Automated Collision Avoidance Of Unmanned Maritime Vehicles In Single Agent Environment Using Policy Gradient Methods And Deep Reinforcement Learning.

MIGRATION OF WHALES

DECLINE IN WHALES

Main causes

Propsed methodology

• Deep reinforcement learning based collision avoidance method for a single agent environment.

• Deep Reinforcement Learning (DRL) algorithms like:

• Proximal Policy Optimization (PPO)

• Soft Actor Critic (SAC)

• Deep Q Networks (DQNs)

• Develop our own DRL algorithm after learning existing available algorithms.

• Simulated Environment with whales and a ship using Unity Engine.

Flowchart

PPO

sample

extending it to real-world