Many high school students require costly private lessons to prepare for their final exams. These lessons go beyond what regular schools can provide. As a proof of concept, a distance learning program for exam preparation has been created using Moodle. It offers study materials that incorporate gaming and real-life applications. Each lesson begins with a diagnostic test and follows a specific training program to address students' weaknesses. An AI-based intelligent tutoring system has been developed to take on the role of a teacher and assist students with their work. After the system analyzes errors, the tutor provides specific hints and guidelines to help students solve problems accurately. The developed study material was tested on a group of students preparing for their final exams in mathematics.@en

The grow of urban environments caused many problems in traffic networks. On the roads in and around an urban environment many traffic incidents happen all the time, especially during rush hours or bad weather conditions. To minimize the negative impacts, an incident scene has to be cleared as soon as possible. Additional delay has been caused by bad communication and missing information. In this paper we present a new IT system based on the framework JADE (Java Agent DEvelopment). The system offers a blackboard like functionality for communication of first responders at the incident scene and control rooms. The system has been implemented and tested in laboratory settings and shows a significant reduction of incident process time. As a consequence, the resilience of traffic network in urban areas has been improved by better incident management and communication.@en

At many times we observed disturbances of traffic flow on highways. This may be caused by traffic accidents, bad weather conditions, road maintenance or rush hours. The Road Traffic Management takes many rules and regulation, to make road sections more robust to disturbances and improve the recovery from disturbances in traffic flow (traffic resilience). To study the effect and impact of these rules and regulations assessment models of traffic resilience was needed. In this paper, we designed and tested such an assessment model. The model is inspired by the well-known Resilience Triangle. The assessment of traffic resilience was based on measurements of the speed of traffic flow on highways. Neural Networks were used to model and smooth recorded speed data. The Traffic Resilience model has been tested on real life data.

@en

For many years the Dutch NS Railway Company runs all trains in the Netherlands. It can be expected that next year different transport companies will exploit the Dutch railway network. Possible that they will use different systems to design timetables, generate travel plans and to handle delays. In this paper we research the coordination between different systems. A prototype of distributed systems in different areas has been developed using agent technology. The system has been tested in simulation studies.

@en

Recently articles in Newspapers, University News Bulletins and Scientific Literature report about negative aspects of the wellbeing of students caused by COVID-19 epidemic. Half of the students have mental problems and don't participate in the teaching learning process anymore. In the Netherlands, Universities are surveyed by questionnaires, researching the mental health problems of students. In this paper we focus on students of Delft University of Technology. It proved from surveys, that many students complain about loneliness, fear, sleep deprivation and lack of study motivation. In this paper we report about experiments at the Faculty of Electrical Engineering, Mathematics and Computer Science, how students can be activated, motivated and socialized via study activities presented at the website of one of the Study Societies and via study-buddy groups. Students were personally invited to take part in discussions via a Forum, to enroll in group activities and to visit special lectures. A special COVID-19 didactics has been developed to stimulate students to make assignments in Calculus and Programing via Massive Open Online Courses developed in the framework EdX, an online learning destination and MOOC provider.

@en

In 1979 Fukushima developed a hierarchical, multilayered neural network called Neocognitron and used it for the automatic recognition of handwritten Japanese symbols. We combined the Neocognitron classifier with a special image and segment processor and applied the system in 2001 for automatic recognition of license plates in laboratory experiments. In this paper we report about a special image acquisition module and a postprocessor. We tested the system in real life conditions in an application of automated access control.@en

The areas of many cities in the Netherlands are covered by a network of stationary sensors, measuring special components of air pollution such as CO2, NO2, PM2.5 and PM10. The pollution with fine dust along roads, surrounding and crossing the city is primarily related to traffic density. To measure traffic density, we used a license plate recognizer based on a special Neural Network Neocognitron, analyzing the video footage of surveillance cameras along the roads. We also studied the onset and offset of traffic density to predict traffic density, using the first recorded sparse traffic data. In cooperation with MIT Senseable City Lab the Technical University of Delft has developed special mobile, low cost sensors to measure air pollution. These mobile sensors are integrated with stationary sensors to a heterogeneous sensor network and enable measurement of air pollution out of the reach of the stationary sensor network..

