Atari Up and Down Environment

Overview

Up’n Down is a vertically scrolling game that employs a pseudo-3D perspective.[citation needed] The player controls a purple dune buggy that resembles a Volkswagen Beetle.[citation needed] The buggy moves forward along a single-lane path; pressing up or down on the joystick causes the buggy to speed up or slow down, pressing right or left causes the buggy to switch lanes at an intersection, and pressing the “jump” button causes the buggy to jump in the air. Jumping is required to avoid other cars on the road; the player can either jump all the way over them, or land on them for points.[citation needed]

To complete a round, the player must collect 10 colored flags by running over them with the buggy. If the player passes by a flag without picking it up, it will appear again later in the round. The roads feature inclines and descents that affect the buggy’s speed, and bridges that must be jumped. A player loses a turn whenever the buggy either collides with another vehicle without jumping on it, or jumps off the road and into the grass or water.

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
105728.7 A3C LSTM Asynchronous Methods for Deep Reinforcement Learning
74705.7 A3C FF Asynchronous Methods for Deep Reinforcement Learning
54525.4 A3C FF 1 day Asynchronous Methods for Deep Reinforcement Learning
29443.7 Prioritized DDQN (prop, tuned) Prioritized Experience Replay
24759.2 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
22681.3 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
19086.9 DDQN (tuned) Deep Reinforcement Learning with Double Q-learning
16626.5 Prioritized DQN (rank) Prioritized Experience Replay
12157.4 Prioritized DDQN (rank, tuned) Prioritized Experience Replay
9896.1 Human Massively Parallel Methods for Deep Reinforcement Learning
8747.67 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
4721.1 DDQN Deep Reinforcement Learning with Double Q-learning
3311.3 DQN Massively Parallel Methods for Deep Reinforcement Learning
707.2 Random Massively Parallel Methods for Deep Reinforcement Learning

No-op Starts

Result Algorithm Source
436665.8 ACKTR Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
332546.75 IMPALA (deep) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
273058.1 IMPALA (shallow) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
194989.5 Reactor ND The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
103557 NoisyNet A3C Noisy Networks for Exploration
93931 DuDQN Noisy Networks for Exploration
89067 A3C Noisy Networks for Exploration
88148 IQN Implicit Quantile Networks for Distributional Reinforcement Learning
82155.3 IMPALA (deep, multitask) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
71260 QR-DQN-1 Distributional Reinforcement Learning with Quantile Regression
70790.4 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
64354.2 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
61326 NoisyNet DuDQN Noisy Networks for Exploration
53585 QR-DQN-0 Distributional Reinforcement Learning with Quantile Regression
44939.6 DDQN A Distributional Perspective on Reinforcement Learning
44939.6 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
33879.1 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
16769.9 DDQN Deep Reinforcement Learning with Double Q-learning
15612 C51 A Distributional Perspective on Reinforcement Learning
14255 NoisyNet DQN Noisy Networks for Exploration
12561.58 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
11693.2 Human Dueling Network Architectures for Deep Reinforcement Learning
11652 DQN Noisy Networks for Exploration
9989.9 DQN A Distributional Perspective on Reinforcement Learning
9082.0 Human Human-level control through deep reinforcement learning
8456 DQN Human-level control through deep reinforcement learning
3533 Linear Human-level control through deep reinforcement learning
2449 Contingency Human-level control through deep reinforcement learning
533.4 Random Human-level control through deep reinforcement learning

Normal Starts

Result Algorithm Source
145051.4 ACER Proximal Policy Optimization Algorithm
95445.0 PPO Proximal Policy Optimization Algorithm
17369.8 A2C Proximal Policy Optimization Algorithm