erc-bpgc/gennav

Python Package for Robot Navigation

GenNav

Status: Under Development

A python package for robot navigation algorithms.

Installation

The package is currently under development so we suggest installing from source.

From Source (Recommended)

git clone https://github.com/ERC-BPGC/gennav.git
cd gennav
python -m pip install .

Using pip

pip install gennav

Usage

To plan a path using the Rapidly-exploring random tree algorithm in a polygon based environment representation.

from gennav.planners import RRT
from gennav.envs import PolygonEnv
from gennav.utils import RobotState
from gennav.utils.geometry import Point
from gennav.utils.samplers import UniformRectSampler

obstacles = []  # obstacles can be added here
env = PolygonEnv()  # polygon environment to keep track of obstacles
env.update(obstacles)  # updating environment with obstacles

sampler = UniformRectSampler(-5, 15, -5, 15)  # for sampling random states
planner = RRT(sampler=sampler, expand_dis=0.1)  # creating the planner

start = RobotState(position=Point(1, 1))  # starting state
goal = RobotState(position=Point(10, 10))  # goal state

path, info_dict = planner.plan(start, goal, env)  # planning path through obstacles

Note that the environment have been left blank empty here, they should be updated as per use case.

For more details you can refer to our documentation.

ROS Integration

If you wish to use gennav in a ROS based stack, check out gennav_ros.

Contributions

Contributions are always welcome. We reccomend you check out contribution guidelines and view the docs beforehand.

gennav's Issues

Add nearest_obstacle_distance method to scan environment

The nearest_obstacle_distance method returns the distance to the nearest obstacle. Check the base.py file in envs to know about the general implementation of the method. As an example you can also take a look at the method implemented in the PolygonEnv.

Test PRM on trotbot

Good issue for those who want to get famililar with ROS also

Improve docs

1. Organise docs into logical sections
2. Add intro text
3. Add examples
4. Fix minor formatting issues in docstrings (good first issue)

We are using this style guide

Add Hybrid RRT

Resources:
https://www.researchgate.net/publication/338721855_Hybrid_RRT_A_Semi-dual-tree_RRT-based_Motion_Planner

Add PRM* Path Planner

Refactor PRM to follow planner base class

Refactor RRT Based planners to use single base and reduce redundancy

Adding odometry simulator

A class for an odometry simulator would be very useful for testing controllers. SInce our courrent controllers output a velocity, the simulator would take the velocity command at each step and numerically integrate (basically add) it to get distance travelled.

Now to simulate real life conditions, instead of giving back the position derived directly by integrating (the true position) the class will add gaussian noise (using np.random.randn) to it and then return it back to controller. The simulator would be added in /utils

Basic structure would be

class Simulator
    def __init__(controller)
        ...
    def run(parameters)
        ...

Add bugflood implementation

Add PRM path planner

Set up docs

Add RRT*-Smart Path Planning

Fix bug in RRT

There is a bug in RRT which is causing tests to fail randomly. Sometimes the planned path is correct while sometimes it isnt.

Add pybind11 based geometry module

add differential pid controller

Implement a method to convert Laser Scan data into an Occupancy grid

Most of the robots are equipped with a laser scanner. A sample implementation of algorithms using a laser scanner data can be found in the scan_env.py in envs. Implement a function that takes an input of such data and updates an occupancy grid.

Path Post Processing

Implement the following path post processing algorithms

1. Interpolation based path smoothening
2. Bezeir Curves
3. B-splines
4. Cubic-Splines

Add **kwargs to to planner __init__s

Add potential field planner

Add RT-RRT* Path Planning

RT-RRT*.pdf
This is the Real Time RRT* research paper which contains the algorithm.

Updated RRT and PRM to use new node class

Add RRG path planner

Add example for rrg

In /examples/rrt.py add a short example of how PRM can be used by an external user

Add examples for prm

In /examples/prm.py add a short example of how PRM can be used by an external user

Add contribution guidelines

Make samplers a callable class

Add wavefront planner

https://www.youtube.com/watch?v=Ls8EBoG_SEQ&t=1694s

This lecture is a good starting point to understand apf planning algorithms. A unique requirement in wavefront planner is that you need to represent the planning area in the form of a grid world. Asking the user for the coordinates of the points enclosing the grid environment(for eg. the coordinates of end points in case of a polygonal grid world) would be the preferable way to go about this. The dimension of each cell in the grid depends upon the intended resolution

Add utils/common.py

Add a file utils/common.py for storing common datatypes and functions

Add benchmarks in the documentation

Implement general compute_vel method for OmniPID controller

Add methods for path shortcutting

Implement the algorithms given below in the resource.
Link: http://www.osrobotics.org/osr/planning/post_processing.html

Add bi-directional RRT

You can refer to these resources:
http://planning.cs.uiuc.edu/node237.html
http://people.csail.mit.edu/aperez/obirrt/
https://www.cs.cmu.edu/~motionplanning/lecture/lec20.pdf

https://www.cs.cmu.edu/afs/cs/academic/class/15494-s14/readings/kuffner_icra2000.pdf

Adding replan to all the planners

Add polygon list env

Add LBT-RRT Path Planning

LBT-RRT.pdf
This is the Lower Bound Tree- RRT research paper which contains the algorithm.

Minor error in CONTRIBUTIONS.md file

There is a "python" spell error in the bash code on line 31.

Streamline RRT and RRT*

Add omni-PID controller

Add Occupancy Grid Environment

Add occupancy gird environment

Add plan function in PRM Path Planner

Fix linting issues

Create djikstra function

This will call astart with 0 heuristic internally.

Add tests for envs

Refactor RRT to follow planner base class