Browsing by Keyword "Decision Sciences (miscellaneous)"
Now showing 1 - 3 of 3
Results Per Page
Sort Options
Item Designing a generalised reward for Building Energy Management Reinforcement Learning agents(IEEE, 2021-09-08) Martinez, Ruben Mulero; Goikolea, Benat Arregi; Beitia, Inigo Mendialdua; Martinez, Roberto Garay; Mulero, Rubén; Arregi, Beñat; Mendialdua, Iñigo; Garay, Roberto; Solic, Petar; Nizetic, Sandro; Rodrigues, Joel J. P. C.; Rodrigues, Joel J.P.C.; Gonzalez-de-Artaza, Diego Lopez-de-Ipina; Perkovic, Toni; Catarinucci, Luca; Patrono, Luigi; DIGITALIZACIÓN Y AUTOMATIZACIÓN DE LA CONSTRUCCIÓN; EDIFICACIÓN DE ENERGÍA POSITIVA; Tecnalia Research & InnovationThe reduction of the carbon footprint of buildings is a challenging task, partly due to the conflicting goals of maximising occupant comfort and minimising energy consumption. An intelligent management of Heating, Ventilation and Air Conditioning (HVAC) systems is creating a promising research line in which the creation of suitable algorithms could reduce energy consumption maintaining occupants' comfort. In this regard, Reinforcement Learning (RL) approaches are giving a good balance between data requirements and intelligent operations to control building systems. However, there is a gap concerning how to create a generalised reward signal that can train RL agents without delimiting the problem to a specific or controlled scenario. To tackle it, an analysis and discussion is presented about the necessary requirements for the creation of generalist rewards, with the objective of laying the foundations that allow the creation of generalist intelligent agents for building energy management.Item Fast Real-Time Trajectory Planning Method with 3rd-Order Curve Optimization for Automated Vehicles(IEEE, 2020-09-20) Lattarulo, Ray; Perez, Joshue; CCAMAutomated driving (AD) is one of the fastest-growing tendencies in the Intelligent Transportation Systems (ITS) field with some interesting demonstrations and prototypes. Currently, the main research topics are aligned with vehicle communications, environment recognition, control, and decision-making. A real-time trajectory planning method for Automated vehicles (AVs) is presented in this paper; the contribution is part of AD’s decision-making module. This novel approach uses the properties of the 3er order Bézier curves to generate fast and reliable vehicle trajectories. Online execution and vehicle tracking capacities are considered on the approach. A feasible trajectory is selected based on the criteria: (i) the vehicle must be contained by a collision-free corridor given by an upper decision layer, (ii) the vehicle must be capable to track the generated trajectory, and (iii) the continuity of the path and curvature must be preserved in the joints. Our approach was tested considering a vehicle length (automated bus) of 12 meters. The scenario has the dimension of a real test location with multiple roundabouts.Item Transfer Learning and Online Learning for Traffic Forecasting under Different Data Availability Conditions: Alternatives and Pitfalls(Institute of Electrical and Electronics Engineers Inc., 2020-09-20) Manibardo, Eric L.; Laña, Ibai; Del Ser, Javier; IAThis work aims at unveiling the potential of Transfer Learning (TL) for developing a traffic flow forecasting model in scenarios of absent data. Knowledge transfer from high-quality predictive models becomes feasible under the TL paradigm, enabling the generation of new proper models with few data. In order to explore this capability, we identify three different levels of data absent scenarios, where TL techniques are applied among Deep Learning (DL) methods for traffic forecasting. Then, traditional batch learning is compared against TL based models using real traffic flow data, collected by deployed loops managed by the City Council of Madrid (Spain). In addition, we apply Online Learning (OL) techniques, where model receives an update after each prediction, in order to adapt to traffic flow trend changes and incrementally learn from new incoming traffic data. The obtained experimental results shed light on the advantages of transfer and online learning for traffic flow forecasting, and draw practical insights on their interplay with the amount of available training data at the location of interest.