EKF SLAM Applied to a Mobile Robot

As part of a recent university project, I undertook the exciting challenge of designing a robot capable of performing EKF SLAM. This project used a PenguinPi robot equipped with a single front facing camera and wheel encoders (odometry). The robot was placed randomly in an arena with beacons at unknown positions, which served as landmarks for the EKF algorithm. By leveraging computer vision, the robot was able to autonomously explore and map its environment, while accurately estimating its own position. This project exemplified the use for EKF SLAM in real-world robotic applications.

EKF SLAM is favoured in mobile robot applications for its ability to concurrently localize the robot and map its surroundings in unknown environments. By effectively fusing sensor measurements with a dynamic model, particularly in handling non-linearities, EKF SLAM provides a robust solution for real-time navigation and mapping tasks, making it a practical choice for autonomous mobile robots.

For this project, the robots design and functions for movement were provided (i.e., functions for retrieving images from the Pi-camera, getting number of ticks from the encoders, and setting voltage to the wheel motors were already built in). The final operation of the robot in various scenarios can be seen below.

In the unit, we explored motion control for car-like vehicles, internal and external sensors for mobile robots, navigation strategies (reactive and map-based), dead reckoning, map-based localization, probabilistic estimation, map creation techniques, and advanced topics like SLAM and EKF SLAM. Beyond the formal curriculum, I delved into research for machine learning algorithms for object recognition, communication protocols for multi-robot systems, and ethical considerations in autonomous decision-making.

If you would like to know more about the project, feel free to reach out.

Asier :)