0
# Gymnasium
1

2
Gymnasium is a comprehensive Python library that provides a standardized API for developing and comparing reinforcement learning algorithms, serving as the maintained fork of OpenAI's Gym. It offers a diverse collection of environments including Classic Control (physics-based problems), Box2D (physics simulations), Toy Text (simple discrete environments), MuJoCo (complex multi-joint physics control), and Atari (game emulation environments).
3

4
## Package Information
5

6
- **Package Name**: gymnasium
7
- **Language**: Python
8
- **Installation**: `pip install gymnasium`
9

10
## Core Imports
11

12
```python
13
import gymnasium as gym
14
```
15

16
For working with specific components:
17

18
```python
19
from gymnasium import Env, Wrapper, Space
20
from gymnasium import make, register, spaces, wrappers
21
```
22

23
## Basic Usage
24

25
```python
26
import gymnasium as gym
27

28
# Create an environment
29
env = gym.make('CartPole-v1', render_mode='human')
30

31
# Reset environment to get initial observation
32
observation, info = env.reset(seed=42)
33

34
for _ in range(1000):
35
    # Take a random action
36
    action = env.action_space.sample()
37
    
38
    # Execute the action
39
    observation, reward, terminated, truncated, info = env.step(action)
40
    
41
    # Check if episode is done
42
    if terminated or truncated:
43
        observation, info = env.reset()
44

45
env.close()
46
```
47

48
## Architecture
49

50
Gymnasium's architecture is built around several core concepts:
51

52
- **Environment (Env)**: The main interface that defines the reinforcement learning task through step, reset, render, and close methods
53
- **Space**: Defines the structure and constraints of action and observation spaces
54
- **Wrapper**: Modifies environment behavior without changing the underlying environment
55
- **Registration System**: Manages environment discovery and instantiation through string IDs
56
- **Vector Environments**: Enables parallel execution of multiple environment instances
57

58
This design provides a flexible, extensible framework that serves as the standard interface for reinforcement learning in Python.
59

60
## Capabilities
61

62
### Core API
63

64
The fundamental classes and interfaces that form the foundation of all Gymnasium environments and wrappers.
65

66
```python { .api }
67
class Env:
68
    def step(self, action): ...
69
    def reset(self, *, seed=None, options=None): ...
70
    def render(self): ...
71
    def close(self): ...
72
    
73
class Wrapper(Env):
74
    def __init__(self, env): ...
75
```
76

77
[Core API](./core-api.md)
78

79
### Environment Registration
80

81
Functions for discovering, creating, and managing environments through the global registry system.
82

83
```python { .api }
84
def make(id: str | EnvSpec, max_episode_steps: int | None = None, **kwargs) -> Env: ...
85
def register(id: str, entry_point: str, **kwargs) -> None: ...
86
def spec(id: str) -> EnvSpec: ...
87
```
88

89
[Environment Registration](./registration.md)
90

91
### Spaces
92

93
Defines action and observation spaces with various types including continuous, discrete, composite, and structured spaces.
94

95
```python { .api }
96
class Box(Space): ...
97
class Discrete(Space): ...
98
class Dict(Space): ...
99
class Tuple(Space): ...
100
```
101

102
[Spaces](./spaces.md)
103

104
### Environment Wrappers
105

106
Pre-built wrappers for modifying environment behavior including observation transformation, action modification, reward shaping, and rendering enhancements.
107

108
```python { .api }
109
class TimeLimit(Wrapper): ...
110
class FlattenObservation(ObservationWrapper): ...
111
class ClipAction(ActionWrapper): ...
112
class NormalizeReward(RewardWrapper): ...
113
```
114

115
[Environment Wrappers](./wrappers.md)
116

117
### Vector Environments
118

119
Batched environment execution for improved performance when training with multiple parallel environments.
120

121
```python { .api }
122
class VectorEnv: ...
123
class SyncVectorEnv(VectorEnv): ...
124
class AsyncVectorEnv(VectorEnv): ...
125
```
126

127
[Vector Environments](./vector-environments.md)