I'm currently working on creating a robot for a capstone that moves in an environment where humans will be milling about indoors and will occasionally approach the robot to interact with it. My first thought was to detect people with AI using a depth camera with RVC2, but eventually found SLAM as a technique that sounded promising. I figured ROS2 was the perfect framework to work with since I have to also integrate robot movement base commands, distance sensors on the outer frame, and a manual override for movement.

I've looked into RTAB-Map for mapping and localization, but the thought of being out of scope on using this package, and subsequently spending more money on potentially pointless equipment, leads me to asking here if I'm looking down the right path with ROS right now. Here's the gist of my questions:

  • Is what I want to accomplish more or less complex than I realize?
  • Should I be looking elsewhere for this kind of complex problem or acquiring other specific equipment?
  • I have a depth camera (that can also run AI models) and successfully ran RTAB-Map as is, but I am considering adding LiDAR for increased performance and "fake" odometry. Is this a right choice, if RTAB-Map is also a good path?
  • I have an Xilinx KR260 FPGA with ROS acceleration via KRS, but are my expectations of doing all of this in real time on it excessive?
  • Anything else I'm missing I may need to consider?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.