Dynamic key-frame selection technique for enhanced odometry estimation based on laser scan similarity comparison

Abstract

A dynamic key-frame selection technique is proposed for enhanced odometry estimation. The odometry estimation problem has a drift problem in which errors are accumulated over time. The proposed technique finds the key-frame which reduces the odometry drift for every time step based on scan similarity comparisons. The enhancement of the proposed method is verified through the performance comparison with the frame-to-frame and the conventional frame to key-frame methods.