The NODE lab research focuses on estimation and diagnosis in cyber-physical systems, human-robot cooperative controls, and distributed control to improve reliability and performance in robotic, assistive robotic, and intelligent transport systems.
Partners
Government and National Partners
Academic Collaborators (on Autonomous Navigation and Assistive Robotics)
Research Topics
Distributed Estimation, Diagnosis, and Event-Triggered Control
In this research theme we focus on: i) Developing a control framework for an agent (mobile robot) in an uncertain environment over a connected sensor network, such that the agent is able to finish a sequence of tasks, namely, reaching certain sets in order. Considering constrained communication between the agent and the sensor network, an event-triggered task-switching controller and a distributed filter (to estimate the agent state) is being developed; and ii) Robustness analysis of other second-order consensus dynamics (e.g. mechanical and power networks) in terms of the physical parameters and the spectrum of the grounded Laplacian matrix associated with the network structure.
Related Publications:
Automatica2021, Robot2020, IFAC2020S, ITS2018, ITS2017_1, IV2017, ITS2017_2, IFAC2017
Autonomous and Connected Vehicles: Safety and Reliability Enhancement
The research projects in this theme include: i) Developing a cyber-physical system framework for reliable vehicle state estimation, robustly to road friction condition variations quantizing the effect of disturbances and adversaries due to occasional GPS signal drops. An integrated trajectory planning-tracking and crash mitigation controller is also being developed for connected AVs to operate in extreme handling conditions under multi-actuation control strategies; ii) Securing connected mobile autonomous agents when an unknown mobile sensor is under positioning attack subject to bounded system uncertainties; and iii) Designing an event-triggered add-on safety mechanism to adjust the controller for timely actuation in a networked vehicular system, while maintaining maneuverability. To this end, two different wireless technologies (LTE and 5G) for communication between the infrastructure and the vehicles, and the effect of slicing-enabled network are being investigated.
Related Publications:
USPatent, USPatent, USPatent, TCST2021, TII2021, ACC2021, ACC2020, IFAC2020, TMech2020, TII2019, Attack2020, Systems2020, CEP2017, CEP2018, TCST2018, VSD2016,
Assistive Robotics
The first theme of this research program focuses on Inverse Optimal Control (IOC) problems and optimal cost function identification using structured classification for natural human walking. This will be used for assistive robotic controller design (e.g. lower limb exoskeletons and human-exoskeleton controls). A nature-inspired curriculum learning (CL) and a deep reinforcement learning (DRL) based framework are developed considering neuromechanical reward functions. This is to generate close-to-natural walking and facilitating decentralized adaptation between the user and the assistive robotic devices.
The second theme includes learning-aided human state estimation through augmenting nonlinear and invariant observers (with stability guarantee) with DL, for a cooperative human-robot control framework, to enhance safety of the users. It also focuses on developing a generic framework for fault detection and recovery aided by explicit kinematic features for assistive technologies. Furthermore, developing optimal preview controller and stable performance of the closed-loop walk robust to disturbances is explored for a biped robot (Nao). To reduce computational load, a closed-form IK solutions are obtained and a UKF state observer is used for the preview control.
Related Publications:
