PlatformTrack

A cooperative control task where a group of k UAVs are connected together by a rigid frame to form an overactuated platform. Each individual UAV, attached by a 2-DoF passive gimbal joint, acts as a thrust generator. The goal for the agents is to collectively control the platform to track a given reference trajectory while orienting the platform towards a given direction.

Observation

The observation is a CompositeSpec containing:

• `obs_self`` (1, *): The state of each UAV observed by itself, containing its kinematic information with the position being relative to the frame center, and an one-hot identity indicating the UAV’s index.

• `obs_others`` (k-1, *): The observed states of other agents.

• obs_frame:

• state_frame: (1, *): The state of the frame.

• rpos (3 * future_traj_steps): The relative position of the platform to the reference positions in the future future_traj_steps time steps.

• time_encoding (optional): The time encoding, which is a 4-dimensional vector encoding the current progress of the episode.

Reward

• reward_pose: The reward for the pose error between the platform and the reference (position and orientation).

• reward_up: The reward for the alignment of the platform’s up vector and the reference up vector.

• reward_spin: Reward computed from the spin of the drone to discourage spinning.

• reward_effort: Reward computed from the effort of the drone to optimize the energy consumption.

The total reward is computed as follows:

\[r = r_\text{pose} + r_\text{pose} * (r_\text{up} + r_\text{spin}) + r_\text{effort}\]

Config

Parameter	Type	Default	Description
drone_model	str	“hummingbird”
num_drones	int	4
arm_length	float	0.85
reset_thres	float	0.5
future_traj_steps	int	4
reward_distance_scale	float	1.2
time_encoding	bool	True