This thesis bridges the gap between generative artificial intelligence and innovation in footwear design, an industry that is quite conservative in its design methods despite the constant demand for innovation. Initial applied research showed that while 3D technologies have the potential to enhance efficiency in development, they are still time-consuming, as the 3D generative AI was not advanced enough yet and therefore left out of the scope. As a result, the research focused on applying AI on image generation, particularly using LoRA fine tuning to effectively capture brand identity and consistently produce high-quality results with reference images and ControlNet. To validate these findings, the tools and workflows were applied in a case demonstration to a design brief. Additionally, a user test was done with a designer of Filling Pieces, comparing the outcomes of the AI framework versus traditional processes in a survey. The findings showed that the AI framework significantly accelerates the generation of design concepts and enhances creativity by producing more novel designs. However, the study identified limitations in clarity, completeness, and production feasibility, emphasizing the need for AI-generated designs to be complemented with technical drawings for communication with the factory. The relationship between AI and human designers should thus be complementary to achieve the best results.

REV’IT! would like their first road racing boot on the racetrack before the racing season 2024 for the riders they sponsor in the MotoGP Championship. These riders wear everything REV’IT! from the neck down, except for boots. For boots, they currently have to arrange another sponsor deal with one of the direct competitors of REV’IT!. As contact is already there, the risk of losing the riders to these competitors for the entire sponsor package increases. REV’IT! has experience developing adventure bike boots, race gloves and race suits, but not in terms of motorcycle racing boots. Research, designing, and testing had to be done to get from this point to a race-ready product. This project served as a jump start in this process.
The initial goal was to find and dive into a design problem for a concept of a boot subsystem by executing different types of analyses to pinpoint current areas of improvement based on experience with competitor boots and create a list of requirements and an overview of the design drivers. As REV’IT! stands for quality, technical innovation and solving user problems, it would be fitting to make a boot that is not only as good as its competition but also has an answer for the problems indicated by the riders. The necessary data was collected through interviews with experts, a co-creation and Q&A session and interviews with the REV’IT! Riders, brainstorming sessions, questionnaires, and literature research. Besides the boot requirements and design drivers, four potential research areas surfaced. With the use of an evaluative method, the choice was made to focus on improving the current ankle brace design to be better at protecting the riders against hyper-rotation injury. The problem statement became: “Limiting the damage caused by hyper-rotation of the joints of the lower limb of REV’IT!’s MotoGP riders during a crash with a lightweight, low-profile solution that does not cause any discomfort in terms of pressure or donning and doffing, and does not limit the rider in their freedom of movement and the tactile feel to a noticeable extend.”
With in-depth research, ideation, calculations, prototyping, tests and multiple iterations, a final concept was generated that limits hyper-rotation in the direction of inversion and eversion even when combined with plantar and dorsiflexion. 2.1 (when plantar flexed), 7.0 (when neutral), and 3.7 (when dorsiflexed) times more force is required to invert this brace compared to the brace of one of the best racing boots currently available. Meaning this brace absorbs more impact energy during a crash. The concept combines a comfortable, lightweight and low-profile pivoting 3D printed PA11 hard part based on a 3D scan of the rider with six reinforcing Dyneema® strings that mimic the ligamentous structures of the ankle in terms of placement and functionality. It allows for the complete active range of motion in the direction of plantar and dorsiflexion while limiting inversion and eversion in all flexion positions. The strings allow for the personalisation of the amount of limitation based on rider anatomy and preferences.

