Assistive Subsystems

Consider an automated vacuum cleaner with the following subsystems under the command of an AI system to be designed. These subsystems limit the AI complexity to just intelligent guidance of the vacuum, much like the behavioral components of the nervous systems of any arthropods, vertebrates, or other bilaterally symmetric animals.

  • Accurate and reliable vacuum body position and orientation control through four 8" diameter rubber wheels with independent suspension and radial positioning via a separate compass and floor plan positioning system — the AI can simply command that system to go to position $(x, y)$ in room $R_i$ and face "NNW" (north-north-west).
  • Accurate and reliable tool control — the AI can simply command "brush", "crevice", "rectangle", "rug", "right-angle", or a coordinate in $\mathbb{R}^3$ with a tool approach angle for the tube onto which tools are attached, given as a compound angle $(\theta, \phi)$ relative to the orientation of the vacuum, where $\theta = \arctan(\Delta x, \Delta y)$ is the yaw and $\phi = \arctan(\Delta z, \sqrt{(\Delta x)^2 + \Delta y)^2})$ is the pitchdirection from which the tool
  • Access to a tactile interference an event queue containing a time series of significant changes in force vectors. The queue contains elements that include the force vector in $\mathbb{R}^3$ and whether the force change is being reported for force on the vacuum or on the tool.

Product Limitations

Assume the following as product limitations. (Whether or not these are the best limitations, these are the directives from the Board of Directors.)

  • The vacuum is blind. (A vision system cost to train across multiple households, manufacture, test, and update, and the associated impact on consumer price tags was evaluated and a tactile only sensory system is specified as a device engineering constraint.)
  • The room is free from foot traffic, children playing, and other such human motive activity.
  • The floor of the room is free from balls, toys, spills, cloth items, and other such artifacts that would frustrate a vacuuming operation.

Requirements of Maintaining Two Maps

An accessibility map must be maintained to make use of past information about where the vacuum can go without bumping into something. It is obviously easier to maintain an accessibility map than to rediscover a room upon every room vacuuming. This map must include tool positions also.

A coverage map must be initialized at the beginning of every room vacuuming and continuously updated to ensure that all the floor area is covered with appropriate overlapping of tool paths.

Distinct Question

Note that this challenge is distinct from the [How do autonomous robotic vacuum cleaners perceive the environment for navigation?] question(How do autonomous robotic vacuum cleaners perceive the environment for navigation?) in that it is not about discovery of room and furnishings location but rather assumes floor plans are installed upon product deployment (which is more likely to be a highly effective short term product strategy). The problem of adapting to changes in the accessibility map is the challenge here.

How can a fully automated vacuum cleaner use and update room information?

See above two maps.

  • $\begingroup$ My impression is, that in the last time some questions here are repetition of research from the past. If my memory is right, we had last month a question in which somebody wanted to parse a mathematical text description into an equation, which was discussed in a paper from the 1980s. This time, a problem with map building for a vacuum cleaner is there. The question itself was very accurate formulated, so I've searched a bit at Google scholar and found that exactly the same problem was discussed in the literature of the early 1990s. The reference is available with a keyword search for $\endgroup$ – Manuel Rodriguez Oct 16 '18 at 16:47
  • $\begingroup$ ... “agent architecture grounding map vacuum cleaner”. So I have to ask back: 1.) Is the question about a problem, which is already solved? 2.) Is it a literature question, which means the answer would to identify correct paper titles and there is no need to describe the architecture again? $\endgroup$ – Manuel Rodriguez Oct 16 '18 at 16:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.