Best indoor positioning system/technology for in-home use?

  • Thread starter Thread starter Comprehensive-Ad3963
  • Start date Start date
  • Tags Tags
    Navigation
AI Thread Summary
The discussion focuses on developing an indoor positioning system for a robot designed to assist users in a residential setting. Key technologies considered include computer vision for identifying doors and users, Bluetooth Low Energy beacons for user location tracking, and inertial navigation for the robot's positioning. There is a debate over the effectiveness and cost of hardware solutions like lidar versus software-based approaches, with a suggestion to utilize inexpensive cameras and software libraries for image processing. Additional ideas include using wearable devices for the user to enhance tracking accuracy and memorizing the user's appearance for better identification. The conversation emphasizes the importance of combining various technologies to achieve reliable indoor navigation and user interaction.
Comprehensive-Ad3963
Messages
4
Reaction score
1
I'm building a robot to be used in a residential setting. This robot's primary directive is going to be to open a door in someone's house, and then come to them and tell them this has been done. This will require the robot to be aware of its location within their house (or, at least, its location relative to the user)

Computer vision might be sufficient for identifying the door to be opened and then for identifying the human once either is in the robot's line of sight, but I'm not entirely sure about any other case.

Maybe I could have the user place Bluetooth low-energy beacons in every room of their house. Maybe I could combine that with computer vision for the robot to determine which room it's in (if it looks like the kitchen and the kitchen beacon is closest, the robot's probably in the kitchen.)

Maybe I could have the bot start from the charging station and then go back once its tasks are finished, and then the bot could use inertial navigation to determine ITS position, and then I can use a different technology to calculate the user's position.
 
Physics news on Phys.org
It is a hardware versus software issue. Cameras interfaced to computers are very inexpensive today.

There are software libraries for image processing that you could use to find the image of the robot. You'll need CPU power and you'll need a lot of learning to figure out how to use the libraries.

Hardware solutions like lidar may be simpler to understand but much more expensive and hard to make them versatile enough.

My recommendation, do it with software and get a 14 year old to do the software for you.
 
  • Haha
  • Like
Likes hutchphd, sophiecentaur, russ_watters and 1 other person
anorlunda said:
It is a hardware versus software issue. Cameras interfaced to computers are very inexpensive today.

There are software libraries for image processing that you could use to find the image of the robot. You'll need CPU power and you'll need a lot of learning to figure out how to use the libraries.

Hardware solutions like lidar may be simpler to understand but much more expensive and hard to make them versatile enough.

My recommendation, do it with software and get a 14 year old to do the software for you.
OK, but how do I let the robot locate the user? What if the bot's in the kitchen and the user's at his desk in the office?

I could use Bluetooth Low Energy beacons sprinkled throughout the house to locate the user, but then I might as well make the robot aware of them, too, and thereby (potentially) increase the robot's ability to accurately identify the room it's in.
 
Can the robot just call or text the person on their cell phone? Is there a reason the robot has to go find the person and use voice synthesis to tell the user?

Can you provide more context for this project? Is it some kind of a challenge where you are competing against other Makers to come up with the best system for a set of requirements? Or are you trying to come up with something that can help an elderly family member be more self-sufficient in their home?
 
berkeman said:
Can the robot just call or text the person on their cell phone? Is there a reason the robot has to go find the person and use voice synthesis to tell the user?

Can you provide more context for this project? Is it some kind of a challenge where you are competing against other Makers to come up with the best system for a set of requirements? Or are you trying to come up with something that can help an elderly family member be more self-sufficient in their home?
A challenge, sure.
 
anorlunda said:
You'll need CPU power and you'll need a lot of learning to figure out how to use the libraries.
Even if I implemented some simple neuron network to classify images for me I never used Edge TPU by myself, but I was told by a friend of mine it is surprisingly powerful for the price - and is sold as just a USB device that can be plugged into anything. It could be a wise choice for the problem at hand.
 
  • Informative
Likes anorlunda and berkeman
OK, so I have a few ideas.

1) Have the user wear two antennae a set distance apart and use time-of-arrival/time-difference-of-arrival to calculate the robot's distance to these antennae and then its position relative to the antennae. The bot would also know the user's position relative to these antennae and could therefore attempt to calculate the best route.

2) Have the user wear a device that is connected to at least three beacons that are located throughout the house. Measure the time-difference-of-arrival between the device and the beacons.

2) Have the robot memorize what the user looks like from all angles, so that once the robot gets close enough to the user, it can rely on its camera feed to find the user.
 
Comprehensive-Ad3963 said:
and then come to them and tell them this has been done.
That does not specify the "come to them" distance. Open the door, and loudly play "I'm here." How does that not satisfy the challenge?

If you have a more precise version of the challenge, please post it.
 
Comprehensive-Ad3963 said:
1) Have the user wear two antennae a set distance apart and use time-of-arrival/time-difference-of-arrival to calculate the robot's distance to these antennae and then its position relative to the antennae. The bot would also know the user's position relative to these antennae and could therefore attempt to calculate the best route.

Your user doesn't look to happy...

1641512036030.png

https://www.zoo.com/quiz/its-your-favorite-my-favorite-martian-trivia-quiz
 
  • Love
  • Haha
Likes dlgoff and anorlunda
  • #10
This Youtube video is a little long but for those interested, it's Season 1, Episode 1:

 
  • #11
I was wonder if perhaps a bright dog could do the required job. I guess that wouldn't count. (The solution - not the dog)
 
  • Like
Likes berkeman
Back
Top