An experimental validation of reinforcement learning applied to the position control of UAVs

2012 ◽  
Author(s):  
Sergio Ronaldo Barros dos Santos ◽  
Sidney N. Givigi ◽  
Cairo Lucio Nascimento Junior
Keyword(s):  
Reinforcement Learning ◽  
Experimental Validation ◽  
Position Control

A new hybrid control methodology for a morphing aircraft wing-tip actuation mechanism

2019 ◽  
Vol 123 (1269) ◽  
pp. 1757-1787
Author(s):  
M. J. Tchatchueng Kammegne ◽  
R. M. Botez
Keyword(s):  
Control System ◽  
Power Amplifier ◽  
Experimental Validation ◽  
Position Control ◽  
The Other ◽  
Feedback Signal ◽  
Aircraft Wing ◽  
Control Loops ◽  
Flexible Skin ◽  
Wing Tip

ABSTRACTThe focus of this paper is on the modelling of miniature electromechanical actuators used in a morphing wing application, on the development of a control concept for these actuators, and on the experimental validation of the designed control system integrated in the morphing wing-tip model for a real aircraft. The assembled actuator includes as its main component a brushless direct current motor coupled to a trapezoidal screw by using a gearing system. A Linear Variable Differential Transformer (LVDT) is attached on each actuator giving back the actuator position in millimetres for the control system, while an encoder placed inside the motor provides the position of the motor shaft. Two actuation lines, each with two actuators, are integrated inside the wing model to change its shape. For the experimental model, a full-scaled portion of an aircraft wing tip is used with the chord length of 1.5 meters and equipped on the upper surface with a flexible skin made of composite fibre materials. A controllable voltage provided by a power amplifier is used to drive the actuator system. In this way, three control loops are designed and implemented, one to control the torque and the other two to control the position in a parallel architecture. The parallel position control loops use feedback signals from different sources. For the first position control loop, the feedback signal is provided by the integrated encoder, while for the second one, the feedback signal comes from the LVDT. For the experimental model, the parameters for the torque control, but also for the position control-based encoder signal, are implemented in the power amplifier energising the electrical motor. On the other hand, a National Instruments real-time system is used to implement and test the position control-based LVDT signal. The experimental validation of the developed control system is realised in two independent steps: bench testing with no airflow and wind-tunnel testing. The pressure data provided by a number of Kulite sensors equipping the flexible skin upper surface and the infrared thermography camera visualisations are used to estimate the laminar-to-turbulent transition point position.

scholarly journals Human–robot interactive control based on reinforcement learning for gait rehabilitation training robot

2019 ◽  
Vol 16 (2) ◽  
pp. 172988141983958 ◽  
Author(s):  
Guo Bingjing ◽  
Han Jianhai ◽  
Li Xiangpan ◽  
Yan Lin
Keyword(s):  
Reinforcement Learning ◽  
Lower Limb ◽  
Position Control ◽  
Human Robot Interaction ◽  
Experimental Results ◽  
Gait Rehabilitation ◽  
Sigmoid Function ◽  
Interactive Control ◽  
Rehabilitation Training ◽  
Active Rehabilitation

A human–robot interactive control is proposed to govern the assistance provided by a lower limb exoskeleton robot to patients in the gait rehabilitation training. The rehabilitation training robot with two lower limb exoskeletons is driven by the pneumatic proportional servo system and has two rotational degrees of freedom of each lower limb. An adaptive admittance model is adopted considering its suitability for human–robot interaction. The adaptive law of the admittance parameters is designed with Sigmoid function and the reinforcement learning algorithm. Individualized admittance parameters suitable for patients are obtained by reinforcement learning. Experiments in passive and active rehabilitation training modes were carried out to verify the proposed control method. The passive rehabilitation training experimental results verify the effectiveness of the inner-loop position control strategy, which can meet the demands of gait tracking accuracy in rehabilitation training. The active rehabilitation training experimental results demonstrate that the personal adaption and active compliance are provided by the interactive controller in the robot-assistance for patients. The combined effects of flexibility of pneumatic actuators and compliance provided by the controller contribute to the training comfort, safety, and therapeutic outcome in the gait rehabilitation.

Design and experimental validation of doubly salient permanent magnet linear synchronous motor for precision position control

Mechatronics ◽  
2013 ◽  
Vol 23 (2) ◽  
pp. 172-181 ◽  
Author(s):  
Shi-Uk Chung ◽  
Ji-Won Kim ◽  
Byung-Chul Woo ◽  
Do-Kwan Hong ◽  
Ji-Young Lee ◽  
...  
Keyword(s):  
Permanent Magnet ◽  
Experimental Validation ◽  
Position Control ◽  
Synchronous Motor ◽  
Linear Synchronous Motor ◽  
Doubly Salient

scholarly journals Decentralized Coordinated Motion Control of Two Hydraulic Actuators Handling a Common Object

2007 ◽  
Vol 129 (5) ◽  
pp. 729-741 ◽  
Author(s):  
Mark Karpenko ◽  
Nariman Sepehri ◽  
John Anderson
Keyword(s):  
Reinforcement Learning ◽  
Position Control ◽  
Control Strategies ◽  
Interaction Force ◽  
Learning System ◽  
Multiagent Learning ◽  
Credit Assignment ◽  
Interaction Forces ◽  
Actuator Dynamics ◽  
Hydraulic Actuators

In this paper, reinforcement learning is applied to coordinate, in a decentralized fashion, the motions of a pair of hydraulic actuators whose task is to firmly hold and move an object along a specified trajectory under conventional position control. The learning goal is to reduce the interaction forces acting on the object that arise due to inevitable positioning errors resulting from the imperfect closed-loop actuator dynamics. Each actuator is therefore outfitted with a reinforcement learning neural network that modifies a centrally planned formation constrained position trajectory in response to the locally measured interaction force. It is shown that the actuators, which form a multiagent learning system, can learn decentralized control strategies that reduce the object interaction forces and thus greatly improve their coordination on the manipulation task. However, the problem of credit assignment, a common difficulty in multiagent learning systems, prevents the actuators from learning control strategies where each actuator contributes equally to reducing the interaction force. This problem is resolved in this paper via the periodic communication of limited local state information between the reinforcement learning actuators. Using both simulations and experiments, this paper examines some of the issues pertaining to learning in dynamic multiagent environments and establishes reinforcement learning as a potential technique for coordinating several nonlinear hydraulic manipulators performing a common task.

scholarly journals Position control of a mobile robot using reinforcement learning

2020 ◽  
Vol 53 (2) ◽  
pp. 17393-17398
Author(s):  
G. Farias ◽  
G. Garcia ◽  
G. Montenegro ◽  
E. Fabregas ◽  
S. Dormido-Canto ◽  
...  
Keyword(s):  
Reinforcement Learning ◽  
Mobile Robot ◽  
Position Control
Sign in / Sign up

Export Citation Format

Share Document