ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A

Autobotware-AI new All in one Autonomous UGV Platform

Ladies and Gentlemen,
Glad to announce that Autobotware platform(Roboware previously) is now available at:

Autobotware is an all-in-one open-source software for autonomous UGVs and industrial robots. the target industries that Autobotware is working on are farming industry, mining industry, warehouses industry, and construction industry.

Autobotware Featues

  • Modular software components can work with different sensors(benwake/velodyne)
  • 3D semantic segmentaiton object detection.
  • Freespace estimation.
  • Integration of Autobotware with ROS packages such as gmapping, move base, and amcl.
  • Free space fusion for cocoon Setup.
  • 3D point cloud assembly.

Currently under development features:

  • a new state of the art 3D slam.

There are much more features planned in the future :slight_smile:

We believe that the world is way bigger than a single team would win alone :), lets all win together :).

please try it and provide feedback :slight_smile:
please try it and provide feedback :slight_smile:

Contact us and let’s build the future together:


I read this as “” Is the name intentional? It is really confusing.


I didn’t realize it WASN’T a post about until your post.


That is true we intentionally named it Autobotware as the idea of this software framework was inspired from Autoware, and we are looking forward to have the same impact of Autoware left in the autonomous driving industry but in Robotics and UGVs fields.

In Autobotware you can the same algorithms in Autoware, but it is more optimized, better performance, and more robust for example:

lidar clustering and ground segmentation in Autoware(core_perception/lidar_euclidean_cluster_detect at master · Autoware-AI/core_perception · GitHub)
Spherical Grid Map Clustering in Autobotware(Autobotware/sgm_lidar_clustering at main · Robotion-AI/Autobotware · GitHub)

Both of the two packages are providing the same features but our SGM clustering is more optimized, more robust, and better performance.

same for Lidar localization (currently under development in Autobotware).

Autoware requires high preformance hardware targets to process its algorithms, Autobotware can process LiDAR point clouds on Raspberry pi.

All in all, we were inspired by them, we hope we leave their impact in the Robotics and UGV industry.

1 Like

Congratulations on releasing the code!

However, I have to say that is very unfortunate name. It’s gonna confuse your potential users and Autoware’s. Given that it’s still in early stages, have you considered renaming it? Two people in this thread thought the post was about Autoware. I myself, work on Autoware, and my first reaction was of surprise because I thought the post was about Autoware too.


Hi Khaled,

Have you considered talking to the Autoware team about the name and how you could better collaborate? It seems a bit impertinent to create a derivative project name and then claim that your project is “more optimized, better performance, and more robust”, especially without proving metrics to back up your claim. If your approach is better why wouldn’t you upstream it to Autoware so everyone could benefit? I think I speak for most of the community when I say that we should try to foster collaboration instead of competition. In any case, I would be more than happy to make an introduction to the Autoware team if you think it would help.




Thanks for your feedback, I took it to our team and there is an on going discussion regarding renaming it (Y), let’s see where this will Go :slight_smile:

Hi Kat,
Thanks for your feedback as well, but I may reply to you in points:

  • Auto(BOT)ware is a software platform provided by Robotion, a new startup seeking to have a place in the Robotics and autonomous UGV industry, and releaseing an open source software platform would be the best advertisement for the company, and to promote our competence and capabilities.
  • Unfortunatly we hadn’t talked to Autoware team, but we are open to cooperate with anyone, and it would be honor to work with the Autoware team, if you can help in that please let me know (, they started the opensource autonomous driving thing :slight_smile: , and I believe the community owe them that :slight_smile: .
  • Regarding KPIs you are correct I should have presented my claim with KPIs I take responsibility of this mistake. but atleast what is obvious is no other platform can process LiDAR point cloud on raspberry pi, but still you are right, quantitative KPIs are the language to speake with not claims. in the mean while we are not looking for a competition, more than a cooperation and completion between different entities and different groups.
  • We are presenting a novel software stack blocks are arranged in a different way from Autoware, and the approach of solving the problems are different, and our assumptions are different, as a result the performance and accuracy are different.
  • I might disagree with you diversity always produce better quality products, for example Autoware is not the only opensource SW for autonomous driving, there is appolo and other opensource SW platforms, diversity is always better, and we are here to be part of that :slight_smile: .

at the end of the day thanks for your feedback, we are open to cooperate and work together, as I mentioned before the world is so big for a single team to win, let’s all win together :slight_smile: .

