Development of a Human Tracking Indoor Mobile Robot Platform

Gürkan Küçükyıldız, Suat Karakaya

In this paper, a differential drive mobile robot platform was developed in order to perform indoor mobile robot researches. The mobile robot was localized and remote controlled. The remote control consists of a pair of 2.4 GHz transceivers. Localization system was developed by using infra­red reflectors, infrared leds and camera system. Real time localization system was run on an industrial computer placed on the mobile robot. The localization data of the mobile robot is transmitted by a UDP communication program. The transmitted localization information can be received any computer or any other UDP device. In addition, a LIDAR (Light Detection and Ranging; or Laser Imaging Detection and Ranging) and a Kinect three­dimensional depth sensor were adapted on the mobile robot platform. LIDAR was used for obstacle and heading direction detection operations and Kinect for eliminating depth data of close environment. In this study, a mobile robot platform which has specialties as mentioned was developed and a human tracking application was realized real time in MATLAB and C# environment.

Obstacle and Optimal Heading Direction Detection Algorithm On a Mobile Robot Platform

Gürkan Küçükyıldız, Suat Karakaya

In this study, Sick­LMS100 Lidar was used for detecting the obstacles around a mobile robot platform and finding the best heading direction. The computer and the LIDAR were communicated via Ethernet TCP/IP in order to gather position information of the objects around. The algorithm, which was developed in Visual Basic 6.0 environment, chose the optimal heading direction relative to the positions of the obstacles. The gathered path information was then sent to a DSP for motor control via serial port. A mobile robot platform was developed during the study and the optimum heading direction finding algorithm was tested on this mobile robot platform in real time. The results which were gathered in several conditions were compared.

Kinematic Model Based Path Tracking Algorithm for Differential Drive Mobile Robots

Gürkan Küçükyıldız, Suat Karakaya

In this study, it was studied on a path tracking method which is based on fuzzy logic, PI and P control for 4­ wheeled differential drive autonomous mobile robots. Major problem is to force the mobile robot which is assumed to be located on a static map, to track a path that was planned by planning algorithms on the same map. Therefore, a mobile robot simulator was developed regarding a real mobile robot’s mechanical and physical specs. The developed method was tested on this simulator by using the control algorithms. Performance criterions were given as the length of the route taken by the robot and tracking duration

A Hybrid Indoor Localization System Based on Infra-red Imaging and Odometry

Gürkan Küçükyıldız, Suat Karakaya

In this study, a real-time indoor localization system was developed by using a camera and passive landmarks. A narrow band-pass infra-red (IR) filter was inserted to the back of the camera lens for capturing IR images. The passive landmarks were placed on the ceiling at pre-determined locations and consist of IR retro-reflective tags that have binary coded unique ID’s. An IR projector emits IR rays at the tags on the ceiling. The tags then reflect the rays back to the camera sensor creating a digital image. An image processing algorithm was developed to detect and decode the landmarks in captured images. The proposed algorithm successfully estimates the position and the orientation angle based on relative position and orientation with respect to the detected tags. To further improve the accuracy of the estimates, extended Kalman filter (EKF) was adapted to the measurement algorithm. The proposed method initially estimates the position of a mobile robot based on odometry and kinematic model. EKF was then used to update the estimates given the measurement obtained from the image processing system. Real time experiments were performed to test the performance of the system. The results prove that the proposed indoor localization system can effectively estimate position with an error less than 5cm.

A Bug-Based Local Path Planning Method for Static and Dynamic Environments

Gürkan Küçükyıldız, Suat Karakaya

In this study, a local path planning method was proposed for both static and dynamic environments. In obstacle-free cases, the mobile robot was forced to basic motion-togoal movement. In case where direct movement towards the global target is not possible, the algorithm searches for possible gaps which satisfy certain clearance criteria. The gaps were detected by taking the gradient of one dimensional distance vector acquired from the SICK LMS100 Light Detection and Ranging (LIDAR) sensor. The detected gaps were filtered by various methods which finally led to the optimal gap. Points on the line passing through the optimal gap were evaluated through a cost function and the point having the minimum cost was assumed to be the current local target. The points which were close to the two opposite corners of the gap less than a certain threshold were discarded to avoid collision. The threshold was determined based on the robot size and the kinematic model. Proportional and integral (PI) speed controller for left and right steered wheels was adapted to the proposed method. A graphical user interface (GUI) was developed to visualize the outputs of the method. On the GUI, offline LMS100 vectors and location data were visualized considering differential drive kinematic constraints for the mobile robot. The algorithm was developed at MATLAB environment.

