Atari Road Runner Environment

Overview

The player controls Road Runner, who is chased by Wile E. Coyote. In order to escape, Road Runner runs endlessly to the left. While avoiding Wile E. Coyote, the player must pick up bird seeds on the street, avoid obstacles like cars, and get through mazes. Sometimes Wile E. Coyote will just run after the Road Runner, but he occasionally uses tools like rockets, roller skates, and pogo-sticks.

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
73949.0 A3C LSTM Asynchronous Methods for Deep Reinforcement Learning
58549.0 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
57207.0 Prioritized DQN (rank) Prioritized Experience Replay
56990.0 Prioritized DDQN (prop, tuned) Prioritized Experience Replay
56086.0 Distributional DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
54630.0 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
54261.0 Rainbow Rainbow: Combining Improvements in Deep Reinforcement Learning
52264.0 Prioritized DDQN (rank, tuned) Prioritized Experience Replay
43884.0 DDQN Deep Reinforcement Learning with Double Q-learning
43156.0 DDQN (tuned) Deep Reinforcement Learning with Double Q-learning
43079.8 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
34216.0 A3C FF Asynchronous Methods for Deep Reinforcement Learning
31769.0 A3C FF 1 day Asynchronous Methods for Deep Reinforcement Learning
9264.0 DQN Massively Parallel Methods for Deep Reinforcement Learning
6878.0 Human Massively Parallel Methods for Deep Reinforcement Learning
200.0 Random Massively Parallel Methods for Deep Reinforcement Learning

No-op Starts

Result Algorithm Source
234352 NoisyNet DuDQN Noisy Networks for Exploration
111310.0 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
71168 Reactor The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
69524.0 DDQN A Distributional Perspective on Reinforcement Learning
69524.0 DuDQN Dueling Network Architectures for Deep Reinforcement Learning
67780 QR-DQN-0 Distributional Reinforcement Learning with Quantile Regression
66790.5 Reactor ND The Reactor: A fast and sample-efficient Actor-Critic agent for Reinforcement Learning
64262 QR-DQN-1 Distributional Reinforcement Learning with Quantile Regression
64051 DuDQN Noisy Networks for Exploration
63366.0 Distributional DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
62151.0 PDD DQN Dueling Network Architectures for Deep Reinforcement Learning
62041.0 Rainbow Rainbow: Combining Improvements in Deep Reinforcement Learning
57900 IQN Implicit Quantile Networks for Distributional Reinforcement Learning
57121.0 IMPALA (deep) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
55839 C51 A Distributional Perspective on Reinforcement Learning
53446.0 ACKTR Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
51007.99 Gorila DQN Massively Parallel Methods for Deep Reinforcement Learning
48377.0 DDQN Deep Reinforcement Learning with Double Q-learning
45993 NoisyNet DQN Noisy Networks for Exploration
45315 A3C Noisy Networks for Exploration
39544.0 DQN A Distributional Perspective on Reinforcement Learning
37910 DQN Noisy Networks for Exploration
37505.0 IMPALA (shallow) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
30454 NoisyNet A3C Noisy Networks for Exploration
24435.5 IMPALA (deep, multitask) IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
18257 DQN Human-level control through deep reinforcement learning
7845.0 Human Dueling Network Architectures for Deep Reinforcement Learning
7845.0 Human Human-level control through deep reinforcement learning
89.1 Contingency Human-level control through deep reinforcement learning
67.7 Linear Human-level control through deep reinforcement learning
11.5 Random Human-level control through deep reinforcement learning

Normal Starts

Result Algorithm Source
35466.0 ACER Proximal Policy Optimization Algorithm
32810.0 A2C Proximal Policy Optimization Algorithm
25076.0 PPO Proximal Policy Optimization Algorithm