Max van IJsseldijk

As a Teaching Assistant for the Multiped Robots DBL, I developed MATLAB and Arduino starter code to help first-year students kickstart their coding journey. I guided them through programming and problem-solving, enabling the design and implementation of functional robotic systems. Additionally, as a Teaching Assistant at PROTO/Zone, I gained hands-on experience in mechanical and electrical prototyping by assisting students in building their prototypes. This included supporting design optimization and operating 3D printers and a laser cutter to bring their concepts to life.

During my internship at the Computational Autonomy and Robotics Laboratory (CURLY), I conducted research on optimal control and reinforcement learning techniques. My focus was on applying reinforcement learning to approximate the optimal value function in order to enhance the efficiency of a Model Predictive Controller, while ensuring safety guarantees. This research experience in the US was incredibly rewarding, and I received an 8.5 for my work.

During my time with the Formula Student team, I gained valuable experience in CAD, software development, and hands-on prototyping. Using tools like NX, Python, and openFOAM, I helped optimize the race car's performance, by designing parts like the rear wing, front wing, and undertray, as well as improving radiator and accumulator cooling systems. I created the entire automated workflow using Python to streamline processes and improve efficiency. Through iterative design, we improved the car's aerodynamic efficiency by 50% through the years. In addition to CAD work, I worked on the structural design and assembly of key components, which strengthened my understanding of mechanical and structural principles. I also gained practical experience fabricating parts, including carbon fiber components. Working with the team, I helped solve problems and improve the race car design. This experience taught me how to communicate better, work through challenges, and collaborate effectively with others.

As a member of several committees at Simon Stevin, I helped organize events, dinners, and activities for the general members. This experience helped me build teamwork and communication skills while contributing to the association's events and community.

For this internship, I utilized JavaScript to generate a dynamically updating view of the Knowledge Graph of the ASTRID system, showing its learning capabilities in real-time. The visualization was made using the Sigma JavaScript library.

For my master's, I specialized in machine learning techniques, including Bayesian methods and reinforcement learning. I applied these methods in my thesis, where I developed a system for socially aware robot navigation using reinforcement learning. To implement this, I created a custom reinforcement learning platform in C++ and Python, integrated with ROS. My research demonstrated that by learning a navigation policy based on physical points in space, defined by a convex safe corridor and pedestrian data, pedestrian disturbance could be reduced by over 70% compared to standard dynamic window approaches. This work was awarded an 8.5 and is currently being developed into an academic paper, as well as a coding contribution to the ROS community.

For my bachelor end project, I worked on an optimal control model for an active tire actuator to improve car handling and comfort performance. This was achieved with a LQR controller which optimized a constrained optimal control problem. This work was awarded a 9.0.