# Development of an AI-based control algorithm for 8 legged holonomic robot

This work consists of an AI-based control algorithm for an eight-legged holonomic robot. It also contains an octapod modelled in Gazebo and few digital elevation models to train and test the algorithm.
The entire simulation developed consists of three main parts i.e. 
1. A RNN that takes joint configurations of octapod at any time 't' and predicts the next stable set of joint configurations for time 't+1'
2. A Knowledge transfer networkthat takes the joint dynamics as an input and classifies the type of terrain model that needs to be used in predicting the next stable set of joint configurations for Octapod
3. The ROS wrapper that allows our Neural networks to communicate with our robot model in Gazebo

This SiL simulation is developed and the results are interpreted using python 3.6 and there are several environments that have been used in the code. Out of which notable environments are 
1. ROS Melodic
2. Gazebo 9
3. Tensorflow 1.14
4. Numpy
5. matplotlib

the entire simulation is done on Ubuntu 18.04 operating system. The code required for simulation are divided into catkin packages, each of which contains the necessary modules i.e.
1. spider_control
   b. leg_dynamics.py
   c. my_network.py
   d. control1.py
2. spider_description
3. spider_gazebo