Real time obstacle motion prediction using neural network based extended Kalman filter for robot path planning

Abstract EN

Navigation for mobile robots in dynamic environments necessitates estimating the path of dynamic obstacles, which is accomplished in this study using an enhanced kalman filter.

The measured data, however, contains bias and noise.

The SDAE, a deep learning-based neural network structure, delivers noise-free data that the Kalman filter uses to construct an optimal measurement noise covariance matrix.

This matrix is then used by the Kalman filter to estimate an error-free obstacle path.The SDAE is trained using both the Adam and stochastic gradient learning algorithms.

To ensure safe navigation, the robot’s path is re-planned based on the estimated obstacle path.

Numerical simulations using MATLAB demonstrate that the novel methodology is more relevant and superior to traditional Kalman and Particle filter approaches, and that it can be applied in a variety of navigational applications.

In terms of computing time and robustness in closely spaced obstacles, simulation testing indicated that path planning using the proposed technique excels the hybrid A star, artificial potential field, and decision algorithms.