The Kalman uses sensor measurements to aid the core process model determined by the IMU data. Gains within the system control the weighting placed on the process model and the sensor measurements. With a low gain, the filter weights the process model and the state estimate will more closely follow the system state estimated by the IMU data. With a high gain, the filter weights the sensor measurements and the state estimate will more closely follow the system state estimated by the sensor measurements. Greensea uses two gains to specify a confidence in the IMU and the sensor data independently. We determine these gains through data sheet analysis, testing, and field experience. A configuration file in the INS processor provides the gains for each sensor channel. The term "tuning", when applied to an INS, typically refers to updating the gains of the aiding sensors to optimize the way their measurements are used in the solution. These are based on the comparison of the estimated state to the true-life state in field trials.
The goal of the INS is to provide a total state estimation of the submersible body that is as close to the actual state as possible, and certainly closer than any one discrete measurement would provide. More specifically, we want the average of our state estimate to equal the average of the true-life state and we want any variation of our state estimate from the true-life state to be as small as possible.
