Project dedicated for sensiitivity experiments. Robust trajectory generation based on the PX4 controller for Acanthis drone below.

Getting Started

• Installation
  

Use the following sudo commands

sudo apt install clang
sudo apt-get install libomp-dev

• Make Directory and clone Repo
  

mkdir sens_exp && cd sens_exp
git clone https://gitlab.inria.fr/asrour/sensitivity_exp.git

• Configure a virtual enviroment OR skip this if you prefer to install global on machine
  

In the same directory that you cloned the project, you can create and source the virtual environment by:

cd sensitivity_exp
python3 -m venv env
source env/bin/activate

• Install dependencies in virtual environment
  

pip install -r requiremnts.txt

• Build the framework as a package
  

The -e is essential as this would make it possible to modify the package without building it upon each modification-

pip install -e .

• sens package
  You can view the package by

pip list

• Run the Following script that generates the ODE problem
  This script should generate the model with all sensitivity related stuff

python3 -m sens.gen.models.jetson_pd

Framework Overview

After installing the sensitivity framework as the sens package, you are ready now to start discovering the framework. Mainly the sens package consists of 5 folders "directories" which are:

• cnst
  
  • Contains a python file constant.py which includes a class with all constants to be used in the framework
• gen
  

  • Contains Folder lib for symbolic generation (No need to edit scripts in this directory)

    • a script base_model.py defines an abstract class of the model and functions declarations
    • a script sym_gen.py which contains a class called Jitparam based on Jitcode for generating symbolic model .so file

  • Contains a Folder models where you have to define your ODE problem, model, controller and sensitivity

    • a script jetson_pd.py is the script containing all these definitions and responsible of generating the model depending on jtiparam class in sym_gen.py
• opt
  
  • a script optimize.py where different functions implemented for different optimization problems depending on Nonlinear optimizar from nlopt package in python
• script
  
  • This folder contains several scripts that you can run as examples and they are documented very well
• utils
  
  • Functions.py defines some symbolic functions used in jetson_pd.py and also other functions like evaluation of the tubes on states.
  • trajectory.py contains a class named PiecewiseSplineTrajectory where we define the trajectory and do all the planning and pre optimization depending on this class

Learn by examples

Inside sens.script you would find several examples that you need to run to explore the package These examples/turorials are for the sake of using the framework:

• How to optimize trajectories?