On-Line Learning for Planning and Control of Underactuated Robots With Uncertain Dynamics

Giulio, Turrisi,; Marco, Capotondi,; Gaz, Claudio; Modugno, Valerio; Oriolo, Giuseppe; Luca, Alessandro De

doi:10.1109/lra.2021.3126899

Cited by 8 publications

(3 citation statements)

References 27 publications

(34 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The main method of motion planning is to preset a reasonable reference trajectory from the output of the underactuated systems, and then design the controller based on the reference trajectory to meet the control requirements. The motion planning methods mainly include offline trajectory planning method (Turrisi et al, 2021; T. Yang et al, 2018) and input‐shaping method (K. T. Hong et al, 2003). There is a strong coupling between the motions of the boom, the trolley, and the lifting load, and the nonlinear relationship between the state parameters increases with the rope length.…”

Section: Overview Of Control Theory and Methodsmentioning

confidence: 99%

A review for control theory and condition monitoring on construction robots

Shi

Bai

et al. 2023

Journal of Field Robotics

View full text Add to dashboard Cite

The application of robotic technologies in building construction leads to great convenience and productivity improvement, and construction robots (CRs) bring enormous opportunities for the way we conduct design and construction. To get a better understanding of the trends and track the application of CRs for on‐site conditions, this paper conducts a systematic review of control models and status monitoring of CRs, which are two key aspects that determine construction accuracy and efficiency. Control accuracy and flexibility are primary needs for CRs applied in different scenes, so the control methods based on driving models are vitally important. Status monitoring on CRs contains knowledge in fault detection, intelligence maintenance, and fault‐tolerant control, and multiple objectives need to be met and optimized in the whole drive chain. Moreover, the state‐of‐the‐art is comprehensively summarized, and new insights are also provided to carry on promising researches.

show abstract

Section: Overview Of Control Theory and Methodsmentioning

confidence: 99%

A review for control theory and condition monitoring on construction robots

Shi

Bai

et al. 2023

Journal of Field Robotics

View full text Add to dashboard Cite

show abstract

“…In [28], the control law to learn the target velocity with the Gaussian process model was considered. In [29], the authors proposed an iterative learning method for underactuated robotic systems with the Gaussian process regression. The authors in [30] considered probabilistic model predictive control based on reinforcement learning.…”

Section: Introductionmentioning

confidence: 99%

Cooperative Target Tracking by Multiagent Camera Sensor Networks via Gaussian Process

2022

View full text Add to dashboard Cite

In this paper, we present learning-based robust cooperative control for camera sensor networks. The dynamics of each camera agent with the pan and tilt mechanism is modeled by the Euler-Lagrange equation. In the target tracking problem, the modeling errors caused by high nonlinearity of camera system dynamics and changes in the tracking environment may incur significant deterioration of tracking performance. To address this, we propose a distributed control method based on the Gaussian process regression to compensate for the model uncertainty of the dynamics of the camera sensor network. We consider the error dynamics of the multiagent Euler-Lagrange system to evaluate the tracking errors of the camera agents. We show that the ultimate boundedness of the error dynamics is achieved by the proposed learning-based method. A numerical example confirms that the proposed control law enables each camera agent to track the targets with a bounded tracking error in the presence of model uncertainty.

show abstract

“…Yet, to the best of the Authors' knowledge, no solution ensuring convergence independently of the step size is available in the literature. Moreover, such a requirement is essential in several contexts such as deep learning [10], iterative learning control [11], [12], inverse map [13], observer design [14], and barrier functions [15], where the amplitude of the step size cannot be apriori fixed.…”

Section: Introductionmentioning

confidence: 99%