Freedom of mobility is a crucial aspect of our daily lives. Consequently, engineering solutions for mobility, includ- ing smart wheelchairs, are becoming increasingly important for those with disabilities. However, the lack of a reliable solution for indoor localization has affected the pace of research in this direction. GPS signals cannot be measured indoors and envi- ronment modifications for wheelchair localization can be expen- sive and intrusive. This research explores the feasibility of us- ing ambient magnetic fields for indoor localization by exploit- ing the spatial non-uniformity due to ferromagnetic objects in ordinary working environments. A non-parametric density esti- mation technique was developed to build magnetic field maps. This approach is compared to an existing regression technique. Two different approximate kinematic models for the wheelchair are presented and implemented in a particle-filtering frame- work. Finally, the efficacy of these mapping techniques and motion models, including and excluding odometry information, are compared via tracking experiments conducted with a smart wheelchair.