Publication: Data-driven modelling translational and rotational speed of differential drive wheeled mobile robot with varying payloads
Loading...
Date
2023-07
Authors
Muhamad Husaini Asyraf bin Adnan
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Differential drive mobile robot is one of the most popular robots that has been used in various applications around the world. By changing the speed of either side of the wheels, the differential drive mobile robot may alter the direction of its movement [1]. The robot's two wheels are typically positioned on its two sides. Both wheels must spin at the same rate for the robot to move forward or backward. The robot will turn to its right if the speeds of the two wheels differ, such as the left wheel revolving more quickly than the right wheel. A mobile differential drive robot's navigation system may impacted by the weight of the payload until it starts to produce significant systematic inaccuracy. Therefore, a system identification procedure is needed when developing a load-carrying mobile robot in order to model the uncertainty of how the different payloads may impact the robot's linear and rotational velocities. In this paper, the translational and rotational speed of differential drive wheeled mobile robot is modeled using a data-driven approach. A system identification process is done by acquiring data from the real-physical robot using sensors such as Inertial Measurement Unit (IMU), load sensor and built-in encoder. The system identification process includes the data acquisition, selecting transfer function model order, estimating the parameters of model, and validating the estimated model. These modelling techniques should address the problem of improperly perceived system behavior and the difficulty of theoretically simulating the system due to its complex features and unpredictability. In this project, performance measures including comparison of estimated model output and validation data to find the fit to estimation data in percentage and mean square value are utilized to assess the model's correctness. All the transfer function models that meet the project
criteria of 80 % fit to estimation data are used to find the nominal transfer function of the robot.