Abstract
Modular robots are versatile systems whose compositions can be adapted and optimized for a wide range of applications and environments. However, the practicality of their reconfigurations is often limited by the necessity of manually deriving models and adapting control software for each new configuration. This manual process can be time-consuming and complex, making it less feasible to fully utilize the potential of modular robotics. Existing approaches for automatic derivation of models and effortless deployment of model-based controllers consider robotic manipulators only. In contrast, we consider modular wheeled mobile robots and present an approach for automatically recognizing their kinematics and deploying their navigation capabilities given mobility and perception modules. Our approach has been tested through both simulations and experiments. The simulations validate the approach across different platforms, each with varying compositions of modules. The experimental results further show the effectiveness of our approach, by deploying the autonomous navigation of a platform composed of four steering wheels, and by focusing on car-like, differential, and omni-like kinematic configurations. © 2025 Elsevier B.V., All rights reserved.