RoKON: Efficient and Adaptive Data-Enabled Predictive Control for Robotic Manipulation

Akshun Sharma; Zhongqi Wei; Buqing Ou; Frederike Dümbgen

cs.ROOct 26, 2026

RoKON: Efficient and Adaptive Data-Enabled Predictive Control for Robotic Manipulation

Authors: Akshun Sharma, Zhongqi Wei, Buqing Ou, Frederike Dümbgen

Organizations: Maryland Applied Graduate Engineering, University of Maryland, College Park, MD 20742, USA · Department of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, PA 15213, USA

Abstract

Data-enabled predictive control has recently emerged as a promising control paradigm, enabling a transparent optimal-control formulation without the need to learn an explicit dynamics model. This paper presents a number of extensions paving the path towards its use in real-time robotics tasks. Following recent work, we utilize a nonlinear predictor based on Gaussian kernels to represent the relationship between input and output data. We propose a damped Gauss-Newton method that enables real-time control by efficiently exploiting the least-squares structure of the control problem. In addition, we introduce a hybrid buffer management strategy that incorporates informative online data while retaining relevant offline data, allowing the predictor to account for changes in the dynamics. We demonstrate the tracking performance and computational suitability of the proposed method in MuJoCo using a Franka Research 3 arm pushing a block across surface, without prior knowledge of its changing friction values.

Figures & tables

Fig. 1: Image of the setup on MuJoCo. An FR3 arm pushing a block across three variable friction surfaces.

Panel of Figure 2 — Fig. 2: Closed-loop velocity tracking performance on uniform-friction surfaces for $\mu\in\{0.08,\,0.25,\,0.50\}$ and $y^{\mathrm{ref}}\in\{0.05,\,0.10\}\,\mathrm{m/s}$ .

Method	Surface RMSE (m/s)			Complete-run RMSE	$H_{c}$ columns	Mean time (ms)
	$\mu=0.2$	$\mu=0.5$	$\mu=0.25$	(m/s)
Stribeck MPC	0.00370	0.01909	–	0.01766	–	7.20
Linear DeePC	0.00709	0.00721	0.02502	0.01331	294	2.76
RoKDeePC	0.00113	0.01703	–	0.01552	294	14.44
Windowed DeePC	0.01496	0.00490	0.03687	0.02029	294–350	4.37
RoKON (Ours)	0.00202	0.00308	0.01084	0.00602	294–350	3.81

TABLE I: Varying Friction Experiment Tracking Accuracy and Computation Time

Fig. 3: Ablation study of the proposed RoKON method across the three friction surfaces. The bars report the velocity-tracking RMSE on each surface and over the complete evaluated trajectory. N/R indicates that the block did not reach the corresponding surface.

Data-Driven Control

Data-driven control uses measured data and learned models or policies to control systems when accurate first-principles models are unavailable.

See all 36 papers

Adapting Neural Robot Dynamics on the Fly for Predictive Control
Apr 5, 2026Abdullah Altawaitan, Nikolay AtanasovRobot Dynamics Learning Robot Dynamics Identification
Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA 92093, USA · Kuwait University
Deep Reinforcement Learning-Enhanced Event-Triggered Data-Driven Predictive Control for a 3D Cable-Driven Soft Robotic Arm
Jun 24, 2026Cheng Ouyang, Moeen Ul Islam, Kaixiang Zhang +3RL Control Robot Trajectory Tracking
Department of Agricultural & Biological Engineering, Mississippi State University, MS, USA · Department of Mechanical Engineering, Michigan State University, MI, USA · Department of Electrical and Computer Engineering, Michigan State University, MI, USA
RK-MPC: Residual Koopman Model Predictive Control for Quadruped Locomotion in Offroad Environments
Apr 5, 2026Sriram S. K. S. Narayanan, Umesh VaidyaModel Predictive Control Quadruped Locomotion
Department of Mechanical Engineering, Clemson University, Clemson, SC 29630, USA

Apr 5, 2026J/K move · Enter open · S save

Adapting Neural Robot Dynamics on the Fly for Predictive Control

Abdullah Altawaitan, Nikolay Atanasov

Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA 92093, USA · Kuwait University

Robot Dynamics Learning Robot Dynamics Identification Online System Identification

Open paper AskPDF arXiv

Accurate dynamics models are critical for the design of predictive controllers for autonomous mobile robots. Physics-based models are often too simple to capture relevant real-world effects, while data-driven models are data-intensive and slow to train. We introduce an approach for fast adaptation of neural robot dynamic models that combines offline training with efficient online updates. Our approach learns an incremental neural dynamics model offline and performs low-rank second-order parameter adaptation online, enabling rapid updates without full retraining. We demonstrate the approach on a real quadrotor robot, achieving robust predictive tracking control in novel operational conditions.

Abstract

Data-Driven Control

Explore similar work

Adapting Neural Robot Dynamics on the Fly for Predictive Control

Deep Reinforcement Learning-Enhanced Event-Triggered Data-Driven Predictive Control for a 3D Cable-Driven Soft Robotic Arm

RK-MPC: Residual Koopman Model Predictive Control for Quadruped Locomotion in Offroad Environments