1 Like

I’d be careful with your claims of performance and novelty. I took a glance and while I’m not commenting on the merits of the approaches, the vast majority of the code is using common PCL and openCV APIs meant for exactly (and commonly used with) the way you’re using it. Just a friendly note to not oversell and not list proper tracking/recall metrics to support your claims.

there are three types of KPIs performance KPIs, qualitative KPIs and quantitative KPIs,

  • Quantitative KPIs are the only KPIs that are missing, and I will give it the highest periority.
  • Performance KPIs is reported on a raspberry pi, and I think as far as I know no one had processed LiDAR point cloud on a raspberry pi before.
  • Qualitative(visualization) KPIs can be tested just by cloning it and checking how good it looks.

Quantitative KPIs was not presented ((at this stage)) as this SGM is one module of our sw stack, where we concentrated on adding features more than concentrated on comparing it with current platforms.

in all the ways, I am calling you to test it and to provide feedback on it :slight_smile: , we are open to learn if we are wrong :slight_smile:

The Autoware team has a dedicated Slack and a Discourse section here on the ROS forum, all the info on how to reach out to them is on the website

That’s not how it works, the burden of proof is on who makes the claims. If you claim your project to have better performancer than another project, you have to back those claims with benchmarks, not ask other people to prove you you’re wrong.


I agree with you, we will give the benchmarking task the highest priority now, beside we already provide the performance and qualitative KPIs, but you are correct :slight_smile:

Laddies and Gentlemen,
All feedback is much appreciated and we take all of it really seriously, but please try to know how to give a feedback.
I have received some negative feedbacks that the least that I can say it has lack of knowledge and have no idea of what we are providing in our software framework, that’s why I called you to test it before providing feedback. And for real I have no idea how people will act when we release the benchmark or when we will release our 3D slam algorithm, I think we will witness some mad waves “how dare you publish something so nice to be true!!”
In all the ways best of luck for you all and hope all of you all the success.
Khalid elmadawi
Sr Software Engineer,
Environment perception engineer.

Is there any community standards here? :slight_smile:

As a Dutch guy, I’m all for being direct and to the point, but could we please keep things a bit more civil?

I agree that the onus is on @Khaled_Elmadawi et al. to provide some proof for their performance claims, and the name is a bit ambiguous (which is not a very nice thing to do), but that’s not really a justification for being impolite like this.


@Khaled_Elmadawi: if you provide benchmarks which are repeatable and fair, and which substantiate your claims about performance, that should help convince potential users.

At the moment, all we have are your comments here and on your website.

I can fully appreciate the enthusiasm you have for the results of your work, but if you intentionally use a name similar to an existing project/product (piggy-backing on the reputation of an established software project is not a form of attribution nor acknowledging where you got your inspiration from), then claim repeatedly that what you have is very much more performant without backing up those claims with evidence it seems it wouldn’t be so unexpected to see some backlash.

1 Like

thank you so much for your feedback, I have finished the Test cases and made some conclusions: here is the primary image of the virtual test rig, where I provide same input and make measurements on the output of each algorithm:

what is taking time from me is:

  • I am on a vacation.
  • defining the KPIs.
  • writing the conclusion.
  • editing the images to project the conclusions on it.
    everything should be finished either today or tomorrow.

Thanks again for being professional, and best of luck in everything :slight_smile:

I hope this time you’ll also publish a paper where you’ll point what’s YOUR contribution to YOUR 3D slam. New 3D slam approaches are always welcomed.
That reminds me of a linkedin post I’ve seen recently from an indian startup where they wrote “we implemented teb_local_planner”. No you don’t, you just typed “ros2 launch teb_local_planner”

1 Like

what you are saying is optimal way and best way to present your work, but from around 8 years+ experience 6 of them in autonomous driving industry technology/state of the art changes with significant rates, so make your move in the way that fulfill your priority and my priority now is to reach people and show them our competence so we can gain their trust in solving problems, we might be wrong, we might be right, lets see how things will work :slight_smile:

Healthy approach. Good luck with benchmarks and report.


pon., 17.05.2021, 10:39 użytkownik Khaled Elmadawi via ROS Discourse <> napisał:

1 Like