Autonomous Vehicle Prototype Development and Navigation using ROS
Nehal Borole · International Journal for Research in Applied Science and Engineering Technology · 2019
This paper describes an approach to the development of an autonomous vehicle, which can navigate in indoor environments like warehouses and assembly lines.Specifically, the paper focuses on phases from prototyping the vehicle to developing algorithms for navigation using the ROS framework.Along with the autonomous mode, the vehicle has a semiautonomous mode through which the vehicle can be controlled remotely using a PlayStation 2 controller connected to the host system.The semi-autonomous relies on a client system (for teleoperation) to which the PlayStation controller is connected.The ROS implementation for autonomous navigation where different nodes come together and work in sync for the task of autonomous navigation will be explained as well.I. INTRODUCTION Autonomy has been of utmost importance in places where the roles of humans to handle certain tasks are either risky or timeconsuming.From disposing of nuclear waste to search and rescue operations, autonomous robots have played a crucial role.Until now, tremendous progress has been made in the field of autonomous indoor navigation, but some approaches assume the structural parts of the environment to be completely static.The following autonomous vehicle (AV) takes into consideration the dynamic changes in the environment by utilizing sensory data about the environment and vehicle states, and perform localization, dynamic obstacle recognition, and optimal trajectory computation.The AV implementation is carried out in two phases: (1) semiautonomous mode; (2) autonomous mode.II.HARDWARE USED A. Nvidia Jetson TK1 B. Kinect sensor C. Four BS48H motors along with Roboteq SBL1310 motor controller with Hall effect encoders D. PlayStation 2 Controller E. PC (client system) F. Arduino G. GY-80 Multi sensor board H. Wireless USB adapter I. USB to RS232 converter III.VEHICLE DEVELOPMENT As the first step of the AV, the model of the vehicle was developed in Solid Works.The vehicle is fitted with on-board computer and sensors.As a computational platform, Nvidia Jetson TK1 embedded computer has been used, supporting CUDA with on-board parallel computation.The vehicle is mounted with a Kinect sensor, which provides depth images that are helpful to segment objects based on depth rather than intensity.In order to visualize, the vehicle in action in a 3D environment, a ROS package named Rviz is used.Using sw_urdf_exporter plugin in Solid Works a URDF file is generated, which is used by Rviz to generate the vehicle.Fig 3.1 Model of AV in Solid Works.