Graduate Student in Robotics
Tracking & monitoring of underground pipeline with Pioneer 3dx Mobile robot.
Intern project, Intelligent Systems Lab, IIT Kanpur
Project Guide - Dr. Laxmidhar Behera, Electrical Lab, IIT Kanpur
Project Members - Anirvan Dutta, Hritwik Shukla, Himanshu Ranjan
The goal of the project was to navigate an autonomous ground vehicle to detect and follow underground pipelines. The robot has to localize itself in outdoor environment and follow a presumed trajectory of the underground pipeline. The challenge involved was to perfectly localize the robot in outdoor environment and implement a trajectory controller to follow the trajectory of pipeline. The trajectory of the pipeline was generated from actual GPS points from Google Map. To improve the localization, extended Kalman Filter as well as unscented Kalman filter was utilized to fuse GPS, IMU and Odometry data of the Mobile Robot. Thereafter a kinematic trajectory controller was implemented to follow the robot along the desired trajectory.
​
The robot used in our project was Pioneer 3-dx manufactured by Adept Mobile robots. The robot was a two wheeled differential drive robot, where each wheel was driven independently. The kinematic analysis was done to control the motion of the non-holonomic differential drive robot.
​
To track the underground pipeline, the most essential aspect was localization of the robot in outdoor environment. However, sensors are noisy, and their measurements are prone to errors. By fusing the data from multiple sensors, we can obtain an overall position estimate with less error than a single sensor in isolation. It is often the case that a greater amount of sensor input data will produce more accurate position estimates.
Therefore, to improve the GPS pose estimate we planned on used Kalman filter for sensor fusion and perform state estimate and since our system was non-linear an Extended Kalman Filter was used. We used the data from GPS, IMU and Robot's Odometer measurement. We used 5 states for our state estimation model. These states were the position in Easting and Northing Measurements or Universal Transverse Mercator co-ordinate system (UTM measurements), the forward velocity, orientation, and rotational velocity. These states were chosen because they line up with the most general sensing environments.
After localization, the next step to path planning. The aim was development of efficient trajectory controller, taking care of uncertainties and dynamic scenarios while under tracking. The design of the kinematic controller was based was kinematic model of the robot and theory proposed by Martin et al. (An adaptive dynamic controller for autonomous mobile robot trajectory tracking, F.N. Martins, Control Engineering Practices, v.16). We had the GPS data of the underground pipelines and a trajectory estimation methodology was implemented which took GPS points of the underground pipelines and generated a trajectory for the robot to follow using the kinematic trajectory controller. Since underground pipelines are mostly linear in nature, least square curve fitting was implied which was less computationally exhaustive. The overall control scheme consisted of -
The objective of the project was achieved and the robot could traverse a given trajectory generated from Google map points. The simulation of the trajectory controller was done in Matlab. The robot was driven in real time using ROS platform. All the nodes were written in C++/Python. The critical aspect of the project was localization and navigation. It was essential for the robot to localize itself accurately as without this navigation would be prone to errors. We observed that in spite of fusing sensors using extended Kalman filter, the robot had an error of around 100cm while localizing itself in outdoor environment. While this involved an error for trajectory based navigation, it could be reduced either using more accurate sensors. Navigation using the given trajectory controller showed excellent performance.
To conclude, we presented a methodology which involved improved localization and trajectory control for navigating a mobile robot in outdoor environment in order to detect and monitor underground pipelines.