This thesis describes the design and development of an accurate and low cost 4D foot scanner in the context of podiatry. The intention is to make a first step towards the development of an affordable 4D foot scanner with commercial grade performance at a low cost, using commodity hardware. By doing so, 4D scanning can hopefully become more accessible to the general public, and accelerate the development and adoption of ultra personalized footwear and digital manufacturing in healthcare. Podiatrists focus on the treatment of physical conditions in the lower regions of the human body, which are often related to foot conditions. Personalized footwear solutions (e.g. orthotics) are widely used to relieve a patient of such conditions, which are primarily designed based on static 3D scanning data of the foot. With additional input of motion analyses and professional experience, a podiatrist can adjust the design of an orthotic to fit the patient. However, the change in foot measurements during different phases of the gait cycle are not accounted for in the design of these personalized solutions, which could vary up to 8 mm. With 4D foot scanning (dynamic 3D scanning), podiatrists will be able to observe and acquire data of the dynamic morphology of a foot during the gait cycle. This should allow the design and development of truly personalized orthotics that are able to support a patient during the entire gait cycle. To support this vision, a proof of concept of a 4D foot scanner has been developed. The presented proof of concept iterates over a previously built 4D foot scanner (Vidmar, 2020). The proof of concept includes: an optimized embodiment for improved scanning quality and scanning consistency, a trigger system to improve the human-computer interaction of the scanner, and a scalable camera configuration for up to 9 cameras. The hardware performance of the scanner in combination with the newly designed data acquisition pipeline has been evaluated in terms of acquisition consistency, speed, and memory usage. The outcome is that the scanner is able to manage dynamic data acquisition with a camera configuration of 9. Also, for a camera configuration of 7, the scanner shows a linear trend in memory consumption, acquired frames, and acquisition speeds, which suggests that performance of the scanner is predictable and constant for this configuration. More elaborate analyses should give better insights into the long time performance of the scanner for different camera configurations. The evaluation of the quality of both static and dynamic scanning data has been done with the implementation of nonrigid ICP. The accuracy of the scanner showed a minimal accuracy error of 2.274 mm. Compared with international 3D scanning standards (minimum accuracy error of 2 mm), the performance of the scanner is considered as a desirable outcome for this graduation project.

This thesis explains the research into and implementation of overuse lower-limb injury prediction among military recruits, using wearable plantar-pressure sensing and biomechanical gait algorithms. The research has been done with the purpose of designing a usable, affordable and accurate injury-prevention tool, increasing both military wellbeing and operability. For years, researchers have gathered evidence of correlations between specific gait biomechanics of individuals and their influence on injury incidence. Using various types of laboratory equipment, such as pressure plates, walkways and treadmills, researchers found a strong correlation between parameters like cadence, vertical loading, (time to peak) heel rotation and local peak pressure values and different types of common overuse injuries in the lower limbs. However, laboratory tests often neglect the influence of footwear, distraction and fatigue. In addition, they are expensive and time-consuming tasks. One specific user group that suffers from a high incidence of overuse injuries is the military. About 3000 recruits follow basic military training each year, of which about 13% end up with one of three most common injuries: MTSS, iliotibial band syndrome and tibial stress fractures. Of those injured recruits, about 8% gets discharged from the military. This high rate is caused by the intensity of training, often accompanying high carried loads and pressure to perform. Specifically, the cumulation of repeated smaller impacts during marches is bound to cause overuse injury. The ability of commercially-used pressure-sensitive insoles to measure abnormality in injury-predictive gait parameters was tested in a series of studies. First, measurements from the insoles were compared to a commonly used gait-analysis tool: a GAITRite walkway (n=20). This study found no statistical agreement in sample-to-sample predictability. To validate whether those results could be caused by methodical differences, a second study was performed comparing the insole measurements of injured soldiers to those of control subjects (n=10). This test did result in significant between-group differences for all measured parameters. Furthermore, significant differences in parameters were found between walking on military boots and running shoes. A strategic study revealed that the product could find a competitive advantage in service and software innovation, with a focus on (a) accuracy while maintaining usability for specific user groups and (b) multi-diagnostic ability and spread market targeting (both b2b and b2c). Specific user demands were found and defined as a reduction in insecurity at various levels (e.g. where a soldier wants to be reassured about their personal fitness, a commander wants to ensure operation-readiness). A design proposal was created (Figure 1), based on research insights. The concept insole not only meets military-specific embodiment, hardware needs and improved sensor placement, but a service-design proposal enables both direct users and important stakeholders to use the data for injury prevention, rehabilitation, adaptation of footwear, operational management, training customization and general overuse-injury research. Besides meeting the needs within a military context, the product could be adjusted to meet the needs of other (occupational) overuse-injury sensitive users, such as police, hospital staff and factory workers. Clinical applications include automation of insole orthosis customization and prevention of ulceration among diabetic patients.

