Context
Our Track 4 project at NAICT (中国全国青少年人工智能大赛) centered on a supermarket-shelf task: one robot retrieves goods from a shelf and places them in a designated area, then a robotic arm transfers the goods.
Participation
I served as team captain. In August 2026, our team received National First Prize in Track 4.
My Contribution
My main responsibilities were:
- Collecting data for model training.
- Transferring the collected data to a server and training an ACT model from scratch there.
- Deploying the model for the robotics task.
- Debugging the deployed system.
Selected Work
The task connected two stages of goods handling: shelf retrieval and placement by one robot, followed by transfer using a robotic arm. My work focused on the data, training, deployment, and debugging needed for the task.
Learning and Reflection
This project gave me hands-on experience with the ACT training workflow, from collecting data and sending it to a server to training the model from scratch, deploying it, and debugging the system. It strengthened my practical skills in connecting data collection, server-side training, and deployment in a robotics setting.
背景
我们参加了 NAICT 中国全国青少年人工智能大赛赛道 4。项目围绕超市货架场景展开：一个机器人从货架取货并将货物放到指定区域，随后由另一个机械手转移货物。
参与情况
我担任队长。2026 年 8 月，我们的团队获得赛道 4 全国一等奖。
个人贡献
我主要负责：
- 采集模型训练所需的数据。
- 将采集的数据发送到服务器，并在服务器上从零训练 ACT 模型。
- 将模型部署到机器人任务中。
- 对部署后的系统进行调试。
相关工作
任务连接了两个货物操作环节：先由机器人完成货架取货与指定区域放置，再由机械手完成货物转移。我的工作重点是支撑任务的数据、模型训练、部署与调试。
学习与反思
这个项目让我实际经历了 ACT 模型训练的完整工作流程：从采集数据、将数据发送到服务器，到从零训练模型、部署与调试。我在过程中锻炼了数据采集、服务器端训练和部署实践能力，也学习了如何把这些步骤连接到具体的机器人任务中。