.
In any case the intent was to prove the basic concepts rather than perform all testing in unmanned form, experimental manned planes were to be built along proven aerodynamic lines.
