Unity Korea, '2021 RL Korea Dron Dilibi Challenge'

Unity Korea has been sponsored by MINI Korea and the '2021 RL Korea Droneliburi Challenge' hosted by Reinforcement Learning Korea, Reinforcement Learning Korea, Reinforcing Korea,

RL Korea is the largest open community in Korea for the establishment of reinforcement in 2017, and the first '2021 RL Korea Droneliburi Challenge', which started this year, aimed at learning to learn the goods of the logistics warehouse as a destination. A challenge that learns the reinforced learning algorithm in a simulation implemented simulation environment implemented by Unity. From November 1 to December 7, about five weeks, 28 teams were able to submit a 861st challenge model and confirm hot interest in strengthening learning.

In this challenge, 17 teams in the top 10, and 17 of the total 28 submissions were learned through the algorithms provided by the Unity Machine Learning Agents (UNITY MACHINE LEARNING Agents, ML Agent) and provided by the ML Agent for the rest of the teams. The Python API implemented an algorithm through the API. In 2017, the ML agent, which was first released, is used to prepare games such as a research project that evolves itself through research projects and self-evolving learning research projects under complex simulation environments.

In May 2021, the newly published ML agent V2.0 is a function that trains cooperative behavior, the ability to train collaborative behavior, the agent to observe the ability to observe various entities in the environment, The cooperative environment is fully supported.

The models submitted to the challenge have been reviewed based on the number of delivery completion and time, and a total of 4 teams were won. The best prize was the RMN, the Excellence Award, won the Nova Park, and the prize money was given 2 million won and 1 million won, respectively. Also, the Hwang and Noose teams were selected for the unity, which is given to two best results using the algorithm provided by ML Agent V2.0, and 500,000 won was given.

RL Korea Min-myeon, a researcher, It is possible to develop a relatively simple simulation environment that is relatively simple when utilizing Unity. In addition, using Unity ML agents, we have been developing a challenge environment based on Unity and ML agents, as it would be possible to easily participate in the challenge from a reinforced learning specialist, said the Challenge Environment. I think it would have been a chance to use the various functions of the ML agent to learning through the ML agent from the use of the PISAI API.

Kim In Took Yi-Korea, said, I was glad to see the algorithms that showed excellent results with excellent talents in the field of strengthening through the events that have been held this year through the events, said, I am happy to see the algorithms that showed excellent results with excellent talents and excellent results. It was a good opportunity to confirm the stability and performance again, and will continue to try to make the creators easily and conveniently.