Summer 2022 Race

Introduction

ROAR provides a Python-based racing simulation environment that allows contenders to race their autonomous AI agents. Contenders can fully train and test their AI algorithms without any vehicle hardware in this series. 

Code Submissions

Summer 2022 Submissions

This submission page has been built for our competitors to submit and participate in the 2022 Summer Competition. Final submissions will be due on Friday, August 26. If you are having any issues submitting your agent, please email (hc_kuan@berkeley.edu).

Award Criteria:

Important: By participating in ROAR Competition and submitting your code, you give the ROAR Committee the permission to evaluate and publicly release your competition results on our website.

Contestants who meet the following criteria will be awarded one or more of ROAR Competition Awards:

1. Grand Prize ($500 Amazon gift card):  This is awarded to the team who achieves the fastest total lap time after the completion of 1 lap. Any penalty due to major collision events does not disqualify the team.
2. Second-Place Prize ($200 Amazon gift card): This is awarded to the team who achieves the second fastest total lap time after the completion of 1 lap. Any penalty due to major collision events does not disqualify the team.
3. Third-Place Prize ($100 Amazon gift card): This is awarded to the team who achieves the third fastest total lap time after the completion of 1 lap. Any penalty due to major collision events does not disqualify the team.
4. Prime-Directive Award: This is awarded to the team who achieves the fastest total lap time after the completion of 1 lap and without encountering any major collision during the competition.

Maps

Since 2022, ROAR S1/S2 racing has adopted a CARLA map that was built from 3D LIDAR scan of the Berkeley campus and its surrounding hill roads, codenamed Berkeley Major Map. Contestants who submit their entries to the S1/S2 race must train and run their solutions based on the Berkeley Major Map. The official Berkeley Major Map is provided here. 

(Note: Please do not train or submit your ROAR Competition entries using earlier maps. The following links are only for benchmarking purposes to allow you to compare your new entries with past results. Your solutions trained on older maps may not be compatible with the Berkeley Major Map and subsequently may be subject to disqualification from the competition.)

• For the 2021 season, we used the Berkeley Minor Map
• For 2020 season, we used the Easy Map

Schedule

Voluntary Test Script Submissions are due on Friday, August 19, 11:59PM (PST) Midnight 

• Competitors can submit a test script to Huo Chao Kuan (hc_kuan@berkeley.edu) by providing your GitHub repository.
• Competitors can adjust their script as necessary before the Final Competition

*Submissions have been postponed for a week.

Mandatory Final Submissions are due on Friday, August 26, 11:59PM (PST) Midnight

Final Spring Competition will take place on Saturday, August 27, 2022

How to Submit

1. Make sure that your code is updated on GitHub

2. Make sure that you did NOT modify your submodules

3. Make sure that you can run the agent via runner_competition_evaluator.py by replacing the default agent in line #79

1. Feel free to change the num_trials and num_laps to smaller number to test, no need to do a full run through if it takes a long time.

4. If you added extra dependencies, please provide description of how to install in the readme.md
5. Submit the link to your GitHub repository. 
   Example: https://github.com/wuxiaohua1011/ROAR

How we will run your code

1. If we see that submodules are modified
   1. We will execute git clone YOUR_REPO to fetch your repository
   2. We will manually copy-paste ROAR_Sim into the repository
2. If submodules are not modified
   1. We will execute git clone –recursive YOUR_REPO to fetch your repository
3. We will start the Berkeley Major map 
4. We will modify the num_trials and num_laps in the runner_competition_evaluator.py to just one 
5. We will run python runner_competition_evaluator.py with use_manual_control=False
6. We will take the average of your 10 trials as your final result. For example, we will take the average of the time_elapsed(sec) column.

Guidelines for S1 Series

** Note that these guidelines are enforced in the ROAR_Sim repository and you SHOULD NOT do ANY modifications.

Starting Condition:

Here are the information provided about the map:

1. Please make sure that your vehicle spawns at Spawn ID 12. You may change it at this file on line 101. Also indicated in the screenshot below.

Run time:

1. Your vehicle MUST be able to run CONTINUOUSLY for 1 laps without external interference
   1. Note that you are provided only ONE lap of waypoints, YOU must handle the regeneration of waypoint in lap 1
2. Collision Penalty: If your vehicle collides with the static environment with intensity more than 10000, it will be regarded as a major collision. In such an event, your vehicle will be penalized by restarting from the beginning of the current lap. Depending on the location of the major collision, therefore, your vehicle will have to re-run the current lap.
   1. Note: It is possible that an auto-pilot that encounters a few major collisions could complete the 10 laps with a shorter total lap time than another that has zero major collisions. It is the contestant’s own responsibility to consider the trade off between safety and aggressiveness of the algorithm strategy.
3. Synchronized mode should be utilized.

Ending Condition:

1. Your Vehicle MUST pass the spawn point 12 at bounding box [2530, 99, 4064, 2543, 120, 4076] again to complete 1 lap.
2. All vehicles must complete the 1 lap including any major (intensity < 10000) collision penalty within 30 minutes. If a vehicle could not complete the 1 laps within 30 minutes, the code will be terminated and the result will be recorded as Did Not Finish (DNF)

How are we going to run your code: We are going to run runner_competition_evaluator.py.

NOTES:

• You may ONLY modify the code within the ROAR repo, NOT its submodules