Atari Video Pinball Environment

Overview

Video Pinball is a loosesimulation of a pinball machine: ball shooter, flippers, bumpers and spinners. It includes a unique rollover bonus with an Atari Inc. logo on the playfield; hitting the logo four times results in an extra ball.

Most of the game play involves learning how to perform specific functions, such as launching the ball or activating the flippers, with the Atari joystick. Moving the joystick controller down pulls the pinball machine plunger back while pressing the joystick button shoots the ball into the playfield. The left and right flippers are activated by moving the joystick controller left or right. The ball can be nudged (as in nudging a table gently in real life) by holding down the joystick button and moving the controller in a particular direction.

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
506817.2 Rainbow Rainbow: Combining Improvements in Deep Reinforcement Learning
470310.5 A3C LSTM Asynchronous Methods for Deep Reinforcement Learning
455052.7 Distributional DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
447408.6 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
374886.9 Prioritized DDQN (prop, tuned) Prioritized Experience Replay
367823.7 DDQN (tuned) Deep Reinforcement Learning with Double Q-learning
331628.1 A3C FF Asynchronous Methods for Deep Reinforcement Learning
295972.8 Prioritized DDQN (rank, tuned) Prioritized Experience Replay
214925.3 Prioritized DQN (rank) Prioritized Experience Replay
185852.6 A3C FF 1 day Asynchronous Methods for Deep Reinforcement Learning
148883.6 DDQN Deep Reinforcement Learning with Double Q-learning
112093.37 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
110976.2 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
20452.0 Random Massively Parallel Methods for Deep Reinforcement Learning
20228.1 DQN Massively Parallel Methods for Deep Reinforcement Learning
15641.1 Human Massively Parallel Methods for Deep Reinforcement Learning

No-op Starts

Result Algorithm Source
949604 C51 A Distributional Perspective on Reinforcement Learning
876503 DuDQN Noisy Networks for Exploration
870954 NoisyNet DuDQN Noisy Networks for Exploration
705662 QR-DQN-1 Distributional Reinforcement Learning with Quantile Regression
701779 QR-DQN-0 Distributional Reinforcement Learning with Quantile Regression
698045 IQN Implicit Quantile Networks for Distributional Reinforcement Learning
572898.27 IMPALA (deep) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
533936.5 Rainbow Rainbow: Combining Improvements in Deep Reinforcement Learning
496101.0 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
479197.0 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
478646.7 Distributional DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
469366 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
429936 DQN Noisy Networks for Exploration
322507 NoisyNet DQN Noisy Networks for Exploration
294724 NoisyNet A3C Noisy Networks for Exploration
261720.2 Reactor ND The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
229402 A3C Noisy Networks for Exploration
228642.52 IMPALA (shallow) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
196760.4 DQN A Distributional Perspective on Reinforcement Learning
157550.21 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
100496.6 ACKTR Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
98209.5 DDQN A Distributional Perspective on Reinforcement Learning
98209.5 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
70009.0 DDQN Deep Reinforcement Learning with Double Q-learning
42684 DQN Human-level control through deep reinforcement learning
20125.14 IMPALA (deep, multitask) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
19761 Contingency Human-level control through deep reinforcement learning
17667.9 Human Dueling Network Architectures for Deep Reinforcement Learning
17297.6 Human Human-level control through deep reinforcement learning
16871 Linear Human-level control through deep reinforcement learning
16256.9 Random Human-level control through deep reinforcement learning

Normal Starts

Result Algorithm Source
156226.6 ACER Proximal Policy Optimization Algorithm
37389.0 PPO Proximal Policy Optimization Algorithm
19735.9 A2C Proximal Policy Optimization Algorithm