Track Introduction
This track is a high difficulty level segment of the Tencent AI Arena Artificial Intelligence National Open Competition, primarily designed for university students with experience in reinforcement learning competitions and projects. The schedule for this track mainly includes pre-competition training courses in reinforcement learning, as well as three stages: open qualifiers, semi-finals, and finals.
Target Participants
This track is open for registration to full-time junior college students, undergraduate students, master's students, and doctoral students who are currently enrolled. Each team must consist of members from the same university, with a maximum of five members and at least one faculty advisor. There are no restrictions on the members' majors, and each individual can only join one team.
Rewards
Teams that successfully submit models and participate in the rankings for the open qualifiers, semi-finals, and finals can receive the following prize. The amounts are in RMB (pre-tax).
Champion: RMB 200,000 + Competition Certificate + Trophy
Runner-up: RMB 100,000 + Competition Certificate + Trophy
Third Place: RMB 60,000 + Competition Certificate + Trophy
Excellence Award (4-8th place): RMB 20,000 + Competition Certificate
Merit Award (9-16th place): RMB5,000 + Competition Certificate
Additional Rewards
In addition to the prize money, participating teams can also receive the following rewards:
1.Official participation certificates for completing the open qualifiers (successful model submission and validation).
2.Outstanding participants may have the opportunity to obtain internships related to the Tencent AI Arena project, as well as eligibility for Tencent Group's campus recruitment and internship recruitment fast track. Specific details will be announced separately.
3.Outstanding teams will be invited to participate in sharing sessions at the Tencent STAC.
Description: Gorge Walk utilizes the Gorge Forest map, featuring a starting point, an ending point, a road, obstacles, acceleration gains, and treasure chests. The agent does not have a full view of the map and can move within it, use summoner skills, and collect rewards from treasure chests. The agent cannot continue forward upon encountering obstacles. The agent must reach the endpoint within a set time, or it will be deemed as a mission timeout. (After successful registration, a detailed development guide is available on the platform.)
Task Objective: To drive the agent through algorithmically trained models, enabling it to learn movement strategies through continuous exploration of test maps. The objective is to efficiently use summoner skills and speed boosts to collect treasure chests, avoid obstacles, and reach the endpoint within a limited time. Participating teams must train and submit a model locally within the designated time, and control the hero Luban No. 7 in the assessment map to score as many points as possible in the least amount of time, thereby completing the Gorge Walk's map exploration target.

Ranking Rules: In the last week of the competition, the system will automatically run with the latest model submitted by the participating teams. Their total score will be the team's ranking score, which will be the final result of the open qualifiers. The top 80 teams completing the challenge in this track will advance to the semi-finals.
Description: This stage of the competition uses the Honor of Kings' 1v1 Mojia Map, which is elongated in shape. At both ends of the map are the revival points for the two agents, with the team's crystal located in front of the revival point. The crystal continuously produces minions for its team, which will automatically move towards the enemy's camp, attacking enemy towers, crystals, and heroes along the way. In front of the crystal is the team's defense towers, which can attack enemy heroes and minions entering its range. Agents can freely move and use skills within the map. Over time, agents gain gold and experience, with significant amounts earned from killing enemy minions, enemy heroes, or destroying enemy towers. The team that successfully destroys the opponent's crystal wins the round.
Task Objective: Train models within a designated time and given computational resources to learn the optimal winning strategy through continuous exploration of the 1v1 map.
Scoring Rules: In each round of competition between two teams, both heroes must play two matches, switching sides to battle against the opposing hero. Winning a match earns one point, while losing scores no points.
Ranking Rules: In the last week of the competition, the system will automatically conduct battles using the latest models submitted by the participating teams. The number of rounds for each leaderboard battle is to be determined. Teams will be ranked based on their accumulated points. This ranking will be the final result of the semi-finals, with the top 8 teams advancing to the finals.
Description: This stage utilizes the Honor of Kings 3v3 Changping Map. In addition to the main elongated lane, the map includes an upper and a lower jungle area. At each end of the map are the revival points for each team's agents, with the team's crystal located in front of the revival point. The crystal continuously produces minions for its team, which automatically advance towards the enemy's camp, attacking enemy towers, crystals, and heroes along the way. In front of the crystal is the team's defense towers, capable of attacking enemy heroes and minions entering its range.The jungle area is a primary source of economic and experience gains for jungler heroes. Players from both teams share the entire jungle area, although some entrances to the jungle are team-restricted. (Tip: These can be cleverly used to evade enemy pursuit.) Agents can freely move and use skills within the map. Over time, agents gain gold and experience, with significant rewards for killing jungle monsters, enemy minions, enemy heroes, or destroying enemy towers. The team that successfully destroys the opponent's crystal wins the match.
Task Objective: Train models within a designated time and given computational resources to learn the optimal winning strategy through continuous exploration of the 3v3 map.

Points Rules: In each round of battle between two teams, all lineup combinations of both teams must play and match with all lineup combinations of the other team one by one, switching sides for two matches. The specific number of games will be determined according to the number of heroes, with each victory earning one point and losses scoring no points.
Ranking Rules: The number of rounds for each leaderboard battle is to be determined, and teams will be ranked based on their points.
Requirements
It is recommended that each team possesses at least one computer that meets the following recommended specifications, to set up the local environment, carry out local development, and submit tasks to the cluster.
Windows10
i5-9th Generation
16GB
Green passageway
In the 3rd Tencent AI Arena Multi-agent Reinforcement Learning Competition, the 12 teams that advanced to the semi-finals, while retaining their team name and at least one original member, can bypass the open qualifier assessment and directly enter the semi-finals of the 'Gaming Algorithm - Advanced Track' in the 2023 Tencent AI Arena Artificial Intelligence National Open Competition. For more details, please consult our staff.
