Accurate robot localization with noisy sensors


In this project, you will use a team of Khepera-III robots to address the question of accurately locating a set of objects in a given area using a set of 3D landmarks. The only two types of sensors available are the onboard sonar sensors and wheel encoders. Due to slippage of the wheels, the integrated values of encoder readings often give large accumulated position errors. These errors can be reduced if the robot can use sonar sensor data to estimate landmarks with known shapes and known locations, to reset drifts in the encoder readings. The noisy encoder data with a drift can in-turn be used to reduce ambiguity about identifying the 3D landmarks.

The objects to be detected will always be a black patches pasted on a white background. Therefore, each robot will communicate their locations to other robots once a target object is encountered. The robot will detect an object using a set of photo reflectors pointing at the ground.

You will evaluate the localizing algorithm by monitoring the localizing error of each object.


Leave a Reply