Project information

• Category: Robotics, AI
• Project date: Fall 2018 - Fall 2019
• Github Repository: https://github.com/abradat/ball-trajectory-predictor

Project Description
a method for prediciting a rolling ball's trajectory based on the combination of K-NN Regression and Autoregression methods.

Overview

Prediction of the trajectory of the moving objects is of paramount importance in different fields of intelligent systems. In this paper, we proposed a novel method for prediction of trajectory of the ball on the soccer field for the humanoid robots. Our proposed method is based on combination of the k-nearest neighbor regression and autoregression models. First, k-nearest neighbor regression is enhanced by adding a cost function, which constitutes the fixed prediction scheme. Moreover, during the movement of the ball, autoregression method coefficients are updated in fixed predefined steps in order to lessen the error of the prediction, which constitutes the adaptive prediction scheme. Our proposed method is tested by simulation and practical implementation. The results demonstrate a high precision rate, in addition to the effectiveness of the adaptive scheme in case of applying external disturbance on the ball trajectory.

Publication

The results of this work has been published as a conference paper in ICROM 2019. The full-text version of the paper is available through here

Technologies/Languages Used

Technology	Usage
	Python language is for implementing trajectory predictor code.
	C++ is used for writing scripts in ROS framework and Gazebo simulator.
	ROS is the framework for running robot code and connecting to Gazebo.
	Gazebo is the simulator used for testing our simulations.