Rubén Lucas

Rubén Lucas

Site for my documentation

Projects

Note that all the projects below have been developed and executed in linux 20.04

Gazebo F1 Follow line all algorithms comparison

Algorithms comparison

Gazebo F1 Follow line ddpg

Algorithms comparison

Gazebo F1 Follow line dqn comparisson

Algorithms comparison

Gazebo F1 Follow line dqn refinement

Algorithms comparison

Gazebo F1 Follow line dqn & qlearning comparison

Algorithms comparison

Adding deviation from the center and improved graphs

Algorithms comparison

Adding continuous actions and clock control to cartpole problem

Algorithms comparison

Comparing PPO, DQN and QLearn algorithms in cartpole problem resolution

Algorithms comparison

Comparing DQN and QLearn algorithms in cartpole problem resolution

Algorithms comparison

Adding additive random noise and improved monitorization graphs

Experiments refinement

Trainings and implementation experiments to make cartpole solution more solid

Experiments

Included random initial positions and random perturbations to cartpole problem

Cartpole as a guinea pig OpenAIGym simulator related problem

DQN cartpole problem included

Cartpole optimal solution reached

Qlearning cartpole problem included

Cartpole as a guinea pig OpenAIGym simulator related problem

created an inference module for RL-Studio

Implemented inference module in RL-Studio 1.1.0

Migration to RLStudio of basic robot mesh problem

The robot mesh exercise was migrated to RL-Studio

Migration to RLStudio of mountain car problem proposed by openAI-gym

The mountain car exercise was migrated to RL-Studio

MountainCar openAI Gym exercise revisited after reading sutton section 1

Still focusing on minimum computation possible, this exercise has been revisited to apply some lessons learned in sutton RL book

MountainCar customized environment making use of OpenAI Gym libraries

Feel free to review physics and remove libraries dependencies

MountainCar openAI Gym exercise solved using q learning

Note that q learning is not the best option to solve this exercise

Robot following path to goal with q learning and sarsa

Using q learning and sarsa to learn a path from origin to destination through a mesh board