Institutional Repository of Vilnius University
Dirbtinio intelekto metodų panaudojimas autonominiam besimokančiam robotui /
Abstract
dc:descriptionThis study aimed to review and analyze artificial intelligence algorithms Backtrack, Backtrack1 and Depth First Search (DFS) in a maze simulation, as well as an autonomous self-learning robot. The motivation behind this study was the lack of accessible and budget-friendly projects for home and school conditions, in order for younger generations to be presented with artificial intelligence conpects. For this reason, this study presents a solid artificial intelligence project example, its‘ simulation, also a method of implementation and theorethical knowledge behind it. Four variantions of a maze were included in the simulation project in order to test algorithms‘ ability to deal with dead ends and ability to avoid infinite loops. The results were based on the amount of steps that were made by each algorithm, the amount of decisions that were made and memory that was used at its‘ peak. It was discovered, that Backtrack algorithm is unable to complete complex mazes, ones that include cycles and have many partings, but is able to run on almost any device, as uses little memory. Backtrack1 is able to avoid cycles in mazes and solve all of the mazes presented in the study, however, uses more memory, especially if the maze gets larger. Depth First Search is more complex in its’ implementation, however, it can solve larger, more difficult mazes, with different structures; DFS may detect, if there is no exit in the maze as well and be faster. However, Depth First Search requires more memory. To test these algorithms in a low-power robot, another maze was implemented in code and then constructed in real-life. Additionally, components of a so-called low power robot were reviewed. Results conducted in code and robot were compared and it was concluded that the number of steps made, decisions made, and memory used was higher. This was due to the imperfections that come with real-life conditions: cardboard construction of a maze may be (even slightly) uneven, the sensors may work with an error term, additionally the algorithms were modified slightly to match Arduino requirements.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Lebedev, Piotr,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192845541&prefLang=en_US