Reinforcement-learning-based navigation for autonomous mobile robots in unknown environments

Mohammed Alhawary
Presentation MSc presentation
Date 2018-08-31
Time 10:00
Location Hal B 2A

Mobile robot navigation in an unknown environment is an important issue in autonomous robotics. Current approaches to solve the navigation problem (roadmap, cell decomposition and potential field) assume complete knowledge about the navigation environment. Navigation in an unknown or a partially unknown environment can be phrased as a reinforcement learning (RL) problem, because it is only possible to discover the optimal navigation plan through trial-and-error interaction with the environment.

The goal of this project is to control a mobile robot to navigate in an unknown environment with obstacles and a slippery floor using reinforcement learning. The main task studied is navigation to a goal location in the shortest time while avoiding obstacles and learning of the slippage model to determine the best way to move on the floor.

The Q-learning algorithm is widely used to discover (sub)optimal trajectories in unknown navigation environments. However, it converges to these trajectories with undesirably-slow rates. The Dyan-Q platform extends the Q-learning with online-constructed models about the environment properties (obstacles, slippage, etc.). A probabilistic model has been created to describe the state transition probabilities that represent the system dynamics. Moreover, a parametric kinematic model of the mobile robot has been learnt experimentally. It has been proven that utilizing these online-built models can significantly accelerate the learning process. 

Posted on Monday, July 9, 2018