Assumption-free noise suppression for autonomous tractors tracking

摘要

Autonomous tractors have gained high interest from researchers due to the need for increased productivity in Agriculture. Their application include ploughing, weeding and crop spraying. One problem of these tractors which has not been fully addressed is tracking using the noisy measurements from a sensor such as RADAR sensor. Most publications assume the error in the measurement to be Gaussian during the position estimation process. This assumption has seen a poor performance of the estimators in case the sensor noise is non-Gaussian. This research work introduces the use of Separable Monte Carlos based Mean for non-Gaussian noise suppression applied to Autonomous tractor tracking. The Monte Carlos based Means work independent from any assumptions. Gaussian and Cauchy Noise are used in experimentation for RADAR sensor measurement. Results suggest that the Separable Monte Carlos based mean (SMC-MEAN) outperforms the Kalman Filter and the Maximum A Posterior (MAP) in the Mean square error (MSE) sense hence can be of practical use in Autonomous tractor tracking.