A reciprocal sampling algorithm for lightweight distributed multi-robot localization

This work is situated in the context of collaboratively solving the localization problem for unknown initial conditions. We address this problem with a novel, fully decentralized, real-time particle filter algorithm, designed to accommodate realistic robotic assumptions including noisy sensors, and asynchronous and lossy communication. In particular, we introduce a collaborative reciprocal sampling algorithm which allows a drastic reduction in the number of particles needed to achieve localization. We elaborate an analysis of our reciprocal sampling method and support our conclusions with simulation results. Finally, we validate our approach on a team of four real robots within a controlled experimental setup.

Authors

• Amanda Prorok
• Alcherio Martinoli

Keywords

• Robot kinematics
• Manganese
• Robot sensing systems
• Mathematical model
• Collaboration
• Equations
• Multi-robot Localization
• Local Problems
• Particle Filter
• Real Robot
• Scalable
• Probability Density Function
• Detection Model
• Standard Algorithm
• Position Error
• Detection Probability
• Localization Error
• Local Algorithm
• Motion Model
• Mobile Robot
• Geometric Relationship
• Normalization Constant
• Steady-state Error
• Update Step
• Collaborative Strategies
• Global Localization
• Swarm Robotics
• Ideal Filter
• Normal Probability Density Function
• Robot Performance
• Probability Density
• Normal Density Function
• Computational Cost
• Error Equation