Simultaneous Localisation and Mapping with a Tracked Mobile Robot

11/22/2022

This work focuses on the development of a tracked mobile robot capable of performing Simultaneous Localisation and Mapping (SLAM). The robot’s kinematics model was approximated by a differential-drive one, and its odometry estimate was experimentally calibrated. There is also focus on state estimation before SLAM by fusing odometry and inertial measurement unit (IMU) data with an Extended Kalman Filter (EKF). Both the calibrated odometry estimate and the EKF pose estimate are compared to motion-captured data in simple and complex trajectories. The EKF estimate is a considerable improvement on the odometry estimate. The SLAM part of the work focuses on two existing open-source graph-based algorithms, one for 2D mapping and the other for 3D mapping. The performance of both SLAM algorithms is reviewed by comparing their map and trajectory estimates with motion-captured data in a custom-made map. Both algorithms tested show more accurate localisation estimates compared to the EKF pose estimate when performing complex trajectories. The map obtained, although congruent with the real map geometry, showed significant noise, especially the 3D mapping algorithm, due to hardware limitations.

Authors in the community:

• 

  F

  Francisco Vieira Pinheiro Gonçalves

  ist190255

Supervisors of this institution:

• 

  A

  Alexandra Moutinho

  ist13959

• 

  M

  Mário António Da Silva Neves Ramalho

  ist12229

Master of Science in Mechanical Engineering

mechanical-engineering - Mechanical engineering

Keywords

• Robotics
• Simultaneous Localization and Mapping
• Odometry
• Tracked-Drive
• Extended Kalman Filter

eng - English

Instituto Superior Técnico