Laboratory of Robotics and Engineering Systems

Funder

Organizational Unit

Publications

Hybrid neural network based models for evapotranspiration prediction over limited weather parameters

Publication . Vaz, Pedro J.; Schutz, G.; Guerrero, Carlos; Cardoso, Pedro

Evapotranspiration can be used to estimate the amount of water required by agriculture projects and green spaces, playing a key role in water management policies that combat the hydrological drought, which assumes a structural character in many countries. In this context, this work presents a study on reference evapotranspiration (ETo) estimation models, having as input a limited set of meteorological parameters, namely: temperature, humidity, and wind. Since solar radiation (SR) is an important parameter in the determination of ETo, SR estimation models are also developed. These ETo and SR estimation models compare the use of Artificial Neural Networks (ANN), Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Recurrent Neural Network (RNN), and hybrid neural network models such as LSTM-ANN, RNN-ANN, and GRU-ANN. Two main approaches were taken for ET(o )estimation: (i) directly use those algorithms to estimate ETo, and (ii) estimate solar radiation first and then use that estimation together with other meteorological parameters in a method that predicts ETo. For the latter case, two variants were implemented: the use of the estimated solar radiation as (ii.1) a feature of the neural network regressors, and (ii.2) the use of the Penman-Monteith method (a.k.a. FAO-56PM method, adopted by the United Nations Food and Agriculture Organization) to compute ETo, which has solar radiation as one of the input parameters. Using experimental data collected from a weather station (WS) located in Vale do Lobo (Portugal), the later approach achieved the best result with a coefficient of determination (R-2) of 0.977. The developed model was then applied to data from eleven stations located in Colorado (USA), with very distinct climatic conditions, showing similar results to the ones for which the models were initially designed ((R2) > 0.95), proving a good generalization. As a final notice, the reduced-set features were carefully selected so that they are compatible with free online weather forecast services.

2023-01Journal article

Open Access

A decision-support system to Analyse Customer Satisfaction Applied to a Tourism Transport Service

Publication . Ramos, Celia; Cardoso, Pedro; Fernandes, Hortênsio C. L.; Rodrigues, João

Due to the perishable nature of tourist products, which impacts supply and demand, the possibility of analysing the relationship between customers’ satisfaction and service quality can contribute to increased revenues. Machine learning techniques allow the analysis of how these services can be improved or developed and how to reach new markets, and look for the emergence of ideas to innovate and improve interaction with the customer. This paper presents a decision-support system for analysing consumer satisfaction, based on consumer feedback from the customer’s experience when transported by a transfer company, in the present case working in the Algarve region, Portugal. The results show how tourists perceive the service and which factors influence their level of satisfaction and sentiment. One of the results revealed that the first impression associated with good news is what creates the most value in the experience, i.e., “first impressions matter”..

2022-12-31Journal article

Open Access

Autonomous temporal pseudo-labeling for fish detection

Publication . Veiga, Ricardo; Exposito Ochoa, Iñigo; Belackova, Adela; Bentes, Luis; Parente Silva, João; Semiao, J.; Rodrigues, João

The first major step in training an object detection model to different classes from the available datasets is the gathering of meaningful and properly annotated data. This recurring task will determine the length of any project, and, more importantly, the quality of the resulting models. This obstacle is amplified when the data available for the new classes are scarce or incompatible, as in the case of fish detection in the open sea. This issue was tackled using a mixed and reversed approach: a network is initiated with a noisy dataset of the same species as our classes (fish), although in different scenarios and conditions (fish from Australian marine fauna), and we gathered the target footage (fish from Portuguese marine fauna; Atlantic Ocean) for the application without annotations. Using the temporal information of the detected objects and augmented techniques during later training, it was possible to generate highly accurate labels from our targeted footage. Furthermore, the data selection method retained the samples of each unique situation, filtering repetitive data, which would bias the training process. The obtained results validate the proposed method of automating the labeling processing, resorting directly to the final application as the source of training data. The presented method achieved a mean average precision of 93.11% on our own data, and 73.61% on unseen data, an increase of 24.65% and 25.53% over the baseline of the noisy dataset, respectively.

