soraxas / sbp-env

Motion planning environment for Sampling-based Planners

Sampling-Based Motion Planners' Testing Environment

Sampling-based motion planners' testing environment (sbp-env) is a full feature framework to quickly test different sampling-based algorithms for motion planning. sbp-env focuses on the flexibility of tinkering with different aspects of the framework, and had divided the main planning components into two categories (i) samplers and (ii) planners.

The focus of motion planning research had been mainly on (i) improving the sampling efficiency (with methods such as heuristic or learned distribution) and (ii) the algorithmic aspect of the planner using different routines to build a connected graph. Therefore, by separating the two components one can quickly swap out different components to test novel ideas.

Have a look at the documentations for more detail information. If you are looking for the previous code for the RRdT* paper it is now archived at soraxas/rrdt.

Quick start with custom arbitrary environments

pip install sbp-env

import sbp_env

from math import exp, sin, cos

for functor in [
    # simple inequality
    lambda x: (x[1] < x[0] + 1) and (x[1] > x[0] - 1),
    # equation adopted from https://au.mathworks.com/help/matlab/ref/peaks.html
    lambda x: 0
    < (
        3 * (1 - x[0]) ** 2.0 * exp(-(x[0] ** 2) - (x[1] + 1) ** 2)
        - 10 * (x[0] / 5 - x[0] ** 3 - x[1] ** 5) * exp(-x[0] ** 2 - x[1] ** 2)
        - 1 / 3 * exp(-((x[0] + 1) ** 2) - x[1] ** 2)
    ),
    lambda x: -0.22 < (cos(x[0]) * sin(x[1])),
    lambda x: 0.05 < (cos(x[0] ** 2 + x[1] ** 2)),
]:
    engine = sbp_env.engine.BlackBoxEngine(
        collision_checking_functor=functor,
        lower_limits=[-5, -5], upper_limits=[5, 5],
        cc_epsilon=0.1,  # collision check resolution
    )
    planning_args = sbp_env.generate_args(
        planner_id="rrt", engine=engine,
        start_pt=[-3, -3], goal_pt=[3, 3],
        display=True, first_solution=True,
    )

    env = sbp_env.env.Env(args=planning_args)
    env.run()
    print(env.get_solution_path(as_array=True))

Installation

Optional

I recommend first create a virtual environment with

# assumes python3 and bash shell
python -m venv sbp_env
source sbp_env/bin/activate

Install dependencies

You can install all the needed packages with pip.

pip install -r requirements.txt

There is also an optional dependency on klampt if you want to use the 3D simulator. Refer to its installation guide for details.

Quick Guide

You can get a detailed help message with

python main.py --help

but the basic syntax is

python main.py <PLANNER> <MAP> [options]

It will open a new window that display a map on it. Every white pixel is assumed to be free, and non-white pixels are obstacles. You will need to use your mouse to select two points on the map, the first will be set as the starting point and the second as the goal point.

Demos

Run maps with different available Planners

This repository contains a framework to performs quick experiments for Sampling-Based Planners (SBPs) that are implemented in Python. The followings are planners that had implemented and experimented in this framework.

Note that the commands shown in the respective demos can be customised with additional options. In fact, the actual command format used for the demonstrations is

python main.py <PLANNER> maps/room1.png start <sx>,<sy> goal <sx>,<sy> -vv

to have a fix set of starting and goal points for consistent visualisation, but we omitted the start/goal options in the following commands for clarity.

RRdT*

python main.py rrdt maps/room1.png -vv

RRT*

python main.py rrt maps/room1.png -vv

Bi-RRT*

python main.py birrt maps/room1.png -vv

Informed RRT*

python main.py informedrrt maps/room1.png -vv

The red ellipse shown is the dynamic sampling area for Informed RRT*

Others

There are also some other planners included in this repository. Some are preliminary planner that inspired RRdT*, some are planners with preliminary ideas, and some are useful for debugging.

