Robot localization in a mapped environment using Adaptive Monte Carlo algorithm
Sagarnil Das
TL;DR
The work addresses robust robot localization in known environments using Adaptive Monte Carlo Localization (AMCL), validating its effectiveness through two ROS/Gazebo-based robot simulations. By adopting a particle-filter approach with adaptive particle counts, the study demonstrates real-time pose estimation and navigation to a goal, comparing UdacityBot and SagarBot across two map worlds. Key contributions include detailed parameter tuning for AMCL and move_base, a benchmark model for reproducibility, and a parallel development of a second robot variant to assess scalability and dynamics. The findings show AMCL can achieve reliable localization and goal-directed navigation in mapped environments, with practical implications for real-world deployment and guiding future hardware-integrated implementations.
Abstract
Localization is the challenge of determining the robot's pose in a mapped environment. This is done by implementing a probabilistic algorithm to filter noisy sensor measurements and track the robot's position and orientation. This paper focuses on localizing a robot in a known mapped environment using Adaptive Monte Carlo Localization or Particle Filters method and send it to a goal state. ROS, Gazebo and RViz were used as the tools of the trade to simulate the environment and programming two robots for performing localization.