Learning State-Dependent Sensor Measurement Model for Localization

Competitive Advantages

  • Utilizes a stochastic approach
  • Accurately estimates noise
  • Accurately estimates measurement bias
  • Does not require a precomputed noise covariance

Summary

USF inventors have devised a novel concept called stochastic perception or the ability for a robot to dynamically estimate the measurement model given the states of a robot and its environment. This method uses conditional probabilistic model to train itself and later predict the measurement model. This novel concept has a wide range of applications in the field of robotics, as it helps to estimate the measurement model given the states of a robot and its environment.

Robots and Landmarks Used to Collect Data Set

 

Desired Partnerships

  • License
  • Sponsored Research
  • Co-Development 

 

Patent Information:
Title App Type Country Serial No. Patent No. File Date Issued Date Expire Date
Learning State-Dependent Sensor Measurement Model for Localization Utility United States 16/236,715 10,572,802 12/31/2018 2/25/2020 12/31/2038