To succeed in the marketplace, consumer robots must do something consumers value. They also must be priced competitively with nonrobotic means of accomplishing the same task. Both constraints profoundly affect robot design.
Robots must ask certain questions as they perform a task while moving safely about their world: "Am I about to tumble down the stairs?" "Have I collided with an object?" Robots choose actions based on the answers, which sensors provide. Robot-control theory has advanced to where we can be confident in our ability to write a program to perform a given task—if we can obtain unambiguous answers to all of the questions the robot must ask.
Key to why robots currently carry out so few tasks autonomously is the mismatch between the questions a given task forces the robot to ask and the answers existing sensors can provide. To make progress, we have alternatives. We can build special-purpose sensors that answer a single question (point an infrared emitter/detector pair at the floor in front of the robot to determine if there is a drop ahead). We can specialize a general-purpose sensor like an electronic camera with software. (Sensors that track color blobs are built this way.) Or we can configure the task so the robot needn't answer difficult questions that at first seem essential. (Current robot vacuums don't sense dirt remotely or know what part of the floor they've already cleaned.)
Cost constrains the search for new task-enabling sensors. Consumers don't want to pay much more for a robot than they would pay for a manual device or service to do the same thing. This imposes a strict frugality on robot designers. Every sensor and system must earn its keep. That's why the impressive but expensive lab-based robots in the news never make it into commercial production.
The first waves of practical robots are entering consumers' homes, and robotic development should accelerate. But the challenges of devising sensors for each new task and accomplishing tasks at a low cost means the expansion of consumer robotics will be less frenetic than the growth of the Internet.