Master Thesis: Robust Cog Error Estimation Algorithm

Your Role

This master thesis will be carried with assistance from our Development team. This team’s daily work involves maintaining and improving our on-board algorithms to meet tomorrow’s performance requirements. The main products are our Tire Pressure Indicator (TPI) and Loose Wheel Indicator (LWI) – both of which implement advanced signal processing to estimate features, without the need for additional sensors.

Description

Our algorithms are heavily reliant upon high resolution wheel speed signals. The raw format for these wheel speed signals is the elapsed time between two flanks of the wheel cog encoder. As these cogs are subject to mechanical inaccuracies, their errors must be estimated and compensated for. The main focus for this thesis project is to investigate filters and/or algorithms that, in a robust manner, can estimate the mechanical errors of the cog wheel. Two possible approaches are described below.

Kalman Filter Adaptation

Investigate how Kalman filter-principles can be incorporated with the existing algorithm for estimating cog errors.

- How should the state uncertainty be updated?

- How can the filter quickly, but without overshooting, estimate the cog errors after a reset?

- How can we utilize the knowledge that the sum of states is equal to one complete revolution?

Outlier Rejection

Unrepresentative signal samples may be encountered while hitting a pot hole or when a cog is missed by the encoder. These require special attention and should be discarded before entering the cog error estimation. If included in the estimation scheme, an inaccurate wheel speed signal may be calculated.

- How can outlier rejection be implemented to discover and intelligently discard deviating signal samples?

- Can this be combined with the Kalman approach described above to achieve a quick convergence when returning to correct signal samples?

Your Profile

We are looking for an engineering student who are studying their last year of a technical aimed at signal processing, sensor fusion, statistical modeling and software development. An interest in vehicle mechanics is encouraged. We expect you to have excellent study results (average 4 or higher) and that you are driven, can take initiative and work independently. The project will be carried out at our head office in Linköping.

Looking forward to your application! Don’t forget to include personal letter, CV and course listing with grades. Applications are considered on a rolling basis.