Designing Decentralized Controllers for Distributed-Air-Jet MEMS-Based Micromanipulators by Reinforcement Learning

  title={Designing Decentralized Controllers for Distributed-Air-Jet MEMS-Based Micromanipulators by Reinforcement Learning},
  author={La{\"e}titia Matignon and Guillaume J. Laurent and Nadine Le Fort-Piat and Yves-Andr{\'e} Chapuis},
  journal={Journal of Intelligent \& Robotic Systems},
Distributed-air-jet MEMS-based systems have been proposed to manipulate small parts with high velocities and without any friction problems. The control of such distributed systems is very challenging and usual approaches for contact arrayed system don’t produce satisfactory results. In this paper, we investigate reinforcement learning control approaches in order to position and convey an object. Reinforcement learning is a popular approach to find controllers that are tailored exactly to the… 
Multi-Agent Actor-Critic Reinforcement Learning for Cooperative Tasks
The aim of this project is to implement actor-critic RL methods to cooperative MAS to combine the advantages of these two approaches and apply the resulting methods to a real-life control problem as a proof of concept.
Distributed control architecture for smart surfaces
A distributed control architecture to perform part recognition and closed-loop control of a distributed manipulation device based on decentralized cells able to communicate with their four neighbors thanks to peer-to-peer links is presented.
A survey of non-prehensile pneumatic manipulation surfaces: principles, models and control
A comprehensive survey of state-of-the-art pneumatic manipulation from the macro scale to the micro scale with a particular emphasis on the modeling and the control issues is discussed.
Distributed Intelligent MEMS: Progresses and Perspectives
The objective of this paper is to report on the progress and remaining challenges in these areas by examining two examples of embedded intelligence in MEMS systems.
Distributed Intelligent MEMS
A real-time programming framework that can be used to design new real- time programming languages for DiMEMSs is proposed and its design challenges and requirements are investigated.
Decentralized Reinforcement Learning of Robot Behaviors
Decentralized Learning for Optimality in Stochastic Dynamic Teams and Games with Local Control and Global State Information
The results presented here are the first, to the authors' knowledge, to give formal guarantees of convergence to team optimality using independent learners in stochastic dynamic teams and common interest games.
A New Aerodynamic-Traction Principle for Handling Products on an Air Cushion
A new aerodynamic-traction principle to handle delicate and clean products, such as silicon wafers, glass sheets, or flat foodstuffs, to meet the requirements for industrial applications is introduced.
2 . 1 Software challenges , scaling up
The objective of this article is to report the progresses made by taking the example of two research projects and by giving the remaining challenges and the perspectives of distributed intelligent MEMS.
Review on Contactless Transport Using Pneumatic Levitation
: Many industries, such as semiconductor, food and pharmaceuticals, require contactless transport of delicate or clean products during production process. The traditional contact method cannot


Hybrid Dynamic Control Algorithm for Humanoid Robots Based on Reinforcement Learning
In this paper, hybrid integrated dynamic control algorithm for humanoid locomotion mechanism is presented. The proposed structure of controller involves two feedback loops: model-based dynamic
Design, fabrication, and control of MEMS-based actuator arrays for air-flow distributed micromanipulation
This paper reports the design, fabrication and control of arrayed microelectromechanical systems (MEMS)-based actuators for distributed micromanipulation by generation and control of an air-flow
FPGA-Based Decentralized Control of Arrayed MEMS for Microrobotic Application
A decentralized decision-making strategy using field-programmable gate array (FPGA) technology as a prototype for an integrated controller of a microelectromechanical systems (MEMS) array for air-flow planar micromanipulation.
Fuzzy Policy Reinforcement Learning in Cooperative Multi-robot Systems
A multi-agent reinforcement learning algorithm with fuzzy policy used to deal with some control problems in cooperative multi-robot systems and results demonstrate that the control performance can be improved after the learning.
A machine-learning approach to multi-robot coordination
Sensorless manipulation using massively parallel microfabricated actuator arrays
A geometric model for the mechanics of microactuators and a theory of sensorless, parallel manipulation and efficient algorithms for their evaluation are developed.
A conveyance system using air flow based on the concept of distributed micro motion systems
A microactuator array for a planar conveyance system on a plane has been designed and fabricated by micro-machining. Though a typical microactuator accomplishes only a simple motion, the coordination
Multi-robot Box-pushing: Single-Agent Q-Learning vs. Team Q-Learning
  • Y. Wang, C. D. Silva
  • Computer Science
    2006 IEEE/RSJ International Conference on Intelligent Robots and Systems
  • 2006
It is shown that the single-agent Q-learning algorithm does a better job than the team Q- learning algorithm in a complicated and unknown environment with many obstacles.
A planar air levitated electrostatic actuator system
  • K. Pister, R. Fearing, R. Howe
  • Engineering
    IEEE Proceedings on Micro Electro Mechanical Systems, An Investigation of Micro Structures, Sensors, Actuators, Machines and Robots.
  • 1990
The design and testing of a micromotor capable of moving multiple objects in three dimensions is discussed. The fixed surface of the motor, fabricated on a silicon wafer, contains air nozzles that
Dynamic correlation matrix based multi-Q learning for a multi-robot system
  • Hongliang Guo, Y. Meng
  • Computer Science
    2008 IEEE/RSJ International Conference on Intelligent Robots and Systems
  • 2008
A novel dynamic correlation matrix is proposed, which not only handles each agentpsilas Q value, but also deals with the correlation among agents, which shows the efficiency and convergence of the proposed DCM-MultiQ method.