International Journal of Advanced Research in Electrical Electronics and Instrumentation Engineering

editor@imedpub.com

Abstract

Safety of Human-Robot Interaction Based on Neural Networks and Using Intrinsic Joint Sensors

Human-robot cooperation (HRC) is the field that aims to perform a complementary combination between robot abilities and human skills. Robots can assist humans by increasing their capabilities in precision, speed, and force. Humans could contribute to cooperation in terms of experience, knowledge of executing a task, easy learning, adaptation, and easy understanding of control strategies. Safety is the most necessary stage during this collaboration between the human and the robot since the proximity of the operator to the robotic manipulator may lead to the possibility of injuries.

In this work, an approach is proposed depending on the neural network (NN) training for detecting the human-robot collisions. The NN is designed taking into consideration the manipulator dynamics and trained, using data with and without collisions, by the algorithm of Levenberg-Marquardt for detecting the collisions that occur between the human and the manipulator. Prior knowledge to the dynamic model is not necessary and required. Three NN types are investigated:
  • One multilayer feedforward NN (MLFFNN-1) architecture, having one hidden layer, is implemented using the sensors of the joint position and the joint torque of the robot. Therefore, this architecture can be applied to robots that have torque sensors. 
  • Other MLFFNN architecture with two-hidden layers (MLFFNN-2), cascaded forward NN (CFNN), and recurrent NN (RNN) are implemented using only the intrinsic joint position sensors of the manipulator. As a result, these architectures could be used with any robot. MLFFNN-2 is investigated experimentally to 1-DOF and 2-DOF manipulators, whereas CFNN and RNN are investigated experimentally to 1-DOF manipulators.
However, all these architectures can be applied with all robot joints. ISO standards are considered for confirming that the detection of the collisions using the introduced method are in the safe region of the human-robot collaboration. The results show that the developed system can detect the collisions effectively.

Author(s): Abdel-Nasser Sharkawy

Abstract | Full-Text | PDF

Share This Article
Awards Nomination 17+ Million Readerbase
Google Scholar citation report
Citations : 205

International Journal of Advanced Research in Electrical Electronics and Instrumentation Engineering received 205 citations as per Google Scholar report

image
Open Access Journals