@en

In 2002 the city of Prague was affected by a flooding of the river Vltava, caused by heavy rainfall in the Southwest area of the Czech Republic. Successive parts of the city were inundated and inhabitants have to be evacuated from areas threatened by the raising water. After breaching the dikes and upcoming water via caves and sewers, streets were flooded. A special Crisis App was designed to route people to safe areas via safe routes. The routing system is visualized by fishbone maps, commonly used in car routing systems. To compute the shortest path from current location to a safe area a hierarchical version of Dijkstra shortest path algorithm has been used. Fleeing citizens and special observation agents report about changing situations. If crossings and streets are flooded, new routes have to be computed or existing routes have to be adapted by applying Dijkstra again. The application is Prague specific but can be generalized to similar smart cities in case of flooding disasters. More details will be presented in the paper including a simulation study showing the adaptive routing of the Crisis App.

@en

The Corona crisis limits the mobility of people and it proves that many people have an increased body mass index. A lot of people want to increase their fitness and to boost their immune system. There is a run-on fitness schools and fitness equipment for the home environment. At fitness schools, fitness coaches are available for personalized instructions and supervision. Fitness in the home environment lacks personalized instruction and supervision and can cause serious injuries. In this paper a digital coach will be introduced providing digital support and supervision. The digital coach offers personalized fitness program, supervises athletes during fitness exercises and provides a supporting system concerning injuries. A prototype of the developed system can be downloaded as an app on a smart phone.

@en

The INTERSPEECH 2021 Computational Paralinguistics Challenge addresses four different problems for the first time in a research competition under well-defined conditions: In the COVID-19 Cough and COVID-19 Speech Sub-Challenges, a binary classification on COVID-19 infection has to be made based on coughing sounds and speech; in the Escalation Sub- Challenge, a three-way assessment of the level of escalation in a dialogue is featured; and in the Primates Sub-Challenge, four species vs background need to be classified. We describe the Sub-Challenges, baseline feature extraction, and classifiers based on the 'usual' COMPARE and BoAW features as well as deep unsupervised representation learning using the AUDEEP toolkit, and deep feature extraction from pre-trained CNNs using the DEEP SPECTRUM toolkit; in addition, we add deep end-to-end sequential modelling, and partially linguistic analysis.

@en

Social media play an important role in crisis events. People in crisis events can report their observations of the crisis environment, and their mental and physical state. They can ask for help or assistance via Twitter. Some studies report, that some or a lot of compassion, concern and empathy is involved. Verification of this thesis was the main concern of our research. We analyzed huge datasets of Tweets available via Internet of recorded messages posted during natural disaster events. We detected many emotional tweets characterized by keywords. Emotional keywords were extracted and composed in an emotional dictionary, to be used for future flooding disasters.

@en

In this paper we present an information system improving situational awareness, communication and management during a flooding crisis. The system is based on the agent framework (JADE) and a blackboard like functionality, which enables rescue workers and services to improve communication, increase context awareness and activate rescue services. Observers in the crisis field, modelled as an agent, report about their observations using an iconbased crisis App on a smartphone. A prototype has been implemented and tested in field experiments.

@en

Current route planners on smart phones or special routing devices try to minimize the traveling time. But many users are no longer interested in the shortest route in time but they prefer a healthier route. In many smart cities there is a centralized distributed network of sensors measuring the pollution in a city on different locations for different time of the day. In this paper we present a routing App computing the less polluted route along the streets using the measurement of the eco sensors. The council of a smart city takes care of the health of their citizens. The developed routing App enables travelers with lung diseases to choose a healthy route. The routing algorithm is based on a special variant of the well-known Ant Based Control routing algorithm. To combine data measuring pollution and shortest traveling time, we solved a special multi-parameter problem. In some experiments we performed a user test of the routing App on different time and location.

