The brain of the AI agent. It has lifecycle methods, receives actions and sends observations. Inherits from MLAgents.Agent More...

Inheritance diagram for AgentBrain:

Public Member Functions
override void	OnEpisodeBegin ()
	Lifecycle method that gets called in the beginning of an episode (by the mlagents code). The robot is rebuilt and the joints are randomized.

override void	CollectObservations (VectorSensor sensor)
	Observations are collected that we want to send back to the Python side. More...

override void	OnActionReceived (ActionBuffers actions)
	Called when actions are received from the Python side More...

Detailed Description

The brain of the AI agent. It has lifecycle methods, receives actions and sends observations. Inherits from MLAgents.Agent

Member Function Documentation

override void AgentBrain.CollectObservations ( VectorSensor sensor )

inline

Observations are collected that we want to send back to the Python side.

Joint angles and joint positions in the space, and the position of the end effector (= last game object within manipulator).

Like this: [j1_x, j1_y, j1_z, j1_θ, j2_x, j2_y, j2_z, j2_θ, ..., ee_x, ee_y, ee_z]

Parameters

sensor Object on which observations are added

override void AgentBrain.OnActionReceived ( ActionBuffers actions )

inline

Called when actions are received from the Python side

Parameters

actions Object that has the actions. Every action represent how much a joint should move

The documentation for this class was generated from the following file: