<p>This paper presents a systematic and effective modelling and identification method for developing real-time models of a tracked mobile robot. This method uses particle swarm optimization (PSO) improved by combined Information Criteria for robust model selection. Tracked robots face unique challenges due to the complexity of their dynamics and varying terrain characteristics, thus necessitating effective model identification to reliably assist model predictive control (MPC) frameworks. Through PSO, we identify parameters for discrete state-space models across different system orders. PSO is also used to optimize the parameters for the PI controllers, which perform the chain velocity tracking. In the next step, an augmented nonlinear state-space model of the tracked robot is developed and combines the robot’s dynamics with kinematic models. The selected models are validated stringently through runtime analysis to ensure they can be employed in real-time. Experimental validation shows that the identified models achieve high fidelity, with model goodness values exceeding 92% for chain velocity tracking, demonstrating the model’s accuracy and suitability for real-time control. Supported by runtime calculations for the target hardware, our simulations demonstrate the validity and accuracy of the model, especially for implementation on a Raspberry Pi 4 Model B. The results show significant promise in leveraging this approach to improve trajectory tracking and control performance with real-time MPC for tracked robots, which represents an essential step towards efficient and scalable identification methods suited for real-time control problems in mobile robotics.</p>