Reference to this repository

You can use the following citation if you use this repository for your research

@article{lai2021SbpEnv,
  doi = {10.21105/joss.03782},
  url = {https://doi.org/10.21105/joss.03782},
  year = {2021},
  publisher = {The Open Journal},
  volume = {6},
  number = {66},
  pages = {3782},
  author = {Tin Lai},
  title = {sbp-env: A Python Package for Sampling-based Motion Planner and Samplers},
  journal = {Journal of Open Source Software}
}

sbp-env's Issues

Skip-optimality Problem

Hi
1.I am wonderingt that the parameter (use_rtree)in choose_least_cost_parent() function and rewire() funtion (RRT). Is it no longer necessary because we use numpy's calculation method?
2. When i run the informedrrt algorithm, the ellipse display of the graphic drawing does not appear as shown in the document. How can it be displayed？
I'm sorry to interrupt you from your busy schedule.

Python version compatibility with scipy

Mentioning the requirement of python version >= 3.8 in README would also help users the way it's done over here. Python versions < 3.8 are not compatible with scipy 1.6

(Part of the JOSS review openjournals/joss-reviews#3782)

question on (example) usage

According to the submitted paper, with sbp-env "one can quickly swap out different components to test novel ideas" and "validate ... hypothesis rapidly". However, from the examples in the documentation, it is unclear to me how I can obtain performance metrics on the planners when a run a test.

Is there a way to save such metrics to a file or print them when running planners in sbp-env? If not, this might be a nice feature to implement in a future version. Otherwise, you could consider adding an example to the documentation on how to compare different planners in the same scenario.

(this question is part of the JOSS review openjournals/joss-reviews#3782)

Path recognition issue

I tried some source, destination positions with the following command and there seems some issue in recognition of the path.
python main.py rrt maps/4d.png --engine 4d

Attaching screenshot below:

(Part of the JOSS review openjournals/joss-reviews#3782)

Graph building of prm planner without user information

The graph building method in the prm planner (build_graph() in prmPlanner.py) can take quite some time when a large number of nodes is used. However, the user is not notified that the planner is still processing data. The first time I encountered this, I suspected the software got stuck in an infinite loop because the window was not responding anymore. I think this can be easily fixed by adding a tqdm bar in the build_graph() method (at line 83)

(this suggestion is part of the JOSS review openjournals/joss-reviews#3782)

invalid start and goal point can be specified with command-line interface

When specifying a goal and start point in the commands line, it is possible to specify invalid points. Specifying an invalid start and goal will result in an infinite loop.

For example, running: python main.py rrt maps\room1.png start 10,10 goal 15,15, will result in an infinite loop with the following GUI:

Expected behavior when supplying an invalid option would be an exception.

Suggestion to make installation easier

I was wondering why you have the following remark block in your installation instructions:

I think it would be easier to add those two packages to the file requirements_klampt.txt. That way they'll be installed automatically, it saves the user an extra action. Or is there any reason I'm missing why that shouldn't be done?

Test Instructions

Though it's standard, adding instruction to run tests in the documentation might be helpful for users wanting to contribute.

(Part of the JOSS review openjournals/joss-reviews#3782)

Exception after running the example from the documentation

When I run the example from the quick start page in the documentation, an exception occurs.

The command:
python main.py rrt maps/room1.png

The exception:

Traceback (most recent call last):
  File "main.py", line 287, in <module>
    environment.run()
  File "C:\Users\Olger\PycharmProjects\sbp-env\env.py", line 198, in run
    self.visualiser.terminates_hook()
  File "C:\Users\Olger\PycharmProjects\sbp-env\visualiser.py", line 148, in terminates_hook
    self.env_instance.sampler.visualiser.terminates_hook()
  File "C:\Users\Olger\PycharmProjects\sbp-env\env.py", line 126, in __getattr__
    return object.__getattribute__(self.visualiser, attr)
AttributeError: 'PygameEnvVisualiser' object has no attribute 'sampler'

The exception only occurs after the simulation has finished so it seems like a minor problem. Although I'm not really sure what happens at env.py, line 126, in __getattr__ and why. So, I don't have a proposed fix.