ROS based AMR

Github repo: github.com/poplop03/robot1 and https://github.com/poplop03/robot1_cartographer

Bachelor thesis: Bachelor_thesis.pdf

----------System overview----------

----------Hardware overview----------

----------Software overview----------

----------SLAM----------

----------Localization----------

----------Navigation----------

----------ROS node overview----------

----------Result----------

Relative odometry result

• Relative odometry are tested with different linear velocity from 0.2m/s to 0.8m/s. Angular velocity is a constant 0.2rad/s.

• Red line is the robot estimated position using EKF, Black line is the estimated position only using wheel odometry.

• The robot is tele-operated from its origin and coming back to that origin to create a loop.

• As the linear speed increase, the global estimated position tend to shift an angle to the coordinate system. Resulting in shifted occupancy map generated.

• With a proper EKF configuration, the Fusion odometry is more reliable than wheel odometry in this scenario.

SLAM result

Experiments demonstrate that odometry remains reliable across different speeds when supported by a well-designed sensor fusion framework. Among the SLAM methods tested, 2D Cartographer proved to be significantly more robust than Gmapping, especially in dynamic environments, large area mapping. However, the performance of SLAM is heavily influenced by the update rate and synchronization of sensor data – particularly from point cloud sources. Even with reliable odometry, high-speed motion combined with sparse or delayed sensor updates can result in map inconsistencies and degradation. These challenges may stem from the interplay of various hardware components, such as the IMU, depth sensors, or integration methods, highlighting the importance of careful hardware-software coordination in mobile robot design.

Navigation result

The data is collected for 10 trials in 6 target is represented below. The robot moves step by step from origin to 1, from 1 to 2 … from 5 to 6.