Wyniki wyszukiwania - BazTech

1

Handling realistic noise in multi-agent systems with self-supervised learning and curiosity

Szemenyei Marton, Reizinger Patrik

Journal of Artificial Intelligence and Soft Computing Research

|

2022

|

Vol. 12, No. 2

135--148

EN

Most reinforcement learning benchmarks – especially in multi-agent tasks – do not go beyond observations with simple noise; nonetheless, real scenarios induce more elaborate vision pipeline failures: false sightings, misclassifications or occlusion. In this work, we propose a lightweight, 2D environment for robot soccer and autonomous driving that can emulate the above discrepancies. Besides establishing a benchmark for accessible multiagent reinforcement learning research, our work addresses the challenges the simulator imposes. For handling realistic noise, we use self-supervised learning to enhance scene reconstruction and extend curiosity-driven learning to model longer horizons. Our extensive experiments show that the proposed methods achieve state-of-the-art performance, compared against actor-critic methods, ICM, and PPO.

2

Model środowiska wieloagentowego w neuroewolucyjnym sterowaniu statkiem

Łącki M.

Zeszyty Naukowe Akademii Morskiej w Gdyni

|

2010

|

nr 67

30-36

PL

W artykule tym przedstawiono propozycję użycia neuroewolucyjnego systemu wieloagentowego do wspomagania decyzji manewrowych sternika statku płynącego na ograniczonym obszarze wodnym. W środowisku tym sternik jest osobnikiem określonej populacji, która za pomocą algorytmów ewolucyjnych oraz metod uczenia ze wzmocnieniem dostosowuje się do wyznaczonego zadania, jakim jest bezpieczne przepłynięcie danego obszaru przez wyznaczone jednostki pływające.

EN

This paper presents the proposal of multi-agent environment to simulate and demonstrate learning behavior of helmsmen in ship maneuvering. Simulated helmsmen are treated as individuals in population, which through environmental sensing learn themselves to navigate in restricted waters selecting an optimum trajectory of a vessel. Learning phase of the task is to observe current situation and choose the best action. Neuroevolutionary algorithms are used to solve this task. The best fitted individuals from each population become parents for the next generation of helmsmen in simulated multi-agent navigational situation.

3

Reinforcement Learning in Ship Handling

Łącki M.

TransNav : International Journal on Marine Navigation and Safety of Sea Transportation

|

2008

|

Vol. 2, no. 2

157--160

EN

This paper presents the idea of using machine learning techniques to simulate and demonstrate learning behaviour in ship manoeuvring. Simulated model of ship is treated as an agent, which through environmental sensing learns itself to navigate through restricted waters selecting an optimum trajectory. Learning phase of the task is to observe current state and choose one of the available actions. The agent gets positive reward for reaching destination and negative reward for hitting an obstacle. Few reinforcement learning algorithms are considered. Experimental results based on simulation program are presented for different layouts of possible routes within restricted area.