Human Robot Interaction for Hybrid Collision Avoidance System for Indoor Mobile Robots

Abstract

In this paper, a novel approach for collision avoidance for indoor mobile robots based on human-robot interaction is realized. The main contribution of this work is a new technique for collision avoidance by engaging the human and the robot in generating new collisionfree paths. In mobile robotics, collision avoidance is critical for the success of the robots in implementing their tasks, especially when the robots navigate in crowded and dynamic environments, which include humans. Traditional collision avoidance methods deal with the human as a dynamic obstacle, without taking into consideration that the human will also try to avoid the robot, and this causes the people and the robot to get confused, especially in crowded social places such as restaurants, hospitals, and laboratories. To avoid such scenarios, a reactive-supervised collision avoidance system for mobile robots based on human-robot interaction is implemented. In this method, both the robot and the human will collaborate in generating the collision avoidance via interaction. The person will notify the robot about the avoidance direction via interaction, and the robot will search for the optimal collision-free path on the selected direction. In case that no people interacted with the robot, it will select the navigation path autonomously and select the path that is closest to the goal location. The humans will interact with the robot using gesture recognition and Kinect sensor. To build the gesture recognition system, two models were used to classify these gestures, the first model is Back-Propagation Neural Network (BPNN), and the second model is Support Vector Machine (SVM). Furthermore, a novel collision avoidance system for avoiding the obstacles is implemented and integrated with the HRI system. The system is tested on H20 robot from DrRobot Company (Canada) and a set of experiments were implemented to report the performance of the system in interacting with the human and avoiding collisions.

Visual recognition of gymnastic exercise sequences. Application to supervision and robot learning by demonstration

“…Skeleton tracking has been widely used in robotics. A remarkable work was presented by Ghandour et al [16], who created a Human Robot Interaction system (HRI) to give indications to mobile robots to avoid collisions with people. They used the skeleton returned by the sensor Kinect 2.0, consisting of 25 joints.…”

Embedded Artificial Neural Network FPGA Controlled Cart

“…The method to achieve the task of following might vary in the hardware or sensor used. Ghandour et al [9] came out with a collision avoidance mobile robot based on human interaction. Kinect sensor with motion identification became an interface for humans to interact with.…”