Recent technological advancements are revolutionizing last-mile package delivery, with autonomous delivery vehicles (ADVs) emerging as a sustainable alternative. These small ground vehicles autonomously navigate pavements, aiming to operate seamlessly in pedestrian-rich environments without any human intervention. The goal of this project is to design a concept of an autonomous delivery vehicle that portrays predictable behaviour for pedestrians.

Extensive research involving literature analysis, interviews, and observational studies unveiled the challenges that current ADVs are facing, including unpredictable behaviour and distrust. The absence of predictability has a big impact on effective on-road communication, and it potentially leads to unsafe traffic interactions. A viable solution lies in communicating the ADV's driving intent. This study delves into the nuances of intent communication in pedestrian environments, and identifies intent signals that are predominantly communicated through body language. The objective is to replicate human-like signals in the behaviour of ADVs, to ensure an intuitive understanding of such cues from previous traffic experiences.

Focusing on designing predictability through intent communication via body language, a test exploration is conducted, testing multiple different prototypes. Iterative trials resulted in interesting insights and eventually lead to the most promising intent signal – the "Looking" scenario. This intent signal is inspired by a head-turning or someone looking into the desired walking direction. Participants recognised this signal, facilitating an intuitive understanding of the meaning of the signal.

The evaluation test confirmed the intent signal's positive effect on enhancing pedestrians' feelings of safety, trust, and comfort during an ADV interaction. These increased feelings of safety, trust and comfortability came from the fact that they were able to predict the movements of the ADV and because they felt "seen". The robot reacted to the participants by showing its intent, this reaction gave participants the perception that the ADV has indeed detected them and therefore won't collide with them. Consequently, it is recommended to integrate intent communication into ADV designs to ensure safer and more harmonious ADV-pedestrian interactions.

Three design guidelines are formulated based on the insights from the exploration phase. The three guidelines emphasize the interpretation, visibility, and relevance of intent signals. The conceptual ADV design presented in this project, aligns with these guidelines. The intent signal in the design concept of the ADV makes their behaviour more predictable, gives people a sense of control because they feel detected, which both leads to an increase in safety, trust, and comfortability. Although the intent signal shows great promise in creating a safer and more comfortable traffic interaction between ADVs and pedestrians, additional research is needed to fully understand the impact of integrating this intent signal.

The process of physical rehabilitation in most cases consists of visits to a physical therapist and exercises the patient is expected to perform at home. To increase or maintain strength or flexibility these exercises need to be performed regularly. Practice teaches us that not all patients succeed here. In this project, we look into what is keeping them from succeeding and how we can support this process.

In literature, four categories of barriers are described, from which the psychological barriers are most interesting to investigate further within this project. Motivation plays a big part within this category.

The Fogg model gives structure to the different components of motivation and different factors influencing the execution of preferred behavior. It shows us triggers will fail if motivation is low or the task takes too much effort to complete.

In a lot of the cases, the intention to complete the exercises is present with the client, but the behavior does not reflect this intention. The intention-behavior gap can be bridged the same way the fogg model suggests triggers to be efficient, by motivating the user or making the task take less effort.

By providing support during the exercise - providing both motivation and decreasing effort - and giving the patient the opportunity to track their progress after doing the exercises for a longer amount of time we can increase the exercise adherence.

To provide support, we need to gather data on the movements that are made during the exercise. We developed a textile stitched strain sensor that tracks the angle of a joint. The sensor has conductive thread stitched in a tight zigzag pattern onto Kinesio tape. Since the tape is adhered to the skin, the sensor experiences minimal hysteresis.

Using the data gathered by the stitched strain sensor, we tested giving back different kinds of feedback using two different actuators. We found that giving the test subjects more precise data on their movement made them more accurate, but at the same time, made them experience their movement as less accurate. This on the one hand, gave them more motivation to improve, but also made them less confident in performing the exercise.

The project proves that using the real time data gathered by the stitched strain sensor can add to better executing of on exercise and has potential to with that data also contribute to long term exercise-adherence.

As current cleaning targets are not achieved, Amsterdam requested to explore the potential for robots to contribute to a cleaner city. This thesis focused on exploring how a robot could contribute. A contextual analysis, reviewing literature and municipal documents, showed an increasing need for improved waste management efficiency in response to Amsterdam’s growth, increasing tourism and sustainability goals. However, increasing labour shortages and demographic ageing make it difficult to rely solely on additional human labour in the coming years.
To further analyse the context, employees of department Schoon were accompanied and interviewed. Six key values were identified contributing to job satisfaction - safeguarding these is crucial for successful robot implementation, including freedom and dislike of heavy work. Challenges were also identified, including labour shortages and reduced efficiency.
To determine where a robot could add value, three main waste problems in Amsterdam were identified and analysed. Inefficient waste management and littering behaviour were often the cause. To understand this behaviour, key influencing factors were identified and used in the analysis and thesis.
Furthermore, by analysing key factors of waste management behaviour, it was found that a cleaning robot can potentially trigger motivators for littering behaviour, suggesting that the robot's task influences waste management behaviour. Furthermore, when identifying factors influencing HRI, it was found that rewarding interactions can motivate proper behaviour, suggesting that robots can contribute to a cleaner environment without having to clean.
Based on the findings, specific requirements and a design vision for the robot were established, emphasising promoting proper waste management behaviour, ownership and waste management efficiency. To communicate the vision to the municipality, two concepts were developed, the Mobile Robot Bin and Mobile Robot Container. The Robot Bin aims to motivate proper behaviour through rewarding interactions and to increase waste collection efficiency. To align with Amsterdam's identity and raise awareness, the appearance will be created in collaboration with local artists.
The Robot Container focuses on providing waste disposal options at will in the city centre, optimising waste management efficiency and reducing littering behaviour, as the current policy allows residents to place waste outside twice a week. The concept reduces the physical burden on employees by eliminating the need to collect waste from the street and minimises the need for heavy waste trucks in the centre.
To assess the concepts’ value and feasibility in contributing to a cleaner Amsterdam, stakeholders from the municipality of Amsterdam, department Schoon and AMS Institute evaluated them. Both received positive feedback, with the Robot Container being identified as the most promising solution for achieving a cleaner Amsterdam. However, challenges such as vandalism need to be addressed before implementation. Nevertheless, stakeholders were enthusiastic and asked about implementation activities. A roadmap outlining these activities was developed and evaluated by the AMS expert. However, further research is needed to develop and implement the robots to achieve a cleaner Amsterdam.

