Robust Recursive State Estimation: Application in UAV Pose Determination

Background: Determining the pose of vehicles, regardless of the environment, is a crucial task. This task is accomplished by sensing and observing various environmental parameters using passive or active sensors. However, sensor data is often contaminated by different types of errors, including blunders. Blunders can cause large deviations and lead to unreliable parameter estimation. Estimators based on L2​ and L1​ norms either fail to bound the influence of blunders or do so insufficiently. Therefore, for statistically reliable estimation of the parameters of interest, the estimation procedure must be robust against this type of error.

Objective:  This master's thesis initially focuses on investigating and studying various recursive parameter estimators that integrate blunder detection and mitigate their influence. Secondly, it focuses on decision-making and integrating a suitable outlier detector to specific recursive estimators that use fused sensor data to determine the pose of a UAV.

Specific Methodology:

Kalman Filter: As a recursive parameter estimation frame work

RANSAC: An example of outlier detector (not limited to this method)

Provided Materials:

Simulated and real LiDAR, camera, IMU data and environment.

Requirement:

• Strong knowledge of programming; MATLAB is preferred, but other programming languages are also acceptable.
• Solid understanding of recursive Bayesian estimation.
• Willingness to work with various types of data, such as LiDAR, camera, and IMU.
• Strong background in adjustment and error theory.

Basic proficiency in statistical analysis.