Small Outdoor Vehicle Navigation: Comparison of ROS2 Nav2 and Autoware

[Ross Lunan]

In a recent forum I participated , there was a discussion of the merits of current ROS 2 Navigation2 and Autoware, that runs on top of Navigation2 that was developed specifically for autonomous vehicle applications. The target vehicle application was a small, i.e.sidewalk dimensions, autonomous outdoor snowplow. Has anyone done any research or comparison of anything pertaining to this, or any relevant papers or dialog that's been posted? 

Autoware

Ross

[Chris Albertson]

I think I’m right at our level on this… I’ve only done basic research into software.   I think it is clear that if it needs to be Open Source, Autoware is the best you will find for this specific task, driving on roads and paths outdoors.     Tesla and Google can do better but they are not sharing their source code.

I think the biggest common beginner mistake is to look at a big package like Autoware and say “This is too complex, I can’t understand it, so I will re-implement it all myself and improve it at the same time.”    OK, it would take 500 man-years of engineering effort and you would not live long enough.  But in only a month or two you could come up to speed with Autoware basic use.   My platform was a ride-on kiddie car with 80Kg payload.  Enough for a big computer. Then I got into walking robots.

Basic ROS nav is very primitive and uses SLAM which means running in a very controlled environment and not paying much attention to user road users.    Outdoors, THE problem is predicting the future actions of others.   

--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/98b56ab0-01ec-44e8-a985-b0689546ad16n%40googlegroups.com.

[Brian Higgins]

This sounds interesting I have a 4 wheel mobility scooter and Wednesday the custom enclosure will be here.  This is phase 1.  Golf cart is phase 2. Now I need sensors and this software 

Brian Higgins 

VA Researcher for blind mobility 

Sent from my iPhone 13 Pro Max

On Feb 18, 2024, at 8:34 PM, Chris Albertson <alberts...@gmail.com> wrote:

I think I’m right at our level on this… I’ve only done basic research into software.   I think it is clear that if it needs to be Open Source, Autoware is the best you will find for this specific task, driving on roads and paths outdoors.     Tesla and Google can do better but they are not sharing their source code.

To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/90859CC0-9CF3-4B77-AC96-D702109D0A85%40gmail.com.

[Sergei Grichine]

Just to confuse things even more ;-)  - the outdoors are well explored by different autopilots, including the PX4 frome Dronecode project (https://dronecode.org/).

PX4 has been interoperating with ROS for a long time. There's currently a push to develop better support for Rovers of all kinds. I am using PX4 for my full size 23 Hp lawn mower (no ROS needed).

Yes, lawn mowing and drone flying are in some ways less demanding for robotic skills than driving on a sidewalk. But autopilots are VERY good at IMU/GPS sensing and controlling actuators in true real time. And listening to ROS guidance.

So, there are many vehicles in existence now where the "brain" is ROS-based, while the "body" is managed by PX4.

To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/8F3B53D5-BF58-4D87-9EDE-F2FE2D678BAD%40gmail.com.

--

Best Regards,

-- Sergei

[Chris Albertson]

On Feb 19, 2024, at 4:25 AM, Brian Higgins <blind...@gmail.com> wrote:

This sounds interesting I have a 4 wheel mobility scooter and Wednesday the custom enclosure will be here.  This is phase 1.  Golf cart is phase 2. Now I need sensors and this software 

You absolutely MUST set up a simulated environment FIRST.  One does not install Autoware in a car or even a golf cart and then test it on the street, or even in a parking lot.    It is not at all like you just click “Install” and it runs.  You will be into the source code deep for a year.   You will want to set up sensors in a lab bench environment first

[Ross Lunan]

Thanks everyone & I'm certain the group where this discussion took place will value the comments. The organization is "Ottawa Autonomous Vehicle Group" (OAVG), located in the Canadian City of Ottawa. OAVG They are preparing for an annual competition that takes place in Minneapolis  next  January 17-18, 2025. AutoSnowPlow . BRRRRR