Yes, there are many legal hurdles to clear--some of them the same or similar to the driverless car--but I predict that by the time the algorithns are up to the the task, the law makers will have come to some sort of agreement with the service providers.
- Issues of interference in the voice (various noises)
- Distance from the robot (that means a human needs
to be right next to the robot to be heard).
Remote commands are simpler, easier to have clarity,
simpler to recognize/follow without errors.
Think of it has the robot may be in a remote location, and perhaps a hazardous location. (For example space exploration). You would want simple commands to input remotely, and you would have intermitent communication issues and longer duration communication issues.
The robots still need to react to their surroundings.. (Balance, unforseen obstacles, falling, terrain, perhaps even damage like another robot running into them).
The idea of the degraded channel is a good one. In the future I want to see the ability for hackers to try and disable or take over the robot via the communications channel. Need security on these designs...