Accurate Estimation of 3D-Repetitive-Trajectories using Kalman Filter, Machine Learning and Curve-Fitting Method for High-speed Target Interception

Accurate estimation of trajectory is essential for the capture of any high-speed target. This chapter estimates and formulates an interception strategy for the trajectory of a target moving in a repetitive loop using a combination of estimation and learning techniques. An extended Kalman filter estimates the current location of the target using the visual information in the first loop of the trajectory to collect data points. Then, a combination of Recurrent Neural Network (RNN) with least-square curve-fitting is used to accurately estimate the future positions for the subsequent loops. We formulate an interception strategy for the interception of a high-speed target moving in a three-dimensional curve using noisy visual information from a camera. The proposed framework is validated in the ROS-Gazebo environment for interception of a target moving in a repetitive figure-of-eight trajectory. Astroid, Deltoid, Limacon, Squircle, and Lemniscates of Bernoulli are some of the high-order curves used for algorithm validation.