Sprint #10 (July 6)

[pagand]

Yimeng: 1- consider Evolutional model with time series analysis (Consider using Kalman Filter) 2- Compare two versions Pitch as separate input/ or based on table Inputs only being independent variables or last time step of output as well 3- Input and output variables config Time-dependent model Have previous torque / thrust / longitude / latitude as input to predict current values of these fields 4- Datetime fields inconsistency Use “Dati” for getting hour related information Use “Time” for trip duration 5- Torque vs. Thrust: compute resistance See if we can get any information about waves 6- In next biweekly meeting: Ask BC ferries about value Pitch: realistic magnitude and whether we could use it

Jack: 1- Use higher dimension linear model for navigational evolution of the vessel. 2- learn the parameters with linear regression approach 3- Use Kalman filter for updating the latitude and longitude given your linear model and the observation

[pagand]

@yim-fan Please update these Sprint if any of them has finished. Then put them to the review after updating the github.

[yim-fan]

• [x] Consider Evolutional model with time series analysis (Consider using Kalman Filter)

  • Tried Kalman filter for imputing missing data, not working well. Simply imputation seems fine

• [x] Compare two versions: Pitch as separate input/ or based on table vs. Inputs only being independent variables or last time step of output as well

  • not using pitch anymore since the values seems unrealistic

• [x] Input and output variables config

• [x] Datetime fields inconsistency Stick with Time within each trip for time-series transformer

• [x] Torque vs. Thrust: compute resistance

• [x] See if we can get any information about waves No source of wave info acquired, but so far the seems fine without waves