Robot Localization Using Pre-calculated Covariance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robot localization methods using Kalman filters face inefficiencies due to long computation times and errors caused by wheel slip, particularly when calculating covariance of system and measurement noise, which are often done through trial-and-error or evolutionary computation methods.
Innovation Solution
A robot system that includes a gyroscope and encoder modules, with a control module applying calculated covariance of system and measurement noise to the Kalman filter, ensuring no parameter divergence during evolutionary computation, thereby restricting unnecessary computations and optimizing the experiment order of offspring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If evolutionary computation method is used to calculate covariance of system noise and measurement noise, then localization accuracy is improved, but computation time increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing the covariance matrices of system noise and measurement noise through evolutionary computation before actual robot localization operations. The control module loads these pre-computed covariance values from storage, avoiding repeated evolutionary computation during runtime localization, thus achieving high accuracy while reducing real-time computation time.
Solution Approach 2:
The patent extracts the computationally intensive evolutionary computation process from the real-time localization workflow. By separating the covariance calculation (performed offline) from the localization execution (performed online), the system maintains high localization accuracy while eliminating the time penalty during actual robot operations.
2Loss of time
If trial-and-error method is used to calculate covariance values, then computation time is reduced, but localization accuracy deteriorates due to dependence on experimenter input
Solution Approach 1:
The patent applies self-service by enabling the system to automatically calculate optimal covariance values through evolutionary computation without relying on manual trial-and-error input from experimenters. The algorithm autonomously optimizes the covariance matrices based on the robot's actual system characteristics and noise properties, ensuring both accuracy and time efficiency.
3Measurement precision
If covariance values are calculated for every system modification, then localization accuracy is maintained, but system efficiency degrades due to repeated computations
Solution Approach 1:
The patent applies local quality by storing pre-calculated covariance values in a database or storage medium for specific robot models and configurations. When the robot system operates with unchanged parameters, the control module directly retrieves these stored covariance values without recalculation, maintaining localization accuracy while significantly improving system efficiency during normal operations.
Data Source
AI summary
Provided are a robot capable of improving the computation rate by considering whether a parameter will diverge and modifying the experiment order of offspring during evolutionary computation, when the covariance of system noise and that of measurement noise are calculated for the purpose of localizing the robot by using a Kalman filter, and a method and medium of localizing a robot by using a calculated covariance. The robot includes a gyroscope module providing information regarding rotational angle; an encoder module providing information regarding velocity and information regarding rotational angle of a wheel by sensing motion of the wheel; and a control module estimating a current location according to a Kalman filter method based on information provided by the encoder module and the gyroscope module, a covariance of system noise and a covariance of measurement noise being calculated in an evolutionary computation and applied to the Kalman filter method by the control module, the covariance of system noise and the covariance of measurement noise satisfying a condition that no Kalman filter parameter diverges in the evolutionary computation.