The properties of Hardware-in-the-Loop (HIL) for the development of controllers, together with electronic emulation of physical process by Digital Twins (DT) significantly enhance the optimization of design and implementation in nonlinear control applications. The study emphasizes the use of the Raspberry Pi (RBP), a low-cost and portable electronic board for two interrelated goals: (a) the Electronic Control Unit (ECU-RBP) implementing a Lyapunov-based Controller (LBC) for nonlinear trajectory tracking of P3DX wheeled robots, and (b) the Digital Twin (DT-RPB) emulating the real robot behavior, which is remotely connected to the control unit. ECU-RBP, DT-RBP and real robot are connected as nodes within the same wireless network, enhancing interaction between the three physical elements. The development process is supported by the Matlab/Simulink environment and the associated packages for the specified electronic board. Following testing of the real robot from the ECU-RBP in an open loop, the model is identified and integrated into the DT-RBP to replicate its functionality. The LBC solution, which has also been validated through simulation, is implemented in the ECU-RBP to examine the closed-loop control according to the HIL strategy. Finally, the study evaluates the effectiveness of the HIL approach by comparing the results obtained from the application of the LBC, as implemented in the ECU-RBP to both the real robot and its DT.
The automatic generation of semantic environment maps remains a challenging problem in intelligent robotic systems, particularly when relying on sensor-driven perception or manual annotation. Although geometric mapping approaches provide accurate metric information, they generally lack high-level semantic structure and remain difficult for users to interpret, validate, and refine interactively. This paper presents NAREL, a human-centered framework for semantic map generation from natural language using large language models (LLMs). The proposed system operates as a conversational agent within a human-in-the-loop paradigm, allowing users to progressively describe indoor environments through natural interaction. These descriptions are transformed into structured machine-readable representations, which are subsequently validated through deterministic geometric and topological constraints and rendered as visual layouts to support interactive refinement. The proposed architecture combines language-driven generation, structured semantic representation, hierarchical modeling, constraint enforcement, and user feedback within a closed-loop interaction process that preserves map consistency throughout all construction stages. Experimental evaluation demonstrates the progressive construction of complete semantic maps including object insertion, traversal inference, conflict detection and recovery, and nested spatial structures. The obtained results show stable convergence behavior, complete structural instantiation ($\kappa_t = 1.0$), effective conflict isolation through the residual conflict metric $\delta_t$, and a normalized interaction effort of $\hat{T} = 2.75$ turns per region for the evaluated environment. These results confirm the effectiveness of combining large language models with deterministic validation and human-guided refinement for semantic map generation in assistive robotic applications.