rvc2 and rvc3 : which one do i have ?

smoothVision

Hello,

i have a oak-d lite and would like to do slam for a mini autonomous robot that i have ( i'm trying to write the code on ros2 also) and for now i have only an obstacle avoidance algoithm in plain python that prints directions o the terminal if it detects objects. i don't understand 1) which rvc the oak d lite has ? 2)if the oakd -lite is capable of doing so on device. Otherwise if i run the algorithm on the host computer like a raspberry pi 4 or 5 isn't affected with a ceratin delay?

thanks,

PS. any suggestions on how i should proceed to my project is appreciated.

jakaskerl

smoothVision
Oak-d lite is built on RVC2. If you write in plain python, you can do so on the device (keep in mind the CPU is not powerful at all so it may introduce latencies). I'd recommend using RPI for this.

Thanks,
Jaka

smoothVision

what do you mean with "If you write in plain python, you can do so on the device" , do you mean there are other alternatives ?

What is the best approach to reach my goal ? i have both rpi 4 and 5 ( and probably i can order a jetson nano ) , and if you suggest me another luxonis camera i can probably buy another one. which other camera of yours , do you suggest to do vio / slam and object detection on the camera ?

Best apporach ? my approach is not to use raspberry pi to run everything. I would like to run ros2 on the raspberry and object detection model or vio/ slam on the camera ( another option is to change to a more powerful board like rpi 5 or jetson nano ). What do you suggest ? do you think my approach to create an autonomous robot is fine ? i'm open to any of your suggestion while using ros2 as the central software.

Thanks.

jakaskerl

smoothVision what do you mean with "If you write in plain python, you can do so on the device" , do you mean there are other alternatives ?

things like cv2 and numpy and such are not supported inside the Script node.

smoothVision Best apporach ? my approach is not to use raspberry pi to run everything. I would like to run ros2 on the raspberry and object detection model or vio/ slam on the camera ( another option is to change to a more powerful board like rpi 5 or jetson nano ). What do you suggest ? do you think my approach to create an autonomous robot is fine ? i'm open to any of your suggestion while using ros2 as the central software.

Object detection (yolo nano and such) runs on device. VIO/SLAM don't. So you still need a powerful host to run it. If you currently only have RPI4, I'd try with that as I am not sure how demanding the current VIO/SLAM is. If it is not enough, you can still purchase a more performant SBC (like RPI5) later.

We have plans to possibly make VIO/SLAM on device for RVC4 which also has more accurate IMUs, but this won't be out for quite some time (year+).

Thanks,
Jaka