S.A.R.A. the robot

Since the beginning of it's participation in the RoboCup@Home tournament, our team has worked considerably to fully develop our lovely S.A.R.A, whose name stands for Autonomous Robotics Assistance Systems in French. Her main goal is to be able to help people at home for daily tasks and our mission is to make her as functional as possible in order for her to be able to navigate in a domestic environment.


This is a companion discussion topic for the original entry at https://community.robotshop.com/robots/show/s-a-r-a-the-robot
1 Like

In it’s current state, can it fetch me a drink in the fridge?

Super cool project btw!

1 Like

Yes it could, currently we can do object recognition, object manipulation and autonomous navigation. The hardest part in this use case would be opening the fridge door :stuck_out_tongue:
And thanks :slight_smile:

2 Likes

Alright! Are you using OpenCV for image recognition?

We are mainly using YOLO V2 for objects recognition tasks.

This is a pretty good research platform, nice work! Thanks for sharing the softwares used and the BOM, very informative. I didn’t know kinova sells actuators separately, can you tell me how much they cost? I am also interested how you combine the yolo bounding boxes with the depth map.
Thanks!

Thanks a lot !
Currently the price for the KA-58 actuator is 1670$ CAD, I don’t have the exact price for the KA-75+ actuator but it must be something between 2500$ and 3000$ CAD.
For the bounding boxes in 3D you can take a look at https://github.com/WalkingMachine/wm_frame_to_box. Another guy from my team worked on it and the readme is currently in french so I’ll try to translate it ASAP :stuck_out_tongue: but basically what we do is we subscribe to the depth image and the detected bounding boxes topic, then for each boxe we calculate the mean depth.

edit
The repo is now translated in English ! :slight_smile:

2 Likes

Great project!
I saw it in real life and this is pretty impressive!

3 Likes