<p>The increasing use of robots in warehouse automation necessitates robust, cost-effective localization solutions. This work introduces a low-cost odometry system for indoor applications that maps common fiducial markers and lane markings, delivering precise robot pose corrections to an odometry source derived via an Extended Kalman Filter. The system utilizes a monocular camera, IMU, and wheel encoders as sensory inputs. It addresses the initialization challenges faced by existing visual-inertial-wheel algorithms, where limited initial movement hinders accurate estimates, and provides real-time pose corrections wherever landmarks are available, rather than waiting for loop closure, presenting a better correction strategy than post-processed pose graph optimization. The approach effectively eliminates error accumulation and mitigates incorrect z-axis estimation previously encountered for ground robots. Moreover, by promptly recovering localization upon observation of landmarks and ensuring robust tracking even in repetitive, low-texture environments, it inherently estimates odometry that can be used for better localization to overcome challenges like sudden pose loss and feature scarcity. Additionally, it features a fallback method to enhance odometry using lanes, even without the need for creating a lane map. Extensive experiments conducted with a wheeled mobile robot in complex warehouse environments-characterized by curvy trajectories, variable lighting, and multiple loops-and in simulation, demonstrate that the proposed system provides globally accurate, reliable, and drift-free odometry, unlike existing methods in large and visually challenging settings.</p>