The course teaches a 3D localization system called TurboTrack for fine-grained robotic tasks. It covers a pipelined architecture for extracting sensing bandwidth, a Bayesian space-time super-resolution algorithm for accurate positioning, and achieving sub-centimeter accuracy in x/y/z dimensions with low latency. The teaching method includes lectures on RFID localization, Bayesian fusion, and quantitative evaluation. The course is intended for individuals interested in advanced robotics, localization systems, and agile robotic tasks.
3D Backscatter Localization for Fine-Grained Robotics
Overview
Syllabus
Intro
Robots are moving towards fine-grained tasks
Today's robotic systems rely on vision for identification and localization
Can we use RFID localization for fine- grained robotics tasks?
Turbo Track
Localization by Estimating Distance
Approach 1: Measure Time-of-Flight
Approach 2: Measure Phase
Bayesian Super-resolution
Bayesian Fusion using MoGs
Space-Time Super-Resolution
Idea: Combine standard RFID reader with localization helper that has higher time resolution
Quantitative Evaluation
3D localization accuracy (Los)
Partial Implementations
Tracking Error vs Speed
Robotic Tasks: Robot Arms
Conclusion
Taught by
USENIX
