Behavior-based Driver Identification is an emerging technology that recognizes drivers based on their unique driving behaviors, offering important applications such as vehicle theft prevention and personalized driving experiences. However, most studies fail to account for the real-world challenges of deploying Deep Learning models within vehicles. These challenges include operating under limited computational resources, adapting to new drivers, and changes in driving behavior over time. The objective of this study is to evaluate if Continual Learning (CL) is well-suited to address these challenges, as it enables models to retain previously learned knowledge while continually adapting with minimal computational overhead and resource requirements. We tested several CL techniques across three scenarios of increasing complexity based on a well-known dataset for the Driver Identification problem. This work provides an important step forward in scalable driver identification solutions, demonstrating that CL approaches, such as Dark Experience Replay (DER), can obtain strong performance with only an 11% reduction in accuracy compared to the static scenario. Furthermore, to enhance the performance, we propose two new methods, Smooth Experience Replay (SmooER) and Smooth Dark Experience Replay (SmooDER), that leverage the temporal continuity of driver identity over time to enhance classification accuracy. Our novel method, SmooDER, achieves optimal results with only a 2% accuracy reduction compared to the 11% of the DER approach. In conclusion, this study proves the feasibility of CL approaches to address the challenges of Driver Identification in dynamic environments, making them suitable for deployment on cloud infrastructure or directly within vehicles. • We investigate Driver Identification in a realistic setting, adapting to new drivers. • We propose three Continual Learning scenarios with progressive real-world alignment. • We propose SmooDER and SmooER, leveraging driver continuity to boost performance. • We validate the effectiveness of these techniques using the OCSLab dataset.

In this paper, the energy efficient control of carbon dioxide heat pump systems is discussed from an experimental point of view. The performance of this kind of systems strongly depends on the operating conditions and in particular on the cycle high pressure. Because of the limited knowledge of certain system parameters and the difficulty of developing and implementing effective models, the problem of determining the optimal value for the cycle high pressure that leads to the maximum system performance is here faced by means of a model-free approach. Specifically, an Extremum Seeking Control (ESC) scheme, which can search for the unknown or slowly varying optimum input with respect to a certain performance index, is adopted. In particular, a variable water flow rate heat pump unit was considered. In this scenario, the performances of the ESC were compared with those provided by other methods available in literature (e.g. Liao's model). Experimental tests show that the ESC scheme guarantees better performance.

Water chillers play a crucial role in HVAC systems. Faulty operations of chillers can lead to energy wastage, system unreliability and shorter equipment life. Due to the intrinsic complexity of these systems, which are nonlinear with interrelated parameters, and since data regarding unforeseen phenomena and abnormalities are not usually available for air-conditioning installations, the development of fault detection algorithms is a burdensome task. In this paper, a data-driven approach is used in order to develop a fault detection methodology that makes no use of a priori knowledge about abnormal phenomena. In particular, we exploit a local Principal Component Analysis to handle nonlinear cases and to accent novelties with respect to non-faulty operations variability. The performance of the proposed approach is assessed by using a synthetic dataset which is related to an air-condensed water chiller in both fault-free and faulty conditions.

Cyber-physical systems (CPS) refer to novel hardware and software compositions creating smart, autonomously acting devices, enabling efficient end-to-end workflows and new forms of user-machine interaction, in a wide range of application fields. Given their heterogeneous nature, CPS are naturally designed in the so-called simulation-centric process, where physical equipment design are translated into behavioral simulation models. Although many tools exist to ease the design phase in the different disciplines, their full integration is still an open problem. This fact holds particularly true in a control design perspective, given that in a CPS the "plant" has a heterogeneous nature, and the design of model-based, advanced control techniques would strongly benefit from the availability of a common modeling environment. In this paper, we try to enhance this simulation-centric process by introducing a pure simulation kind of prototype, based on the co-simulation of the firmware and of the multi-physical controlled system. We introduce some innovative tools, implemented in μLab/CfL [1], and discuss upon their impact towards a better collaborative design and integration during the design of CPS. An example is given, taken from the HVAC (Heating, Ventilation, and Air Conditioning) field.