Local path planning based on Ridge Regression Extreme Learning Machines for an outdoor robot

Abstract

For mobile robot local path planning under outdoor environment, Ridge Regression Extreme Learning Machines (RRELM) is adopted, it is a fast machine learning classification method to apply in path planning. Firstly, the laser rangefinder data are extracted and marked to describe the outdoor environment. Secondly, ridge regression theory is utilized to improve the generalization ability of ELM for local path planning. Meanwhile, the start-goal point constraint is considered for planning. Additionally, abrupt dynamic obstacle is regarded as a kind of disturbance to plan the path by RRELM. Then the optimal path is estimated by the distance evaluation function among feasible paths. Finally, a great deal of outdoor robot simulation experiments are shown that RRELM find out the safety path for outdoor robot, and the generalization ability, smoothness and rapidity of RRELM for path planning are better than SVM and ELM, furthermore, the performance of RRELM for the dynamic environment is also efficient. © 2015 IEEE.