Localization accuracy is an extremely important aspect of mobile robotics, enabling robot control and navigation. This thesis examines localization accuracy of mobile robot based on the four-wheel-independent steering and four-wheel-independent driving (4WIS4WID) structure. The position and orientation of the robot can be estimated using a nonlinear kinematic model that includes wheel distances from the robot's center, as well as the steering angle and radius of each wheel. Discrepancies in radius and wheel distances from nominal values introduce estimation errors that cumulatively increase with robot movement. The study presents a robot parameter calibration process to reduce estimation errors and improve tracking of reference trajectories. Six different trajectories are proposed to evaluate accuracy, demonstrating the influence of all parameters. A sum of squared errors between reference and actual trajectories is chosen as the objective function for minimization. The actual robot position is obtained using an external OptiTrack measurement system with a declared accuracy of 0.2 mm. Parameter optimization is performed using two gradient-based and two stochastic methods. The newly obtained parameters are used for verification, resulting in a visible improvement of 60 percent. Since numerical integration also affects robot localization error, relying solely on odometry is insufficient for adequate accuracy. Fusion of multimodal measurements from various sensors allows for greater accuracy and robustness in robot localization. The extended Kalman filter (EKF) is used to fuse wheel odometry with tracking camera and inertial measurement unit (IMU) data. The results presented in this study demonstrate significant improvements in localization with the calibrated robot and sensor fusion.