Patrol - MARL formation¶
Close-patrol (formation-keeping) environment: state, parameters and transition.
| Action space | Box((3,)), all actions in [-1, 1] |
| Observation space | Box((26,)) |
| Tracked variable(s) | slot error (m) |
| Episode length | 200 steps (200 s at 1 s per step) |
| Import | from target_gym import PlanePatrol, PatrolParams |
| Cite as | patrol-v1 |
Action space¶
Actions are normalised to [-1, 1] and mapped onto the plant's real
actuator range inside the environment.
| # | meaning | min | max |
|---|---|---|---|
| 0 | power | -1 | 1 |
| 1 | stick | -1 | 1 |
| 2 | aileron | -1 | 1 |
Observation space¶
26 values. Indices (24,) carry the tracked variable(s) that the reward scores.
Rewards¶
See the environment's compute_reward.
Every environment in this suite scores on one contract: the reward is
(tracking terms, multiplied) x (1 - weighted costs), bounded in
[0, 1], and reaches 1 only while the target is held exactly. See
Reward shaping.
Starting state¶
reset samples the initial condition and the target; state has 12 fields.
Episode end¶
Termination. See check_is_terminal.
Truncation. After 200 steps.
Baselines¶
A tuned PID ships with this environment.
Arguments¶
| parameter | default |
|---|---|
delta_t |
1 |
max_steps_in_episode |
200 |
gravity |
9.81 |
initial_mass |
73500 |
thrust_output_at_sea_level |
240000 |
air_density_at_sea_level |
1.225 |
frontal_surface |
12.6 |
wings_surface |
122.6 |
C_x0 |
0.095 |
C_z0 |
0.9 |
initial_fuel_quantity |
19088 |
specific_fuel_consumption |
0.0175 |
power_response_rate |
0.05 |
stick_response_rate |
0.9 |
| … | 57 more, see the params dataclass |