Esineiden etsintä sotkuisista ja rajoitteisista tiloista
Kalle Huhta-Koivisto · Aaltodoc (Aalto University) · 2026
Object search executed in a cluttered or spatially constrained space requires logical inference to be successful. Clutter is caused by other objects in the search space that can hide the target behind them. Constraints consist of floor, walls, and ceiling, which all can be seen in a cabinet like space. This work examines robots that perform mechanical search to find the target. This bachelor thesis aims to categorize different solutions for object search problems. These solution groups are compared, and it is determined whether certain solutions are better than others. Solution groups are divided into probability-based, model-based, semantic-based and learning-based solutions. These solutions focus on home environments containing more than 6 cluttered objects. This bachelor thesis is conducted as a literature review. Results indicate that learning-based solutions are more efficient, especially in cases where there is a large amount of clutter. This can be seen with higher success rate than with other solutions even with search areas with 20 cluttering objects. Learning-based solutions are often used when there exists clutter as they can generalize better for never seen before spaces. In the future it seems like generative artificial intelligence is being used more to find semantic clues about the search space.