Command line tools for calling path server

[arjo129]

We need a way to specify paths from destinations. A simple python script should suffice for this.

[Briancbn]

Hi @arjo129, I recently created a cli binding packages. This was primarily for my team to debug mqtt and amqp related topics similar to ROS2.
It does similar things as ros2cli but without the dynamic look up of extension across multiple packages.

Since my team and I are also exploring using the same prototype msgs in our existing services. I can maybe give this one a crack?

Let me know your thoughts.

[arjo129]

Hey @Briancbn , I'd be happy to receive contributions from you and your team. There is a prototype of the work we are doing in this branch: https://github.com/arjo129/next_gen_prototype/tree/arjoc/feat/path_server

Its not quite ready for merging here yet as I need to split it into smaller working chunks. Things are moving very quickly there. The destination server for instance is broken and the path planner is currently focused on a simple grid world (I expect to upstream things here slowly over the course of the next week).

I think the best place to take inspiration from is: https://github.com/open-rmf/rmf_demos/tree/main/rmf_demos_tasks

We can start by specifying x,y target coordinates. Id be happy to brainstorm design decitions here.

[Briancbn]

Hey @Briancbn , I'd be happy to receive contributions from you and your team. There is a prototype of the work we are doing in this branch: https://github.com/arjo129/next_gen_prototype/tree/arjoc/feat/path_server

I am really glad things are progressing really fast and thanks for compiling PIBT solvers too. We want to give them a try too.

The destination server for instance is broken and the path planner is currently focused on a simple grid world (I expect to upstream things here slowly over the course of the next week).

No worries on this. Let me take a look first!
We had the same approach in res_mapf (without a destination server) to provide a simple grid world in pybullet.
The only add on is to add a dependency resolution and execution part in the plan_executor.
We wanted to re-write this module in rust in the future, so it would be great if it can done here.
My team and I would be happy to contribute reviews for PRs. Feel free to tag me, @sauk2 or @chianfern

I think the best place to take inspiration from is: https://github.com/open-rmf/rmf_demos/tree/main/rmf_demos_tasks

We can start by specifying x,y target coordinates. Id be happy to brainstorm design decitions here.

Got it!
I was thinking to have something like the following

rmf destination send --option1 <value> --option2 ...

rmf path send --option1 <value> --option2 ...

Which later on we can include more commands and verbs such as

rmf vda5050 assign_order --option1 <value>

Since this needs ROS2, a ament_python package similar to ros2cli, might be best. Unfortunately, it does prevent us from the convenience of uv and some of the modern python utilities.

However, ideally if we can write it in rust, people can simply download an installer similar to uv or pnpm would be the best case scenario, in my opinion.

[arjo129]

I am really glad things are progressing really fast and thanks for compiling PIBT solvers too. We want to give them a try too.

It'll be a while before this stuff touches real robots as the goal of Pass 1 in issue #8 is to focus on getting plumbing up. The PIBT solvers are great for open space, but the improvements for corridors are not yet part of them up stream.

The only add on is to add a dependency resolution and execution part in the plan_executor.

We have a rust version of this already. Currently, it lives out here: arjo129/mapf_post#7
In the future we will be merging this upstream with the mapf library.

[Briancbn]

We have a rust version of this already. Currently, it lives out here: arjo129/mapf_post#7

Perfect! We would be happy to switch to your implementation instead in the future, since we are technically using the same interfaces.

[arjo129]

You can take a look at #13 for starters. I expect bulding it to be a bit challenging but nothing a dockerfile won't be able to solve.