Neural network-based calibration for accuracy improvement in lateration positioning system

Abstract

Mobile robot positioning is a crucial problem in modern industrial autonomous solutions. Lateration Positioning Systems base on the distance measurements to estimate the object's position. These measurements are however often affected by numerous sources of noise: obstacles, multi-pathing, signal propagation speed etc. Effective calibration methods are therefore required to eliminate these errors to achieve precise positioning. In this paper, we present the application of neural networks to improve the accuracy of a UWB lateration system. We present the network architecture and demonstrate how it can be used to alleviate the effects of multi-pathing and environment anisotropy in a real positioning setup. We furthermore compare the efficiency of the neural network with the state-of-the-art calibration methods.