Document

README.md

@@ -30,3 +30,65 @@ Create a new pull request from the `Pull Requests` tab on this repo, not the for
 
 Request reviews from at least two people. 
 
+---
+
+## Overview
+
+For localisation we decided to implement SLAM ( Simultaneous Localisation and Mapping ). For this we started off with looking into [OpenVSLAM](https://openvslam.readthedocs.io/en/master/ros_package.html).
+
+Some courses for conceptual understanding we found helpful:
+
+1. [Mobile-robotics](http://ais.informatik.uni-freiburg.de/teaching/ss19/robotics/)
+2. [Visual Navigation](https://vision.in.tum.de/teaching/ss2013/visnav2013)
+
+### Dependencies:
+
+1. [Eigen](http://eigen.tuxfamily.org/)
+2. [g2o](https://github.com/RainerKuemmerle/g2o)
+3. [DBoW2](https://github.com/shinsumicco/DBoW2)
+4. [OpenCV](https://opencv.org/)
+5. [Pangolin](https://github.com/stevenlovegrove/Pangolin)
+
+### Issues:
+
+* Mismatch between OpenCV versions : 
+
+Resolved by properly setting up paths and making changes in CMakefiles to enable it to run on required version.
+
+* Video streaming from gazebo :
+
+To make it work with live video feed from gazebo we had to redirect some topics and attach a camera plugin on our model which was kept in the simulation environment with obstalces and walls around it.
+
+Testing OpenVSLAM with a drone moving through a virtual alley:
+[Click here for video](https://drive.google.com/file/d/1Y_Hu-au4z2JEkjzrlUlrZw7EljRF2b_A/view)
+
+
+To test this on a quadruped we tried to stimulate the model we had using [towr](http://docs.ros.org/lunar/api/towr/html/index.html) and [towr_ros](https://github.com/ethz-adrl/towr)
+
+We had to modify the code to make our own node to publish the final pose for the quadruped to reach.
+
+![Image](https://github.com/isro01/PETcat_localization/blob/document/img/towr_code.png)
+
+[Demo video can be found here](https://drive.google.com/open?id=1MN2vm95H1Dhfs8n3qlgd3xLlkyX3Nfzf
+)
+
+* Improper friction values for joints:
+
+We didnt have proper friction values for the robot joints and hence the random motion in the demo video.
+
+Moved on to a [repository](https://github.com/chvmp/champ) to implement the model with SLAM. 
+
+---
+
+## Benchmarking
+
+Had to benchmark against [OpenVSLAM](https://openvslam.readthedocs.io/en/master/installation.html) and [GMapping](http://wiki.ros.org/gmapping) the following factors:
+
+1. Feature detection
+2. Detection rate
+3. CPU usage and GPU information
+4. Odometry drift
+
+The work done for the same can be found [here](https://github.com/m2kulkarni/PETcat_Benchmark)
+
+---