What is the Best SBC/mobile computing platform for ROS in 2020

asked 2020-11-19 22:26:53 -0600

SamFlynn gravatar image

HI

Hope you are doing well.

I am currently trying to make a scale self-driving rally car. TBH I don't have much idea about what I am doing but I made a some what functional self-driving car(In CARLA simulator), by doing courses from udacity, Coursera and udemy and I have no idea about computational complexity and stuff (I am not a CS/EE student). I have a 1/8 scale rc car as my computer carrying platform.

While doing testing ,I almost broke my laptop, on which ros was running and I am not willing to take that risk again.

I am using an Intel d435, an T265, A kinect v1, A Rplidar A1, A MPU 9250 IMU, A Neo 6m gps, 5 rotary encoders and 3 arudino Megas.

(Must requirement) - I plan to use Gslam, Rtabmap, Particle filter, EKF and some other SLAM related custom packages which involves some image processing, path planning and control.

(Will try to do in future) - I intend to run a very complicated/detailed model for state estimation and control(MPC in real time maybe). I also plan to run a high frequency custom local planner(>100 Hz) and a global planner(~50 Hz).

What sort of computing platform might be suitable for me.

I was thinking about buying a 16 GB jetson Xavier but Since carrying a 4-5 kg payload as a computing platform is not a problem for me, I was wondering should I buy a laptop around that price range. My budget is around 700 $(But higher performance to cost ratio is also a high priority after performance).

So, Is there any kind soul out there to give me some advice on what to do? Should GPU be a priority? I know my question is very vague, please feel free ask question to clarify.

TIA for taking the time to read my huge question.

edit retag flag offensive close merge delete