The paper considers the problem of autonomous navigation of unmanned ground vehicle and the way to solve it by using simultaneous localization and mapping methods based on the data, provided by the laser rangefinder, and path planning algorithms. We propose the control system's architecture (including low-level communication protocols and high-level planning and mapping mechanisms) and it's implementation based on Robot Operating System (ROS) framework. The system is planned to use as a toolbox for pathplanning algorithms evaluation on real robotic system in real environment. We also provide the visualization of current state of system. The evaluation is carried out on a Nexus wheeled robot, which specification is also given. Future work includes multi-robot modification of developed system (shared map over all users of the system), exploration algorithms implementation (including multi-agent exploration), multi-agent pathplanning algorithms embedding and moving the whole system's operation on board of mobile ground robotic system.
Copyright © 2018 for the individual papers by the papers’ authors. Copying permitted for private and academic purposes. This volume is published and copyrighted by its editors. In: K. E. Samouylov, L. A. Sevastianov, D. S. Kulyabov (eds.): Selected Papers of the 1st Workshop (Summer Session) in the framework of the Conference “Information and Telecommunication Technologies and Mathematical Modeling of High-Tech Systems”, Tampere, Finland, 20–23 August, 2018, published at http://ceur-ws.org
Nowadays unmanned ground vehicles (mobile robots) are widely used for academic
and practical purposes. The application area of such robots varies from being evaluation
testbeds for methods and algorithms developed in laboratory environment to real world
applications like search-and-rescue, monitoring, guarding etc. Increasing the autonomy
is one of the core tasks in modern robotics [
Diferent approaches to solve the navigation problem exist. Applicability of diferent approaches depends on the properties of the robot’s environment (known or unknown environment, availability of global position systems and etc.) and on-board sensors. Modern navigation methods can be split to reactive and deliberative. Reactive methods operate by handling "at-the-moment" sensors’ information and perform movement based on the current state of the system and the surrounding environment. This approach is mainly used for navigation in dynamically changing environment, e.g. for obstacle-avoiding tasks, path following etc. This approach is also useful when the time horizon is short and decisions should be made very quickly. Deliberative methods assume that the robot posses some information about the environment (e.g. it has the map) and performs the navigation taking into account this information. This class is represented by path planning algorithms, simultaneous localization and mapping (SLAM) algorithms etc. We follow the deliberative approach in this work.
The main tasks we consider are mapping of unknown environment, localization in
resultant (or known apriori) map and path planning. We follow the typical approach
when localization and mapping are combined into the coherent SLAM, e.g. simultaneous
localization and mapping, framework [
The main goal of this work is to investigate the ways of creating a coherent, modularized software control system used for autonomous navigation of a wheeled robot having SLAM and pathfinding as the main components that can be plugged in and out for evaluating diferent approaches and methods.
2.
– 8 GB flash memory. – Web-camera; – 6-DOF manipulator; – Scanning lidar; – Programmable servo controller; – Wireless router.
The on-board computer is powerful enough to run Linux-based operating system
and is capable of processing complex algorithm in real-time. As the main framework
for interfacing individual components (sensors, controllers and etc.) and autonomous
control we chose Robot Operating System (ROS) [
Using ROS for system’s organization makes low-level robot control (servomotors signal control, sensor access etc.) more abstract for end-user through standard highlevel protocol called topics and services. Also, ROS makes applications, built for this framework, modular by executing diferent parts independently in nodes.
The communication with ground platform is done using Wi-Fi, making possible to control robot remotely in autonomous, semi-autonomous and manual modes. In case of fully autonomous control, all the algorithms may be ported directly to ground robot without additional rework.
The on-board computer grants standard interface to the following control mechanisms and sensors’ data: – Laser rangefinder (lidar) data in LaserScan format – Environment map in OccupancyGrid format – Odometry data in Odom format – Robot’s position in resultant map in geometry_msgs/PoseStamped format – Robot movements using control vector with geometry_msgs/Twist commands – Position following (geometry_msgs/Pose)
Localization and mapping is done using on-board computer with data from lidar
sensor and odometry. For simultaneous localization and mapping problem we use known
algorithm "gmapping" [
The on-board computer also implements the action_server and action_client
mechanism for known-goal-following tasks. In case of path following algorithm, mobile platform
"Nexus" implements the Pure Pursuit [
2.2.
We use C++11 for autonomous control system implementation. Application was
released as a ROS framework node, that uses the standard access interface for system’s
components for normal processing, visualization with RViz [
The system consists of 2 main parts: 1. Pathplanner 2. action_client for pathfollowing and building the trajectory
The software architecture of control system is shown in Fig.3. For demonstration
purposes of pathplanning algorithm we use Theta* [
The system can operate in 2 modes: pathplanning in known and unknown environments. For known environments, unknown parts of map considered as obstacles and the trajectory planner assumes this cells occupied. For unknown environment, unknown parts of the map considered as the new type of cells. As the algorithm proceeds, the system plans the trajectory as if this cells were free. When the system reaches such a cell, system replans the trajectory using new observations and map.
3.
Fig. 5 shows the visualization of our control system’s execution using RViz ROS package. The figure demonstrates the mapping, localization in built map and the trajectory from robotic system to goal point planned with Theta* algorithm.
As an experemental environment, we used the 6x3.5m room. The map was built using lidar and odometry data. The size of a single cell is 0.05m (the size is related to lidar’s error). The trajectory is planned without concidering the robot’s size. The robot’s position and orientation are corrected during movements using inertial measurement unit data. If the robot goes far from planned trajectory, then the robot’s position may be corrected manually.
The pathplannig output then goes to pathfollowing algorithm as an array of midpoints and the robot proceed towards this points one-by-one. In case of significant deviation from planned trajectory, the robot attempts to return to the nearest position of built trajectory. The robot moves with 0.2 m/s speed.
This demonstes the working capacity of our system in real environment using PathPlanning and SLAM algorithms with "Nexus" robotic platform.
This work presents the specification for unmanned ground robot autonomous control system based on "Nexus" platform. We described the software architecture and the mechanism of communication with robotic system. The final system may be executed on remote control system or on-board of ground robot. Also, the system may be adapted to any occupancy grid-based pathplanning algorithm.
The results provide us an opportunity to evaluate modern pathplanning algorithms on real robots.
The publication has been prepared with the support of the "RUDN University Program 5-100" and partially supported by RFBR grant No 17-07-00281.