Building a semi-autonomous robot for in-door navigation and sensing

Hi all,

I need to know if the below project is feesible.

I want to create a semi-autonomous robot for navigating within buildings. This would help disabled persons find things like the toilet, locate various offices etc. It could also be used for security monitoring. As of now, I am considering a device on wheels. My plan is as follows.

  1. Use a board like the arduino Uno to connect the following sensors.
    1.1. A compass.
    1.2. A distance sensor.
    1.3. A wireless module for remote communication.
    1.4. A Bluetooth module for communication with mobile phones.
    1.5. Motors, with encoders.
    1.6. A sd card reader and writer to store maps.
    1.7. A camera.

  2. The way the machine would work is as follows.
    2.1. The interface would be an application which would run on something like the raspberry pi.
    2.2. I may migrate this to a mobile phone which is why I need the Bluetooth module.
    2.3. The user will select where he is and will select where he wants to go.
    2.4. The machine will then guide the user to that location.
    2.5. If there is an obstacle like a chair that has been kept out of place, the machine will go round the obstacle.
    2.6. The other thing the user could do is to assign an area for the machine to patrol. It would then use the camera to check things like if a door was open or if a window was broken etc.

  3. In terms of implementation, I plan to do the following.
    3.1. Get the arduino robot because it already has a compass and a sd card reader and writer.
    3.2. Attach a distance sensor, a wireless module and a Bluetooth module to it.
    3.3. Program away.

Is the above project feasible given the way I have described? Yes, I know, this is a very complex project and will undergo several iterations but am I thinking correctly? I plan to create maps of the building in xml. The maps would consist of the heading and the distance travelled until a given obstacle or turn occurs. I have looked at indooratlas that does what I want but on a mobile phone but I don't like the idea of sharing my building's plan on the cloud.

1.1. A compass.

Compasses find magnetic north. Indoors, that seems a relatively useless piece of information. They tend to drift, too. You'll need a way of knowing which way the robot is facing when the robot fires up, with respect the stored map. You'll need a way to periodically, recalibrate the robot.

1.6. A sd card reader and writer to store maps.

A map implies that you know where you are. Knowing the direction you are heading, without knowing how far you are from some known location is meaningless. Knowing that the toilet is 48 degrees east of where you are is useless unless you also know how far it is. The distance sensor is useful, for avoiding obstructions, but it is useless for determining how far you have traveled.

How will the map be constructed? How will you deal with the janitor leaving the mop bucket in the middle of the hallway? How will you deal with temporary closures while the hall is painted/carpeted/cleaned/mobbed?

1.7. A camera.

For what? Might as well tote around a lucky rabbit's foot.

Is the above project feasible given the way I have described?

Individual pieces are quite feasible. The entire project may not be. That depends on your motivation.

Yes, I know, this is a very complex project

OK. That eliminates the need to tell you that.

but am I thinking correctly? I plan to create maps of the building in xml.

Oops. There I was thinking you were, when you spoiled it. XML has it's places, but mapping is not one of them.

I plan to create maps of the building in xml.

You probably need to background research on what others have done in robotic mapping, and see if you are up to the challenge of developing a mapping robot. At this point have you successfully developed/implemented any of the individual element of your project?

https://www.google.com/search?as_q=mapping+robot&as_epq=&as_oq=&as_eq=&as_nlo=&as_nhi=&lr=&cr=&as_qdr=all&as_sitesearch=&as_occt=any&safe=images&tbs=&as_filetype=&as_rights=

pranavlal:
I need to know if the below project is feesible.

With the hardware config you've listed I doubt very much it would be feasible unless you are using the camera to recognize landmarks for localization, and even then that is not a trivial undertaking.

I've just started my own "NavBot" project and from what little I've done so far it is clear that autonomous navigation on the scale you describe (i.e. inside a building) is extremely hard, maybe even impractical if you cannot tightly control the environment.

You'd need to expand more on the details of the project in order for people to be able to provide useful feedback. For example can you place markings on the walls or floor that the robot can use for localization?

With the information you've given so far I'd have to say it's not feasible.

Hi all,

Many thanks for your responses. I have been looking at mapping robots and need to do a fair bit of reading. In addition, that was the key phrase I was missing.

Yes, I am also reading about path finding algorithms. I plan to implement the project in stages. I have got obstacle detection working and will take things from there assuming I don't get distracted! One interesting reference to localization and mapping I found is
http://rossum.sourceforge.net/papers/Localization/PosPosterv4.pdf
Has anyone tried it?

Pranav