In 2010, an estimated 15 million infants were born prematurely, which makes up 10% of all childbirths that year. Just in the Netherlands there were 11,530 preterm births in 2018. Since these children should be spending additional months inside of the uterus, they are delicate and need to grow until near regular gestational age inside of an incubator in the neonatal intensive care unit (NICU), which provides these children with clean, humid, and warm air. One important part of the care of these infants is keeping track of their development, in order to adjust nourishment, as well as spot possible complications early. A regular part of these measurements is the head circumference (HC) measurement. This measurement is performed by two nurses, one holding the infant still while another winds a tape measure around the delicate head, in order to take the measurement. Unfortunately, this process causes the infant to stress, which is correlated to an increased chance in reduced neurological development of the infant. Therefore, Vectory3, a company specialized in 3D scanning for healthcare, commissioned a project to create an alternative way to take this measurement, which would cause the infant less stress than the current method. During this project, this alternative has been conceived and developed into a functional prototype: MONITOR3D. MONITOR3D makes use of 3 hybrid photogrammetry cameras in order to take a contactless 3D capture of the infant’s head, allowing NICU staff to measure the 3D model after acquisition, instead of measuring the child directly. MONITOR3D’s cameras are housed in an unobtrusive shell which is placed on top of the incubator hood, staying outside of the hygienic zone of the child during use. Though the hardware for MONITOR3D is already in a functional state, providing higher accuracy than the current tape measure-based measurements, the software still consists of unmodified open-source software, for which a simple but relatively time consuming (compared to tape measuring) workflow has been developed. The current prototype of MONITOR3D is thus not ready to be used by regular in a NICU setting. However, it can possibly be used for research purposes in its current state. Future developments should focus mainly on the software part of the product, automating as many of the mesh processing steps as possible, in order to make the device usable in a live NICU environment as alternative to the current tape measure.

Trials is a subdiscipline of mountain biking that consists of two variants: one practiced with a trials bicycle and one with a trials motorcycle. Though these two disciplines have the same basic requirements, there are clear distinctions in the technical sports gestures that the respective athletes employ. This is due to some fundamental differences in the vehicles used, one critical difference being that the trial motorcycle has a suspension system while the trial bicycle does not. This leads to incompatibility between the sports at a high level as cross-training can cause learned skill errors, where athletes adopt bad habits from practicing the other discipline. This is particularly problematic for trial motorcyclists, as there are various benefits from training with a bicycle, namely: bicycles provide a more complete workout for trial riders, bicycles are easier to maintain, laws restrict the use of motorcycles in certain areas, and the fact that many trial riders begin their trial experiences on a bicycle. To solve this problem, this project aims to answer the following research question: “Will the integration of a suspension system in a trial bicycle allow it to mimic more closely the usability of a motorcycle trial, in order to be suitable for cross-training of trial athletes?” To answer this question, I employ the Stanford d-school’s Design Thinking approach to create a functional prototype of a hybrid trial bike. This prototype is subsequently tested and evaluated by a target user, which provides a departure point for further research and development into this topic.

Technology is being used more and more to aid elite athletes in improving their performance. Recently there have been promising developments in the research using IMUs regarding wheelchair rugby and basketball. Applying this same measurement setup to wheelchair tennis could give better insight in the wheelchair mobility of wheelchair tennis athletes. This knowledge could aid the athletes in causing less sport related injuries, altering the focus of their training or even training entirely new aspects of the sport. This is why this research will be answering the research question: “What is/are the most influential factor(s) regarding the mobility of wheelchair tennis athletes to discriminate between the winning and losing player?”. We define the wheelchair mobility performance as all the measurable parameters related to the movements of the athlete. These mainly consist of the velocities and accelerations reached by the athletes. The measurements resulted in three different ways to output variables. Match statistics, these include information about the distance travelled, turns made and points played during the match. This was chosen to give general insight into the match. The wheelchair mobility performance (WMP), shows information about the velocities and accelerations during the match. This was chosen to compare the velocities and accelerations of athletes, they are plotted against the average value of the entire database of all measurements, this will be a standardized baseline to compare the results to. Lastly the speed zone plots, the plots give an indication of how long the athlete has been driving within a certain velocity range. To give better insight in the dynamic of the match all outputs also includes the constraints of winning/losing and serving/receiving. It is therefore possible to compare the differences between won and lost points and served and received points. There was also experimented with the possibility of making a plot that shows the location of the athlete on the field as a heatmap. However, because the IMUs only record accelerations the cumulative error to distance had to be manually adjusted for, therefore this was only done for one match as a pilot case. The match statistics, WMP plots and speed zone plots showed no significant differences between winning and losing points. Therefore, this study found that the mobility performance has no deciding effect on the outcome of a tennis match. The technique used to hit the ball and place it on the right spot in the field will probably be the determining factor here. This said, the mobility does however have to be above a certain threshold for the player to compete at a certain level of play, so neglecting it completely is not recommended, however the focus should be on the technique that involves hitting the ball. The heatmap case shows that the winning player plays more aggressive, but whether he is playing aggressive because he is winning or winning because he is playing aggressive cannot be concluded by this research.