Atari Seaquest Environment

Overview

Seaquest is an underwater shooter in which the player controls a submarine. Enemies include sharks and submarines. The player must ward off the enemies while trying to rescue divers swimming through the water. The sub can hold up to eight divers at a time. Each time the player resurfaces, all rescued divers are emptied out in exchange for points. To add to the challenge, the submarine has a limited amount of oxygen. The player must surface often in order to replenish the oxygen, but if the player resurfaces without any rescued divers, they will lose a life. If the player resurfaces with the maximum amount of divers, they will gain bonus points for the sub’s remaining oxygen. Each time the player surfaces, the game’s difficulty increases; enemies increase in number and speed. Eventually an enemy sub begins patrolling the surface, leaving the player without a safe haven.

Description from Wikipedia

Performances of RL Agents

We list various reinforcement learning algorithms that were tested in this environment. These results are from RL Database. If this page was helpful, please consider giving a star!

Star

Human Starts

Result Algorithm Source
40425.8 Human Massively Parallel Methods for Deep Reinforcement Learning
39096.7 Prioritized DDQN (prop, tuned) Prioritized Experience Replay
37361.6 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
25463.7 Prioritized DDQN (rank, tuned) Prioritized Experience Replay
19176.0 Rainbow Rainbow: Combining Improvements in Deep Reinforcement Learning
14498.0 DDQN (tuned) Deep Reinforcement Learning with Double Q-learning
11848.8 Prioritized DQN (rank) Prioritized Experience Replay
10145.85 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
4199.4 DDQN Deep Reinforcement Learning with Double Q-learning
3275.4 Distributional DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
2793.9 DQN Massively Parallel Methods for Deep Reinforcement Learning
2355.4 A3C FF Asynchronous Methods for Deep Reinforcement Learning
2300.2 A3C FF 1 day Asynchronous Methods for Deep Reinforcement Learning
1431.2 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
1326.1 A3C LSTM Asynchronous Methods for Deep Reinforcement Learning
215.5 Random Massively Parallel Methods for Deep Reinforcement Learning

No-op Starts

Result Algorithm Source
266434 C51 A Distributional Perspective on Reinforcement Learning
50254.2 DDQN A Distributional Perspective on Reinforcement Learning
50254.2 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
42054.7 Human Dueling Network Architectures for Deep Reinforcement Learning
30140 IQN Implicit Quantile Networks for Distributional Reinforcement Learning
20994.1 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
20181.8 Human Human-level control through deep reinforcement learning
19595 DuDQN Noisy Networks for Exploration
16754 NoisyNet DuDQN Noisy Networks for Exploration
15898.9 Rainbow Rainbow: Combining Improvements in Deep Reinforcement Learning
13169.06 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
8425.8 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
8268 QR-DQN-1 Distributional Reinforcement Learning with Quantile Regression
7995.0 DDQN Deep Reinforcement Learning with Double Q-learning
5860.6 DQN A Distributional Perspective on Reinforcement Learning
5286 DQN Human-level control through deep reinforcement learning
5071.6 Reactor ND The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
4754.4 Distributional DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
4163 DQN Noisy Networks for Exploration
2680 QR-DQN-0 Distributional Reinforcement Learning with Quantile Regression
2282 NoisyNet DQN Noisy Networks for Exploration
1776.0 ACKTR Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
1754.0 A2C Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
1753.2 IMPALA (deep) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
1744 A3C Noisy Networks for Exploration
1716.9 IMPALA (shallow) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
943 NoisyNet A3C Noisy Networks for Exploration
931.6 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
844.6 IMPALA (deep, multitask) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
810.4 TRPO Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
675.5 Contingency Human-level control through deep reinforcement learning
664.8 Linear Human-level control through deep reinforcement learning
68.4 Random Human-level control through deep reinforcement learning

Normal Starts

Result Algorithm Source
28010 Human Playing Atari with Deep Reinforcement Learning
2995 UCC-I Trust Region Policy Optimization
1948.571 DQN RL Baselines Zoo b76641e
1908.6 TRPO - single path Trust Region Policy Optimization
1782.687 PPO RL Baselines Zoo b76641e
1740 DQN2013 Best Playing Atari with Deep Reinforcement Learning
1739.5 ACER Proximal Policy Optimization Algorithm
1714.3 A2C Proximal Policy Optimization Algorithm
1705 DQN2013 Playing Atari with Deep Reinforcement Learning
1672.239 ACKTR RL Baselines Zoo b76641e
1383.38 PPO (MPI) OpenAI Baselines cbd21ef
1218.87 PPO OpenAI Baselines cbd21ef
1204.5 PPO Proximal Policy Optimization Algorithm
1201.16 ACKTR OpenAI Baselines cbd21ef
1164.08 DQN OpenAI Baselines cbd21ef
1150.66 A2C OpenAI Baselines cbd21ef
1065.98 ACER OpenAI Baselines cbd21ef
920 HNeat Best Playing Atari with Deep Reinforcement Learning
872.121 ACER RL Baselines Zoo b76641e
800 HNeat Pixel Playing Atari with Deep Reinforcement Learning
788.4 TRPO - vine Trust Region Policy Optimization
746.42 A2C RL Baselines Zoo b76641e
723 Contingency Playing Atari with Deep Reinforcement Learning
710.07 TRPO (MPI) OpenAI Baselines cbd21ef
665 Sarsa Playing Atari with Deep Reinforcement Learning
110 Random Playing Atari with Deep Reinforcement Learning