Since July 2025, our research team has been actively exploring intelligent robotics through three complementary technical pillars. This video opens with our work on 3D Reconstruction based on LiDAR—where we process point cloud data to build accurate environmental maps for robot navigation and scene understanding. Next, we present Simulation-Based Robot Training, a core part of our workflow where robots learn and refine their behaviors in virtual environments before deployment, significantly reducing real-world trial costs. Finally, the video highlights our efforts in Reinforcement Learning, where we train control policies that enable robots to adapt to complex, dynamic tasks through trial-and-error interaction with their surroundings. Together, these segments demonstrate the team's comprehensive approach—from perception and virtual training to intelligent decision-making—built on collaborative research and shared experimentation.

    自2025年7月起，我们的研究团队围绕三个互补的技术方向持续开展智能机器人研究。视频开篇展示的是基于激光雷达的3D重建工作——我们对点云数据进行处理，构建精准的环境地图，以支持机器人导航和场景感知。接下来呈现的是基于仿真的机器人训练，这是我们工作流程中的核心环节，让机器人在部署之前先在虚拟环境中学习和优化行为策略，从而显著降低实际试错成本。视频最后聚焦于强化学习方面的探索，我们通过让机器人与环境进行试错交互，训练出能够适应复杂动态任务的控制策略。整个视频展现了团队从感知、仿真训练到智能决策的一体化研究路径，所有进展都源于团队内部的协作攻关与反复实验。