* Reset Parameters: Two, corresponding to goal size, and goal movement speed.
## Drone
![Drone](../images/drone.png)
* Set-up: Quadcopters have 4 engines to propel them.
* Goal: The agents must move towards a target in 3D.
* Agents: The environment contains 10 agent linked to a single brain.
* Agent Reward Function (independent):
* Between 0 and +0.5 Increasing the closer the agent is to his target.
* +1 For Being inside the target.
* Brains: One brain with the following state/action space.
* State space: (Continuous) 24 variables corresponding to position relative to target, rotation, velocity, angular velocity, acceleration and anglular acceleration.
* Action space: (Continuous) Size of 4, corresponding to the vertical force applied to each engine.
* Observations: None
* Reset Parameters: One, corresponding to goal size.