8,922 Hits in 4.6 sec

Differentiable MPC for End-to-end Planning and Control [article]

Brandon Amos, Ivan Dario Jimenez Rodriguez, Jacob Sacks, Byron Boots, J. Zico Kolter
2019 arXiv   pre-print
Using this strategy, we are able to learn the cost and dynamics of a controller via end-to-end learning.  ...  We present foundations for using Model Predictive Control (MPC) as a differentiable policy class for reinforcement learning in continuous state and action spaces.  ...  We thank Alfredo Canziani, Shane Gu, Yuval Tassa, and Chaoqi Wang for insightful discussions.  ... 
arXiv:1810.13400v3 fatcat:pqls2rb2tzdg7itpgi3argb2oe

MPC-Inspired Neural Network Policies for Sequential Decision Making [article]

Marcus Pereira, David D. Fan, Gabriel Nakajima An, Evangelos Theodorou
2018 arXiv   pre-print
Our results suggest that MPC-type recurrent policies have better robustness to disturbances and modeling error.  ...  We take advantage of this extension to show scalable and efficient training of complex planning policy architectures in continuous state and action spaces.  ...  PI-Net: End-to-end Differentiable Path Integral Control PI-Net is a fully differentiable model predictive control algorithm based on the Model Predictive Path Integral (MPPI) controller (Williams et  ... 
arXiv:1802.05803v2 fatcat:argkqsoccjezlct4pwgz7a5stq

Task-Driven Data Augmentation for Vision-Based Robotic Control [article]

Shubhankar Agarwal, Sandeep P. Chinchali
2022 arXiv   pre-print
To do so, we leverage differentiable MPC methods to calculate the sensitivity of a model-based controller to errors in state estimation, which in turn guides how we synthesize adversarial inputs.  ...  Today's robots often interface data-driven perception and planning models with classical model-based controllers.  ...  End-to-End Differentiable Architecture: Since every aforementioned block in Fig. 1 is differentiable, we have an end-to-end differentiable mapping from scene parameters ν to ultimate task cost J.  ... 
arXiv:2204.06173v2 fatcat:lscjgavtgveltoujgneinihzga

Gnu-RL: A Practical and Scalable Reinforcement Learning Solution for Building HVAC Control Using a Differentiable MPC Policy

Bingqing Chen, Zicheng Cai, Mario Bergés
2020 Frontiers in Built Environment  
Specifically, Gnu-RL adopts a recently-developed Differentiable Model Predictive Control (MPC) policy, which encodes domain knowledge on planning and system dynamics, making it both sample-efficient and  ...  Once it is put in charge of controlling the environment, the agent continues to improve its policy end-to-end, using a policy gradient algorithm.  ...  In the online learning phase, the agent continues to improve its policy end-to-end using a policy gradient algorithm. in this section we review both MPC and RL approaches to HVAC control.  ... 
doi:10.3389/fbuil.2020.562239 fatcat:mavxixol25hbxpp32uzxfkoi5u

Whole-Body MPC for a Dynamically Stable Mobile Manipulator [article]

Maria Vittoria Minniti, Farbod Farshidian, Ruben Grandia, Marco Hutter
2019 arXiv   pre-print
planner, and skillfully planning for end-effector contact forces.  ...  The optimization is performed using a Model Predictive Control (MPC) approach; the optimal control problem is transcribed at the end-effector space, treating the position and orientation tasks in the MPC  ...  To this end, another novelty is the addition of end-effector motion control tasks in the optimal control problem formulation and the planning for contact forces.  ... 
arXiv:1902.10415v2 fatcat:tkmqvlcuuvcubkurtv4j47gv4e

3D Online Path Planning of UAV Based on Improved Differential Evolution and Model Predictive Control

Jia Liu, Xiaolin Qin, Baolian Qi, Xiaoli Cui
2020 International Journal of Innovative Computing, Information and Control  
The algorithm integrates model predictive control (MPC) and differential evolution (DE) as the planning strategy.  ...  Then, the improved differential evolution algorithm based on the theory of MPC, is developed to optimize the objective function to find the optimal path.  ...  The authors also gratefully acknowledge the helpful comments and suggestions of the reviewers, which have improved the presentation.  ... 
doi:10.24507/ijicic.16.01.315 fatcat:7l3gwulgyfchdk2gm47z5jchxa

MPC-based Imitation Learning for Safe and Human-like Autonomous Driving [article]

Flavia Sofia Acerbo, Jan Swevers, Tinne Tuytelaars, Tong Duy Son
2022 arXiv   pre-print
With this strategy, IL is performed in closed-loop and end-to-end, through parameters in the MPC cost, model or constraints.  ...  This work suggests the use of a seamless combination of the two techniques to learn safe AV controllers from demonstrations of desired driving behaviors, by using MPC as a differentiable control layer  ...  Acknowledgements This work was supported by the Flemish Agency for Innovation and Entrepreneurship (VLAIO) in the context of MIMIC (huMan IMItation for autonomous driving Comfort) Baekeland Mandaat [Project  ... 
arXiv:2206.12348v1 fatcat:226h7m4hkff7dptawx4j4x3fxy

