Monash-Connected-Autonomous-Vehicle / ESDA

Software stack for MCAVs annual IGVC entry
0 stars 0 forks source link

Find compute unit alternative #69

Closed AbBaSaMo closed 2 months ago

AbBaSaMo commented 3 months ago

Overview

The hive is good, but the hive is heavy.. and so the electrical and mechanical team require that we find a more lightweight compute unit for running our software on. A few alternatives exist, each with their own pros and cons and preferences from leads so this task aims to get the highest preference working otherwise the next best and so on.

As for the alternatives, refer to the following table

Pref Alternative ProsCons
1MCAV owned Alienware with ubuntu 22.04 install and ROS2 humble No need to spend money
  • One of them is in use by UA and breaks when we install zed-ros2-wrapper dependencies
  • The other one needs upgrading from Ubuntu 18.04 to 22.04 (not a biggie) but might also have other issues iirc
2Buying a Jetson Orin Nano (not nano, ORIN nano) Fulfils our needs costs 1.1k AUD (not too much but still a good chunk of our budget)
3Using our good old (old) Jetson TX2 Already in workshop 2 and we have 2 of them
  • Requires extra effort to get ROS2 humble working on it as it can only install up to ubuntu 20.04
  • Since it's old (and from experience) we might encounter general issues working with it

Relevant resources

Acceptance criteria

Todo

Rikidink commented 3 months ago

Error when booting into the Alienware laptop:

"The (VGA Fan) fan failed to respond correctly" Error was fixed by updating the BIOS on the laptop as per this guide. After that successfully booted into Ubuntu after some waiting on the "booting into insecure mode..." screen. NOTE: This is not really an error, you can press continue on the error screen and it'll still boot.

Rikidink commented 3 months ago

Following this forum post to upgrade from Ubuntu 18.04 -> 22.04:

At the sudo do-release-upgrade When upgrading from 18.04 to 20.04.

Got "Could not calculate the upgrade" error seemingly because of incompatible/broken packages between 18.04 and 20.04.

Error in /var/log/dist-upgrade/main.log : 453 ERROR Dist-upgrade failed: 'E:Error, pkgProblemResolver::Resolve generated breaks, this may be caused by held packages.'

Resolved by removing python-dev: sudo apt remove python-dev sudo apt autoremove

Rikidink commented 2 months ago

Finished update to 22.04. Installing ESDA repo (and dependencies) but installing nvidia and cuda drivers as per ESDA repo readme breaks Ubuntu:

# install driver
sudo apt-get install -y nvidia-driver-550-open
sudo apt-get install -y cuda-drivers-550

Entered recovery mode and deleted drivers to get it in a bootable state again:

sudo apt-get purge nvidia*

(Relevant forum post regarding the issue.)

Installed recommended drivers along with CUDA Toolkit according to the following sources:

AbBaSaMo commented 2 months ago

@Rikidink this is the same issue we encountered with the other Alienware we have that is currently used for the Streetdrone project.

See if any of the suggested fixes below work:

As a summary, some suggest using an older version of the driver, others suggest deleting certain files that somehow fix the issue. Once these and any other issues have been exhausted and we know it definitely won't work, let me know and we can close the issue until we get our hands on the orin nano.

Rikidink commented 2 months ago

The Alienware can now run the stack (NOTE: used the sensor-integration branch of the ESDA repo). Closing issue temporarily until getting the Orin Nano to test on as well.