Probabilistic and Reinforcement Learning Track
2023 International Planning Competition
2023 International Planning Competition
Multi-UAV Trajectory Planning
| Example name | UAV ** |
| Action space | Dict |
| State space | Dict |
** stand for continuous, discrete, mixed.
A UAV fleet, described by first order constrained dynamics, are required to reach a goal position in the x-y-z space from thier initial position. The UAV are represented by their 6 positional coordinates (cartesian coordinates and angles), and the velocity in the direction of the nose. The velocity is a first order integrator, i.e., integration over the acceleration, while the angles are assumed to be also first order integrator, i.e., small directe changes to the angles are assumed to be possible. The goal is to navigate the drones to the goal location. Note that some of the drones are not controllable! and so are there but nothing we do can affect them!
This domain has three version!
| Constant | Type | Desc |
|---|---|---|
| CONTROLLABLE(aircraft) | float32 | Whether an aircraft is contrallable |
| GRAVITY | float32 | Earth gravity acceleration coefficient |
| MIN_X | float32 | Min X of the space |
| MAX_X | float32 | Max X of the space |
| MIN_Y | float32 | Min Y of the space |
| MAX_Y | float32 | Max X of the space |
| MIN_Z | float32 | Min Z of the space |
| MAX_Z | float32 | Max X of the space |
| SCALE_FACTOR | float32 | Time scale factor for dynamic equations (Delta T) |
| RANDOM_WALK_COEFF | float32 | RAndom walk var of the non contrallable aircrafts |
| VEL_REG | float32 | Velocity regularization (prevent division by zero |
| MIN_ACC(aircraft) | float32 | Minmum acceleration value |
| MAX_ACC(aircraft) | float32 | Maximum acceleration value |
| MIN_PHI(aircraft) | float32 | Minimum phi change in the single time step |
| MAX_PHI(aircraft) | float32 | Maximum phi change in the single time step |
| MIN_THETA(aircraft) | float32 | Minimum theta change in the single time step |
| MAX_THETA(aircraft) | float32 | Maximum thera change in the single time step |
| GOAL_X(aircraft) | float32 | Aircraft’s goal X position |
| GOAL_Y(aircraft) | float32 | Aircraft’s goal Y position |
| GOAL_Z(aircraft) | float32 | Aircraft’s goal Y position |
All of these can be read from the RDDLEnv interface and from the RDDL files.
The actions are the acceleration in the direction of the nose, and the small changes to the roll and pitch angles.
| Action | type | Desc |
|---|---|---|
| set_acc(aircraft) | Box(1, MIN_ACC(aircraft), MAX_ACC(aircraft), np.float32) | Propelling force to apply to the aircraft |
| set_phi(aircraft) | Box(1, MIN_PHI(aircraft), MAX_PHI(aircraft), np.float32) | Roll angle change |
| set_theta(aircraft) | Box(1, MIN_THETA(aircraft), MIN_THETA(aircraft), np.float32) | Pitch angle change |
| Action | type | Desc |
|---|---|---|
| set_acc(aircraft) | Discrete(MAX_ACC(aircraft)+MIN_ACC(aircraft)+1, start=MIN_ACC(aircraft)) | Propelling force to apply to the aircraft |
| set_phi(aircraft) | Discrete(MAX_PHI(aircraft)+MIN_PHI(aircraft)+1, start=MIN_PHI(aircraft)) | Roll angle change |
| set_theta(aircraft) | Discrete(MAX_THETA(aircraft)+MIN_THETA(aircraft)+1, start=MIN_THETA(aircraft)) | Pitch angle change |
| Action | type | Desc |
|---|---|---|
| set_acc(aircraft) | Box(1, MIN_ACC(aircraft), MAX_ACC(aircraft), np.float32) | Propelling force to apply to the aircraft |
| set_phi(aircraft) | Discrete(MAX_PHI(aircraft)+MIN_PHI(aircraft)+1, start=MIN_PHI(aircraft)) | Roll angle change |
| set_theta(aircraft) | Discrete(MAX_THETA(aircraft)+MIN_THETA(aircraft)+1, start=MIN_THETA(aircraft)) | Pitch angle change |
The state space represents the position and angles of each UAV, and also the velocity in the direction of the nose.
| State | type | Desc |
|---|---|---|
| pos_x(aircraft) | Box(1, MIN_X, MAX_X, np.float32) | Position of the UAV in the x axis |
| pos_y(aircraft) | Box(1, MIN_Y, MAX_Y, np.float32) | Position of the UAV in the y axis |
| pos_z(aircraft) | Box(1, MIN_Z, MAX_Z, np.float32) | Position of the UAV in the z axis |
| theta(aircraft) | Box(1, -inf, inf, np.float32) | The UAV’s pitch angle |
| phi(aircraft) | Box(1, -inf, inf, np.float32) | The UAV’s roll angle |
| psi(aircraft) | Box(1, -inf, inf, np.float32) | The UAV’s yaw angle |
| vel(aircraft) | Box(1, -inf, inf, np.float32) | UAV’s velocity in the direction of the nose |
The reward function is defined as
The sum of the norm-2 distance of all the controllable drones.