Parameters that used in this challenge is specified as below:
| AGENT | |
|---|---|
| HEIGHT (m) | 0.88m |
| MASS (kg) | 9.2 |
| RADIUS | 0.18m |
| SENSORS | [RGB_SENSOR, DEPTH_SENSOR, POINTGOAL_SENSOR, VELOCITIES_SENSOR] |
| POSSIBLE_ACTIONS | [LINEAR_VELOCITY, ANGULAR_VELOCITY] |
| MAX_EPISODE_STEPS | 500 |
| DEPTH_SENSOR | |
| HEIGHT | 180 |
| WIDTH | 320 |
| Horizontal FOV | 69.4 |
| ORIENTATION (Euler angles) | [0, 0.3490659, 0] # 20 degrees |
| MAX_DEPTH | 10 |
| MIN_DEPTH | 0.1 |
| NORMALIZE_DEPTH* | TRUE |
| POSITION | [0, 0, 0.88] |
| RGB_SENSOR | |
| HEIGHT | 180 |
| WIDTH | 320 |
| Horizontal FOV | 69.4 |
| ORIENTATION (Euler angles) | [0, 0.3490659, 0] # 20 degrees |
| POSITION | [0, 0, 0.88] |
| POINTGOAL_SENSOR | |
| GOAL_FORMAT: | POLAR |
| SUCCESS CRITERIA | |
| SUCCESS_GEODESIC_DISTANCE | 0.36m |
* Depth Normalization:
We do depth normalization as the following (pesudo code):
depth[isnan(depth)] = 0
depth[depth > high] = 0
depth[depth < low] = 0
normalized_depth = depth / high