Differential drive wheeled robot design and implementation for search and rescue

Robot technology has developed rapidly within many different areas of technological innovation. Some of these systems not only developed to reduce the daily load of the life but also designed for entertainment. Nowadays, robot technology is used in many different areas including military applications, biomedical systems, humanoid robots and small appliances. The search and rescue robot technology is another branch, attracting a lot of interest among researchers all over the world. It is obvious that using robots in disaster environments can provide information to the rescue teams in natural disasters or accidents where human intervention is difficult and inefficient. Various search and rescue situations where robots may serve include earthquakes, floods, accidents and radioactive disasters. Some of these robots specialized in a specific area while others are produced for general use. Different robots, designed to serve in different type of accidents and debris, have common needs. For example, each robot has to navigate through the environment, even though it uses different models. In the same way, each of them needs to obtain information about the environment and to recognize the environment in different levels. In this study, a differential drive wheeled robot platform is designed and implemented to respond common requirements identified for search and rescue operations. Hydro version of the ROS framework and existing libraries are used in this study to implement some algorithms essential for search and rescue robots. Drivers provided within ROS are taken into consideration during the design of the robot platform. Main requirements examined can be classified as robot navigation, pose estimation, simultaneous localization and mapping. The robot platform, developed for this study has a four-wheeled differential drive motion model. Each wheel may take a different speed commands. In this way, desired rotation can be provided. Laser_Scan_Matcher (LSM) method was used for location estimation. The method uses laser distance measurements to match consecutive laser scans. After that, the experimental results were collected. For mapping problem, the study focuses on gMapping method. Related experiments and the results obtained by this method are discussed. Another issue that was analyzed in the study is the robot's navigation. For that purpose, "navigation stack" package provided for ROS has been run and the results were examined. This package requires localization information of the robot. To obtain the localization information the Augmented Monte Carlo Localization (AMCL) method, which is an adaptive version of MCL, was used. In the experiments, the real robot platform is used and the results are tested with different parameters to optimize.