Efficiency in logistics: Michael Nederhoed's data science & AI project

Efficiency in logistics: Michael Nederhoed's data science & AI project

07/12/2023 - 09:00

In this article, we delve into the experiences and accomplishments of Michael Nederhoed, a second-year student pursuing Applied Data Science & AI at BUas. Michael shares the details of an interesting project he worked on, collaborating with a robotics company in the logistics industry.  
Data Science & AI
  • Stories
  • Student work

Can you describe the project that you have been working on? 

Michael: ‘Sure! I have been working on a project with a robotics company in the logistics industry, specifically focused on factory automation. The project aimed to automate repetitive tasks using robotics, such as pick and place operations. The company was interested in exploring the potential of AI in their operations, as they were still using older methods like barcodes and pressure sensors.’ 

‘We developed a system where a robot could identify objects using image recognition, determine the correct arm movements, control its 7 motors, and calculate 3D coordinates based on a single image. We integrated three systems to create a sorting system. Our solution solely relied on AI algorithms and images.’ 

What was your role in this project? 

Michael: ‘I worked in a group with three other students, and we alternated roles every two weeks. During this project, every two weeks someone took on the role of the scrum master. My responsibilities also involved data analysis, programming in Python, and utilizing my expertise in AI. We had occasional online meetings with the client to address any questions or concerns. Additionally, we had a computer simulation of the robot they used, which allowed us to test our application.’ 

What was the most memorable aspect of this project for you? 

Michael: ‘The most intriguing aspect of this project for me was observing how different AI systems collaborated to accomplish a single task. We employed various AI techniques, including reinforcement learning, where the AI teaches itself to perform tasks without relying on pre-existing data. For example, we trained the robot arm to move from point A to point B and determine which motors to use. By rewarding the robot when it performed correctly and penalizing it for errors, we created a reward function. The robot earned points for getting closer to its target, and it learned to optimize its movements accordingly. Occasionally, the robot would apply excessive force in order to expedite its arrival at the destination, which could be dangerous. As a result, we implemented a system of penalty points to discourage such behaviour.’  

What are your future plans? 

Michael: ‘As I move into my third year, my future plans remain somewhat open. The field is rapidly evolving, and what exists today may not be the same in a few years. I'm not particularly inclined toward creating dashboards; instead, I want to leverage my skills to build AI models and develop innovative solutions. I aspire to carve out a niche in the intersection of data science, AI development, and consultancy.’


Would you like to know more about Michael? Visit his LinkedIn profile or check out his portfolio website