There are three main challenges that have been outlined:
1. Sorting good data vs. bad data
- Based on human ‘truth’ data provided
1b. Separating Truth Data
- Find distinguishing parameters between different truth data/ labeled maneuvers
- Algorithm that can sort them from each other
2. Identifying Maneuvers
- Automatically identify and label each maneuver flown in this dataset from a catalog of around 30 maneuvers
- Maneuver labels should place a “start” and “stop” label in the time series data
- No truth data – best effort
3. Grading Maneuvers
- Grade each maneuver according to the following grading scale:
- (U) Unsatisfactory- Maneuver was unsafe
- (F) Fair- Maneuver was safe but not performed according to procedures or not within allowable tolerances
- (G) Good- Maneuver was safe and performed in accordance with procedures and within allowable tolerances
- (E) Excellent- Maneuver was performed in an exceptional manner
- No truth data – best effort