Usability (at least for v2 branch) about fs-private HOT 4 CLOSED

After going through some "let's see if I can do this myself", I found out that:

1. I needed to install antlr4 python3 support (be careful because if you install the python2 support first, then you are dead!:))
2. I needed to use the flag --hybrid from the command line

Point 1) needs to be written in the Usage
Point 2) needs to be written in the Installation instruction

Anyway, I am still not able to see a plan (argh:D). That's what I got when I run the planner:
`
...
[INFO][ 0.00270] Number of state variables: 6
[INFO][ 0.00270] Number of problem atoms: 6
[INFO][ 0.00271] Number of action schemata: 12
[INFO][ 0.00271] Number of (perhaps partially) ground actions: 0
[INFO][ 0.00271] Number of goal atoms: 8
[INFO][ 0.00272] Number of state constraint atoms: 4
[INFO][ 0.00276] Deriving control to search engine...
[INFO][ 0.00355] Starting search. Results written to .
Peak memory: 59976 KB
caught signal 11 -- exiting
Segmentation fault (core dumped)

`

I used this:
./run.py -i examples/hybrid/continuous/walkbot/instance_001.pddl --driver sbfws --options "bfws.rs=sim" --hybrid

Which isn't very helpful. I am going to try another couple of things!

from fs-private.

Last update:
I tried with --driver bfs and no option and I got that the problem is unsolvable :( I know this problem requires finer discretisationd, yet I don't believe that this is the issue, at least in this case. If it was, won't be good to provide a feedback for the user say: look since these problems are all undecidable, can you please try with a smaller delta:)?

Enrico

from fs-private.

Hello @enricos83 ,

fortunately, right now there are no random people coming to try the planner :)

When that happens, the Readme.md file should point people interested in setting up FS for hybrid planning to these docs I wrote. There you will find about those things you figured out the hard way... Let me apologise because I thought I had pointed you towards that file, which I wrote for you in the first place.

Regarding the blind search driver bfs being used on these problems... I am surprised that it actually works.

Recommending the user to try with a smaller delta, or other parameters such as more precise integrators is a good idea. I am not entirely sure thought that we want to do this from the code, but rather as part of the documentation. I think it is easier to edit the latter than to update the code to produce meaningful feedback.

from fs-private.

Thanks Miquel:)
The pointer in the Readme would be great. I thought the doc folder only contained documentation for the code.
I went for some of the things which are described in the doc, yet I couldn't make Soplex work, which seems to be replaced by some patched version. Let me try.
Thanks again

from fs-private.