A Hybrid Posture Stabilization Method for Mobile Robots

Gürkan Küçükyıldız, Suat Karakaya

In this study, a point stabilization scheme which takes into account static obstacles around wheeled mobile robots is proposed. Novelty of the algorithm lies under the consideration of the static obstacles and corporation with the static path planning method exact Euclidian distance transform (EEDT). For a given start point, goal point and static obstacle configuration, the EEDT algorithm determines the shortest path. This path is in an open-polygonal form due to the robot’s grid-based workspace. Tangent values of each vertex of the open-polygon are given to conventional model prediction control (MPC) based posture stabilization scheme as sub-start and sub-goal points. These sub points are given to MPC in a shiftedhorizon strategy to determine the stabilized trajectories between the vertex coordinates. Overall stabilized static trajectory is determined by combining the sub-trajectories independently calculated by MPC based posture stabilization algorithm. The experimental results which are performed in a 3D virtual reality interface, confirms that the developed scheme satisfies the posture stabilization criterion successfully in presence of static obstacles.

Detection of Obstacle-Free Gaps for Mobile Robot Applications Using 2-D LIDAR Data

Suat Karakaya

Mobile robotics is one of the most studied scientific and technological fields, which is still in progress. Several research interests such as path planning, point stabilization, localization, obstacle avoidance and passable gap detection are commonly studied fields. Gap detection task affects the path planning characteristics of a mobile robot. Especially under presence of limited information about robot’s environment, passable gap detection is necessary for steering the mobile robot towards a goal autonomously. This paper concentrates on passable gap detection for unconstructed environments, which contain only positive obstacles. The method considers specific obstacle configurations such as presence of wall-type obstacle, maze type environments and random placed small sized obstacles. The method proposed in this study is based on reading distance of the obstacles in a certain range and detecting the borders of passable gaps. The detected gaps are re-organized depending on the priority assigned by the robot’s passage order of the gaps. The proposed scheme not only utilizes simple derivation of the measurement data but also extracts hidden gaps in the environment. The proposed scheme assumes the mobile robot is equipped with laser range sensor (LIDAR). A real LIDAR is modelled and adapted to the developed algorithm. The algorithm was developed in Matlab.

Speed Control Adaptation On Static Trajectories

Suat Karakaya

In this study, a speed vector is defined for static trajectories for mobile robots. In many conventional path-planning methods, the major criterion is to plan the trajectory through obstacle-free regions to satisfy safety. The obstacle configuration is not the main concern for controlling the speed of the mobile robot. This issue is handled as a sub-procedure under path tracking scheme rather than a standalone operation. Thus, it is performed that planning a speed vector within the planned static trajectory. Directory lines are fitted on consecutive path coordinates to check whether an obstacle is available on the motion direction of the mobile robot or not. This procedure is operated continuously to control the instant average speed of the mobile robot.

Brain Computer Interface with Low Cost Commercial EEG Device

Gürkan Küçükyıldız, Suat Karakaya

In this study, a brain computer interface (BCI) system was explored.  Instead of high cost EEG devices, a low cost commercial EEG device (EMOTIV) was used. Raw EEG data was obtained by using research edition SDK of EMOTIV.  EMOTIV EEG device has 14 channels (10-20 placement) for EEG and two channels (x and y axis gyro: GYROX, GYROY) for head movements.  Head movements and eye-blink can affect the EEG data and are usually referred to as artifacts. In this study, raw EEG data was pre-processed using the x and the y axis gyro data and the two front EEG channels, namely AF4, F8, in order to determine whether the data is artifact free or not. EEG data was collected from subjects that were asked to accomplish two cognitive tasks: pushing a cube and relaxing. Subjects performed each task for a duration of five seconds during 20 trials. The acquired EEG data was divided into 0.25 second epochs. Epochs that were determined to have artifacts were discarded. Power spectral density (PSD) and time domain based features were extracted from artifact free epochs. The features were then used to train a Support Vector Machine (SVM) to determine the corresponding task. The performance of the SVM classifier was compared to that of an Artifical Neural Network (ANN) based one. Experimental results show the efficacy of the SVM based scheme.

Vision Based Control of Magnetic Levitation System

Gürkan Küçükyıldız

This work presents vision based control of a single-axis magnetic levitation system. The  system is the fundamental of the high speed maglev trains and magnetic bearings. A ferro magnetic object is levitated at a desired position in the air gap by applying electromagnetic  force against to gravity. The system consists of five main components: a position sensor and a camera, a coil, a controller, a driver and a ferro-magnetic object. Current on the coil, which causes electromagnetic force, was controlled according to position feedback. The mathematical model of the system was obtained based on Newton’s theory and verified by experimental data. The nonlinear relationships between force, current and air gap were found experimentally. The controller was designed using feedback linearization technique based on the nonlinear relationships. In the literature, light based sensors have mostly been used to detect the object’s position. Despite the preferred usage of conventional sensors, some disadvantages are encountered such as: calibration requirement, non-linearity, noise and non robustness. These disadvantages inevitably cause disturbances on the system. In this study, the position of the magnetic object was detected using image processing methods in order to overcome the disadvantages associated with the light based sensors.

Copyright © 2017

Romeda Bilgi Teknolojileri Ltd. Şti.