Deep Neural Network Stabilisation for Oscillatory Based Manipulators Influenced by Varying Delays and Dynamic Exogeneous Disturbances

Background: Robotic manipulators are increasingly being applied in various industrial applications such as telehealth and cargo port robotic cranes. These applications often involve highly dynamic environments where the robotic bases cause significant vibrations. Additionally, when these manipulators are guided over the internet, they face varying and unpredictable delays (internet latency). These factors pose substantial challenges to the stability and performance of the robotic systems.

Research Gap: Conventional stabilizing methods can minimize the impact of oscillatory disturbances and delays to a certain extent. However, these methods are not sufficient to fully suppress the vibrational impact, leading to the impossibility of achieving asymptotic stabilization. There is a critical need for advanced control methods that can comprehensively address these challenges and ensure the stability of robotic manipulators under significant exogeneous disturbances and internet delays.

Aim: The aim of this project is to develop and implement deep neural network control methods to stabilize robotic manipulators, irrespective of significant exogeneous disturbances and internet delays.

Objectives:

Develop deep neural network algorithms to predict and mitigate the effects of oscillatory disturbances and varying delays.

Implement adaptive control strategies to stabilize robotic manipulators.

Validate the proposed methods through simulations and real-world experiments.

Project Description: In this project, we propose to use deep neural network control methods to comprehensively harness information about the vibration as well as varying delays before wielding adaptive control forces to the robotic manipulators. We aim to stabilize the robots irrespective of the significant exogeneous disturbances and internet delays.