Automation in the form of swarm robots is an increasingly feasible opportunity for solving complex physical problems. However, how such a swarm is to be instructed and controlled is one of the many topics that requires further research. For this, we can look to nature for inspiration, as there are many nest building swarms. Perhaps the most notable of which are the mound building termites. These termites build incredibly intricate mounds that contain a royal chamber, nurseries, fungal gardens, covered walkways, and even ventilation shafts to thermoregulate the inside of the mound. There are some attempts to capture this behaviour in custom simulators. The problem is that these simulators have limited capabilities. That is why a new simulator specifically for termite-like builder robot swarms is required for further research in this field. In this project, the different requirements for such a simulator are identified and used to develop an open-source state of the art simulator, TermiteSim. TermiteSim is designed for experimenting with different control strategies and tuning the simulation parameters to recreate the desired behaviour. In other words: a tool for researchers to design desired behaviour of a builder robot swarm. In this thesis TermiteSim is used to simulate several termite-like building behaviours described in the literature and the resulting structures are compared to those of other simulators, which TermiteSim was successfully able to recreate.

Providing a bodily injury-free experience might be a promising opportunity for the adoption of VR and force feedback-based equipment in a gym context. Nevertheless, it is still uncertain how these systems could compromise the users’ physical state and how they should react a posteriori.

This Graduation project begins with the exploration of the bodily injury risks associated with the Ethereal Engine, an ultimate VR-based equipment that seeks a revolution in the fitness industry. Resulting in four different clusters -collision, posture, balance loss and long-term exposure-related risks-, a study on the current market-ready technologies that could minimize these risks is performed together with an assessment on how safety can be boosted while keeping the experience’s engagement.

Thanks to the research, posture evaluation is selected as the challenge with a differential potential over the rest to bring Ethereal Engine to the next level, providing an experience in which bodily injury risks could be reduced, motivating users to adopt better postural habits when working out. Right after, a design vision is formulated, emphasizing the need for inspiring users to self-correct their posture instead of imposing rules in order to avoid possible immersion breakages.

In this context, three concepts are proposed, which present different feedback alternatives that vary in terms of explicitness, immediacy, and strictness. However, a priori there is no clear answer on which of these alternatives could most positively affect the experience. To overcome the knowledge gap an experiment that not only assesses the performance of the 16 participants but also their perception of the understandability, usefulness, performance, posture correction, general engagement, confusion and criticism of the feedback is conducted. During three different experiments where different feedback alternatives are assessed -immediate visual clues, statistics after certain repetitions, and immediate pop-up messages- participants are asked to perform 10 squats while lifting a water jug and receiving different scores depending on its displacement.

In order to do so, a partial prototype that allows for the first time to experience certain functionalities of the Ethereal Engine is developed. These functions are: a) a hybrid MoCap system that merges a Oculus Quest headset with a multi-camera-based skeletal tracking system -named MoCapForAll-. ; b) a calibration process that adjusts the range of motion according to the participant’s height; and, c) a demo game that enables a more intense workout while keeping track of your posture while squatting.

After discussing the results, including a calibration process that enables fair competitiveness and feedback at three different levels to communicate diverse information are spotted as desired features to include in the Ethereal Engine. Through these recommendations a virtual trainer that inspires users to self-correct their posture will be possible, and, subsequently, enabling a bodily injury-free experience.

Recently in the Netherlands, researchers have found that sleep duration and quality were suboptimal in the hospital. Evidence proved that many modifiable hospitalrelated factors negatively associate with patients' sleep (JAMA Internal Medicine, 2018). The sound factor is the most significant sleep disturbance in the hospital. In this graduation project, collaborating with Reiner de Graaf hospital and Critical Alarms lab, an information exchange system was proposed to raise awareness of sound as sleep disturbance. The system captures the sound-producing events and visualizes them with visually attractive graphics. In this system, we use the smartphone as the sound captor. The recorded sounds are processed locally on the phone and converted into information such as sound level and the category it belongs to (alarm, speech, incidental sounds, or snore). Fitbit is implemented in the system to collect sleep information. To both patients and medical staff, The Doplor sleep system presents the influence of sound on sleep in a friendly and comprehensive way. During this project, a functioning prototype was developed. We have tested its functionality and user experience with the potential users.