20.500.12556/RUL-164575 Metoda za vodenje sistema voziček-palica s preiskovanjem prostora stanj magistrsko delo Method for controlling the cart-pole system with state space search Problem voziček-palica je klasični problem v teoriji vodenja dinamičnih sistemov, pogosto uporabljen za testiranje krmilnih algoritmov in algoritmov strojnega učenja. To magistrsko delo prvič predstavlja vodenje sistema voziček-palica s preiskovanjem zveznega prostora stanj, kar doslej še ni bilo izvedeno. V nalogi je narejen pregled obstoječih pristopov za reševanje problema voziček-palica. Za realizacijo vodenja sistema s preiskovanjem zveznega prostora stanj so izpeljane enačbe dinamike tudi za vožnjo po neravnem terenu. Za preiskovanje zveznega prostora stanj pa so bili preiskušeni različni preiskovalni algoritmi, in sicer algoritem iskanja v globino z iterativnim poglabljanjem, algoritem A*, algoritem A* z omejenim spominom, algoritem kadarkoli A* in algoritem RTA*. Najprimernejši in najpogosteje uporabljen je bil algoritem RTA*. Da je preiskovanje zveznega prostora stanj učinkovito, je uporabljeno strojno učenje za učenje hevristične funkcije in prepoznavanje ciljnega stanja. V ta namen sta uporabljena metoda k najbližjih sosedov in naključni gozdovi. Vodenje sistema s preiskovanjem zveznega prostora stanj z RTA* se izkaže za uspešno že s pogledom naprej do globine 5. V poskusih s 100 naključno generiranimi testnimi stanji je bil sistem vedno uspešno voden do cilja. Vodenje sistema je bilo uspešno testirano tako na ravnini kot tudi na neravnem terenu, vključno s prečkanjem hriba in kotanje, kar kaže na robustnost in učinkovitost razvitega pristopa. The cart-pole problem is a classic problem in the theory of control of dynamic systems, often used for testing control algorithms and machine learning algorithms. This master’s thesis for the first time presents the control of a cart-pole system through the searching of continuous state space. The thesis reviews existing approaches to solving the cart-pole problem. To realize the control of the system by searching the continuous state space, the dynamic equations are derived also for driving over the uneven terrain. Various search algorithms are tested for the searching of the continuous state space, namely iterative deepening search, A* algorithm, memory-limited A* algorithm, anytime A* algorithm, and RTA*. The most suitable and frequently used algorithm was RTA*. To make the exploration of the continuous state space efficient, machine learning was used to learn a heuristic function and recognize a goal state. For this purpose, the k-nearest neighbors method and random forests were used. Controlling the system by exploring the continuous state space proved to be successful by RTA* with lookahead depth 5. In experiments with 100 randomly generated start states, the system was always successfully driven to a goal state. The control of the system was successfully tested on a flat surface as well as uneven terrain, including crossing a hill and a crater, demonstrating the robustness and effectiveness of the developed approach. voziček-palica preiskovanje prostora stanj strojno učenje realno časovni A* naključni gozdovi cart-pole state space exploration machine learning real-time A* random forests true false false Slovenski jezik Angleški jezik Magistrsko delo/naloga 2024-11-01 08:15:14 2024-11-01 08:15:31 2024-11-22 14:54:53 0000-00-00 00:00:00 2024 0 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 519.8 149240 213411331 15492.pdf 15492.pdf 1 59ED895A64EDE3AB31B95695E44E90BB bea75e7d842195572e33d5cdce4920d16ecb37501751f88f6164ee1af7a5f92f cebcf7f8-9820-11ef-b232-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=195099 Fakulteta za matematiko in fiziko 0 0 0