High-Frequency Nonlinear Model Predictive Control of a Manipulator

Sebastien Kleff, Avadesh Meduri, Rohan Budhiraja, Nicolas Mansard, Ludovic Righetti
2021 2021 IEEE International Conference on Robotics and Automation (ICRA)  
Model Predictive Control (MPC) promises to endow robots with enough reactivity to perform complex tasks in dynamic environments by frequently updating their motion plan based on measurements.  ...  Our controller leverages a state-of-the art optimal control solver, namely Differential Dynamic Programming (DDP), in order to replan state and control trajectories at real-time rates (1kHz).  ...  Figure 5 shows the end-effector trajectories and tracking errors between MPC predictions and measurements for both controllers.  ... 
doi:10.1109/icra48506.2021.9560990 fatcat:7qnzupw7pvh55olgyczuxqcvxm

Active and Passive Control of Walk-Assist Robot for Outdoor Guidance

Chun-Hsu Ko, Kuu-Young Young, Yi-Che Huang, Sunil Kumar Agrawal
2013 IEEE/ASME transactions on mechatronics  
The theory of differential flatness is used to plan the trajectory of control gains within the proposed scheme of the controller.  ...  Index Terms-Differential flatness, model predictive control (MPC), slope guidance, walk-assist robot.  ...  The mean time needed for trajectory planning in MPC was only about 0.2 ms with 1.66 GHz CPU and 1 GB memory.  ... 
doi:10.1109/tmech.2012.2201736 fatcat:krrssekc4ngvdguxswunw7lrva

Context-aware robotic arm using fast embedded model predictive control

Shane Trimble, Wasif Naeem, Sean McLoone, Pantelis Sopasakis
2020 2020 31st Irish Signals and Systems Conference (ISSC)  
for designing high performace MPC controllers.  ...  Model predictive control The robot's objective is for its end effector to reach a certain position and orientation and, in some cases, a certain (angular) velocity, [37] .  ... 
doi:10.1109/issc49989.2020.9180217 fatcat:3shrhh5cbnepncxy4zqoviltqm

High-Accuracy Model-Based Reinforcement Learning, a Survey [article]

Aske Plaat and Walter Kosters and Mike Preuss
2021 arXiv   pre-print
In recent years, a diverse landscape of model-based methods has been introduced to improve model accuracy, using methods such as uncertainty modeling, model-predictive control, latent models, and end-to-end  ...  learning and planning.  ...  Acknowledgments We thank the members of the Leiden Reinforcement Learning Group, and especially Thomas Moerland and Mike Huisman, for many discussions and insights.  ... 
arXiv:2107.08241v1 fatcat:tma6xb2uy5fybjfhmzasfx2cta

Path Integral Networks: End-to-End Differentiable Optimal Control [article]

Masashi Okada, Luca Rigazio, Takenobu Aoshima
2017 arXiv   pre-print
PI-Net is fully differentiable, learning both dynamics and cost models end-to-end by back-propagation and stochastic gradient descent. Because of this, PI-Net can learn to plan.  ...  The network includes both system dynamics and cost models, used for optimal control based planning.  ...  The freezed PI-Net was not trained end-to-end. VIN settings VIN was designed to have 2D inputs for continuous states and 1D output for continuous control.  ... 
arXiv:1706.09597v1 fatcat:jbdgo2k2nbg45mel6x4fprpsky

A Fully-Integrated Sensing and Control System for High-Accuracy Mobile Robotic Building Construction [article]

Abel Gawel, Hermann Blum, Johannes Pankert, Koen Krämer, Luca Bartolomei, Selen Ercan, Farbod Farshidian, Margarita Chli, Fabio Gramazio, Roland Siegwart, Marco Hutter, Timothy Sandy
2019 arXiv   pre-print
The approach leverages multi-modal sensing capabilities for state estimation, tight integration with digital building models, and integrated trajectory planning and whole-body motion control.  ...  We present a fully-integrated sensing and control system which enables mobile manipulator robots to execute building tasks with millimeter-scale accuracy on building construction sites.  ...  Two separate motion controllers are used for tracking the whole-body MPC motion references and for precisely tracking end-effector pose references when the base is stationary.  ... 
arXiv:1912.01870v1 fatcat:mivdjmadfzc4nk4v4jbs5vr2hu

Pontryagin Differentiable Programming: An End-to-End Learning and Control Framework [article]

Wanxin Jin, Zhaoran Wang, Zhuoran Yang, Shaoshuai Mou
2021 arXiv   pre-print
with respect to tunable parameters within an optimal control system, enabling end-to-end learning of dynamics, policies, or/and control objective functions; and second, we propose an auxiliary control  ...  This paper develops a Pontryagin Differentiable Programming (PDP) methodology, which establishes a unified framework to solve a broad class of learning and control tasks.  ...  Acknowledgments and Disclosure of Funding We acknowledge support for this research from Northrop Grumman Mission Systems' University Research Program.  ... 
arXiv:1912.12970v5 fatcat:7sse7c3bvzf7tjyu2smubzh3ri

Haptic Teleoperation of High-dimensional Robotic Systems Using a Feedback MPC Framework [article]

Jin Cheng, Firas Abi-Farraj, Farbod Farshidian, Marco Hutter
2022 arXiv   pre-print
In particular, we employ a feedback MPC approach and exploit its structure to account for the operator input at a fast rate which is independent of the update rate of the MPC loop itself.  ...  This work presents a novel framework for transparent teleoperation of MPC-controlled complex robotic systems.  ...  This plan is then passed to the whole-body controller, where it tracks the plan until a new MPC output is available.  ... 
arXiv:2207.14635v1 fatcat:s7zmwv6udjdkdhwpemo6u7hv5y
« Previous Showing results 1 — 15 out of 8,922 results