We are working hard on a new Locify version. We would like to introduce you one of many new features - usage of a Kalman filter.

With just a walk speed and actual available precission of hand GPS recievers, its necessary to implement some kind of filter which smooth paths and reduce inaccuracies. One of the best filters is called Kalman filter (read more).

Short description : Kalman filter can be divided into two parts.

1. Prediction stage - On a measure from last cycle, filter predicts new position of user including suitable error model (mean error).

2. Update stage - Due to actual NEW measure, „Kalman gain“ is computed a predicted measure is adjust. Main influence on a result has HDOP (the horizontal dilution of precision) of the measure. If a new precision is better than in the last cycle, new position is closer to the actual measure than to prediction and vice-versa.

And what are the benefits for the user? For slow speed, Kalman filter comes with great improvement of navigation fluency. For example in geocaching, filter will recognize that you are actually standing at one place and the navigation arrow will not go "nuts". Big improvement is also in route recording - actual smooth route is recorded and correct route distance is calculated.

Labels: client, J2ME, technology