# Learning Control for Air Hockey Striking using Deep Reinforcement Learning

## TL;DR

This paper develops a deep reinforcement learning approach with specific improvements to enable a robot to learn precise puck striking in air hockey, demonstrating successful simulation results and enhanced algorithm stability.

## Contribution

It introduces modifications to deep Q-learning, including prior knowledge integration and experience replay adjustments, to improve learning stability and feasibility for robotic puck striking.

## Key findings

- Successful simulation of aimed puck striking
- Enhanced stability of the deep Q-learning algorithm
- Effective integration of prior knowledge into learning process

## Abstract

We consider the task of learning control policies for a robotic mechanism striking a puck in an air hockey game. The control signal is a direct command to the robot's motors. We employ a model free deep reinforcement learning framework to learn the motoric skills of striking the puck accurately in order to score. We propose certain improvements to the standard learning scheme which make the deep Q-learning algorithm feasible when it might otherwise fail. Our improvements include integrating prior knowledge into the learning scheme, and accounting for the changing distribution of samples in the experience replay buffer. Finally we present our simulation results for aimed striking which demonstrate the successful learning of this task, and the improvement in algorithm stability due to the proposed modifications.