Env API
E95189
application programming interface
interface specification
reinforcement learning environment interface
Env API is the core interface specification in OpenAI Gym that standardizes how reinforcement learning environments interact with agents through methods like reset, step, and render.
Statements (43)
| Predicate | Object |
|---|---|
| instanceOf |
application programming interface
→
interface specification → reinforcement learning environment interface → |
| allows |
vectorized environment wrappers
→
|
| category |
OpenAI Gym core component
→
|
| compatibleWith |
model-based algorithms
→
policy gradient algorithms → value-based algorithms → |
| definedIn |
OpenAI Gym
→
|
| defines |
environment interface for agents
→
|
| documentationAvailableAt |
https://www.gymlibrary.dev/
→
|
| enables |
algorithm-environment interchangeability
→
|
| hasAttribute |
action_space
→
metadata → observation_space → reward_range → |
| hasMethod |
__init__
→
close → render → reset → seed → step → |
| hasVersion |
Gymnasium-compatible API
→
classic Gym API → |
| implementedBy |
gym.Env base class
→
|
| influenced |
design of many RL environment libraries
→
|
| maintainedBy |
OpenAI Gym project
NERFINISHED
→
|
| originatedIn |
Python ecosystem
→
|
| primaryGoal |
standardization of RL environment interfaces
→
|
| renderSupportsMode |
ansi
→
human → rgb_array → |
| requires |
deterministic method signatures
→
|
| resetReturns |
initial observation
→
|
| standardizes |
interaction between agents and environments
→
|
| stepReturns |
done flag
→
info dictionary → observation → reward → |
| stepTakes |
action
→
|
| supports |
continuing tasks
→
episodic tasks → |
| usedIn |
reinforcement learning
→
|
Referenced by (1)
| Subject (surface form when different) | Predicate |
|---|---|
|
OpenAI Gym
→
|
defines |