@en

Last year's we can observe an enormous grow of surveillance systems in the public domain. In many countries surveillance systems are under discussion because of the privacy aspects of people. In the nineties many multimodal systems were developed for surveillance in Dutch trains. The use of camera systems was complicated because of bad lighting conditions, occlusion and posture. Only recently it is possible to use smart cameras to localize and track faces and displayed facial expressions by smart cameras installed in the public domain. Most systems are based on the Active Appearance model. In this paper we present a real version of a surveillance systems called Image Processing System developed in the vision lab of TUDelft. In an experiment we recorded the behavior of stand-up comedians in trains acting in several scenarios. The facial expressions in the video footage were analyzed.@en

The recovery of the economy, after the crisis at the beginning of this century, causes an enormous increase of traffic on the roads in the Netherlands in and around cities. After many years of discussion road pricing is considered as the most reasonable solution. Car drivers have to pay additional tax if they use the entrance roads of a city during peak times. A network of surveillance cameras along the roads have been installed and tested to register entering cars on the access roads. For many years sensors in the road surface are used to measure speed of cars on the roads to generate alarming messages for upcoming traffic jams and to provide alternative routes via panels along the road. Finally there is also a huge electrical network of traffic lights along the roads. By the enormous rise of mobile phones the interest in existing wired networks along the road is decreasing. In this paper we discuss the revival of wired networks and the integration with wireless networks.

@en

A Smart City can be considered as an open living lab, where new technology will be developed, implemented and tested. The focus of smart cities in this paper is on the safety of citizens, sustainable energy, risk prevention of disasters, smart transport systems using information technology and Internet of Things. Smart cameras and smart sensor systems play a crucial role in smart cities. Citizens are surveyed by smart cameras, registrating their presence, their activities but also their needs and security. Big Brother is watching you, the Orwellian Nightmare come true. In this paper we focus on the smart city of Amsterdam. At the entrance road of the city smart cameras register all entering vehicles; their license place will be recognized and checked for access in available databases. Similar smart cameras check car drivers for their access to the inner city and green zones. But not only car drivers are surveyed. Entering boats on the many waterways of Amsterdam are surveyed. Every vessel is supposed to have a transponder on board with an Automatic Identification System (AIS), enabling control of anomaly of shipping on the water and detection of intruders. This paper is a survey paper of developed technology and research around smart cities.

@en

Most routing devices use real time traffic information. It proved that the first hours of a new day can be used to find a good matching day in the past. The traffic data from the matching day can be used to predict traffic streams on the current day. We designed a historic database storing traffic information from real time available public domain databases. A dynamic version of the well-known Dijkstra shortest path algorithm was used to design a dynamic routing algorithm.

@en

In designing a course a teacher defines the educational goals, selects the teaching material and the way of assessment and finally defines a didactic model. In this didactic model the teacher defines his way of teaching taking into account the different ways of learning. Students starting an academic study interact with the educational or teaching learning environment. Students define their own study goals, they are supposed to have the right capabilities to finish the study successfully and are able to set themselves into action. But an important driving force is the affective component. Successful students like their study, enjoy their life at the university and interactions with fellow students but are afraid to fail for exams, dislike the aloneness, angry at failing systems. In this paper, we present well-known and new didactic models with special attention for affective components. We present a system to assess the emotional state of students. The system has been tested using a management game, where a crisis team composed of students reduce the negative impact of flooding in the city of Prague. The results of the experiment are described in this paper.

@en

Negative emotions and stress can impact human-human interactions and eventually lead to aggression. From the perspective of surveillance systems, it is of high importance to recognize as soon as an interaction escalates and human intervention is needed. One of the limitations of deploying a system in real life is that in practice it can only be trained on a limited number of situations. In this paper we examined the generalization capabilities of a trained system given context change. For this purpose we developed scenarios and made audio-visual recordings in four different contexts in which negative interactions might occur. To obtain a quantification of cross-context performance we kept the test context fixed and performed training on itself (cross-validation) and on all the other contexts. To explore whether multiple examples in the training set are beneficial, we also trained the classifier on a merged corpus of the three contexts that were not used for testing. These experiments were done with audio features, video features and audio-visual feature level fusion to investigate which modality generalizes best. We found that context change generates a decrease in performance that is varying with within-contexts similarities. Merging multiple contexts for training in most cases results in performance just below the best predictive single context. Audio is the most robust modality and in most cases the performance of audiovisual fusion is very close to the one of audio. @en