2022-06-10Journal article

Open Access

An individual-based Model of the Red Alga Agarophyton chilense unravels the complex demography of Its intertidal stands

Publication . Vieira, Vasco M. N. C. S.; Engelen, Aschwin; Huanel, Oscar R.; Guillemin, Marie-Laure

Algal demographic models have been developed mainly to study their life cycle evolution or optimize their commercial exploitation. Most commonly, structured-aggregated population models simulate the main life cycle stages considering their fertility, growth and survival. Their coarse resolution results in weak predictive abilities since neglected details may still impact the whole. In our case, we need a model of Agarophyton chilense natural intertidal populations that unravels the complex demography of isomorphic biphasic life cycles and be further used for: (i) introduction of genetics, aimed at studying the evolutionary stability of life cycles, (ii) optimizing commercial exploitation, and (iii) adaptation for other species. Long-term monitoring yield 6,066 individual observations and 40 population observations. For a holistic perspective, we developed an Individual-Based Model (IBM) considering ploidy stage, sex stage, holdfast age and survival, frond size, growth and breakage, fecundity, spore survival, stand biomass, location and season. The IBM was calibrated and validated comparing observed and estimated sizes and abundances of gametophyte males, gametophyte females and tetrasporophytes, stand biomass, haploid:dipoid ratio (known as H:D or G:T), fecundity and recruitment. The IBM replicated well the respective individual and population properties, and processes such as winter competition for light, self-thinning, summer stress from desiccation, frond breakage and re-growth, and different niche occupation by haploids and diploids. Its success depended on simulating with precision details such as the holdfasts' dynamics. Because "details" often occur for a reduced number of individuals, inferring about them required going beyond statistically significant evidences and integrating these with parameter calibration aimed at maximized model fit. On average, the population was haploid-dominated (H:D > 1). In locations stressed by desiccation, the population was slightly biased toward the diploids and younger individuals due to the superior germination and survival of the diploid sporelings. In permanently submerged rock pools the population was biased toward the haploids and older individuals due to the superior growth and survival of the haploid adults. The IBM application demonstrated that conditional differentiation among ploidy stages was responsible for their differential niche occupation, which, in its turn, has been argued as the driver of the evolutionary stability of isomorphic biphasic life cycles.

2022Journal article

Open Access

A study on the prediction of evapotranspiration using freely available meteorological data

Publication . J. Vaz, Pedro; Schütz, Gabriela; Guerrero, Carlos; Cardoso, Pedro

Due to climate change, the hydrological drought is assuming a structural character with a tendency to worsen in many countries. The frequency and intensity of droughts is predicted to increase, particularly in the Mediterranean region and in Southern Africa. Since a fraction of the fresh water that is consumed is used to irrigate urban fabric green spaces, which are typically made up of gardens, lanes and roundabouts, it is urgent to implement water waste prevention policies. Evapotranspiration (ETO) is a measurement that can be used to estimate the amount of water being taken up or used by plants, allowing a better management of the watering volumes but, the exact computation of the evapotranspiration volume is not possible without using complex and expensive sensor systems. In this study, several machine learning models were developed to estimate reference evapotranspiration and solar radiation from a reducedfeature dataset, such has temperature, humidity, and wind. Two main approaches were taken: (i) directly estimate ETO, or (ii) previously estimate solar radiation and then inject it into a function or method that computes ETO. For the later case, two variants were implemented, namely the use of the estimated solar radiation as (ii.1) a feature of the machine learning regressors and (ii.2) the use of FAO-56PM method to compute ETO, which has solar radiation as one of the input parameters. Using experimental data collected from a weather station located in Vale do Lobo, south Portugal, the later approach achieved the best result with a coefficient of determination (R 2 ) of 0.975 over the test dataset. As a final notice, the reduced-set features were carefully selected so that they are compatible with online freely available weather forecast services.

2022Conference object

Restricted