engleski

Neural Network Based Correction of Odometry Errors in Mobile Robots

Odometry is the most widely used method for estimation of the momentary position of mobile robots. It provides easily accessible real-time positioning information. However, with time, odometric localization accumulates errors in an unbounded fashion. This paper describes a way to decrease the odometry error by using a neutral network as a soft sensor. It makes a correction to the estimated orientation of the mobile robot, which is the most influenced on accumulated odometry errors. Trained neural network can be used instead of a gyro or a compass in some mobile robot applications.

Mobile Robots; Odometry; Neural Network

nije evidentirano