Atari Demon Attack Environment

Overview

Marooned on the ice planet Krybor, the player uses a laser cannon to destroy legions of demons that attack from above. Visually, the demons appear in waves similar to other space-themed shooters, but individually combine from the sides of the screen to the area above the player’s cannon.

Each wave introduces new weapons with which the demons attack, such as long streaming lasers and laser clusters. Starting in Wave 5, demons also divide into two smaller, bird-like creatures that eventually attempt descent onto the player’s cannon. Starting in Wave 9, the demons’ shots follow directly beneath the monsters, making it difficult for the player to slip underneath to get in a direct shot.

Description from Wikpedia

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
115201.9 A3C LSTM Asynchronous Methods for Deep Reinforcement Learning
113308.4 A3C FF Asynchronous Methods for Deep Reinforcement Learning
109856.6 Distributional DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
109670.7 Rainbow Rainbow: Combining Improvements in Deep Reinforcement Learning
84997.5 A3C FF 1 day Asynchronous Methods for Deep Reinforcement Learning
73371.3 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
73185.8 Prioritized DDQN (prop, tuned) Prioritized Experience Replay
69803.4 DDQN (tuned) Deep Reinforcement Learning with Double Q-learning
61277.5 Prioritized DDQN (rank, tuned) Prioritized Experience Replay
56322.8 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
19478.8 Prioritized DQN (rank) Prioritized Experience Replay
14880.13 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
13943.5 DDQN Deep Reinforcement Learning with Double Q-learning
12835.2 DQN Massively Parallel Methods for Deep Reinforcement Learning
3442.8 Human Massively Parallel Methods for Deep Reinforcement Learning
208.3 Random Massively Parallel Methods for Deep Reinforcement Learning

No-op Starts

Result Algorithm Source
274176.7 ACKTR Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
132826.98 IMPALA (deep) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
130955 C51 A Distributional Perspective on Reinforcement Learning
128580 IQN Implicit Quantile Networks for Distributional Reinforcement Learning
122782.5 Reactor ND The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
121551 QR-DQN-1 Distributional Reinforcement Learning with Quantile Regression
117577 QR-DQN-0 Distributional Reinforcement Learning with Quantile Regression
115154.0 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
111185.2 Rainbow Rainbow: Combining Improvements in Deep Reinforcement Learning
110626.5 Distributional DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
107264.73 IMPALA (shallow) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
100189 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
72878.6 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
69311 NoisyNet DuDQN Noisy Networks for Exploration
61033 DuDQN Noisy Networks for Exploration
60813.3 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
58044.2 DDQN A Distributional Perspective on Reinforcement Learning
37880 NoisyNet A3C Noisy Networks for Exploration
37085 A3C Noisy Networks for Exploration
36150 NoisyNet DQN Noisy Networks for Exploration
13693.12 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
12696 DQN Noisy Networks for Exploration
12149.4 DQN A Distributional Perspective on Reinforcement Learning
10095.2 IMPALA (deep, multitask) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
9711.9 DDQN Deep Reinforcement Learning with Double Q-learning
9711 DQN Human-level control through deep reinforcement learning
3401.3 Human Human-level control through deep reinforcement learning
1971.0 Human Dueling Network Architectures for Deep Reinforcement Learning
520.5 Linear Human-level control through deep reinforcement learning
152.1 Random Human-level control through deep reinforcement learning
0 Contingency Human-level control through deep reinforcement learning

Normal Starts

Result Algorithm Source
38808.3 ACER Proximal Policy Optimization Algorithm
11378.4 PPO Proximal Policy Optimization Algorithm
6639.1 A2C Proximal Policy Optimization Algorithm