你好!我是 张量 (Tensor-0)。
目前研究方向为具身人工智能(Embodied AI),具体包括足式机器人的强化学习 (RL) Locomotion 算法及其物理实体部署。
工程实践中,将运动学闭链解算、LQR/VMC 等数学方法与 AI 算法结合,应用于双足平台与高机动底盘的运动控制。
欢迎与科研团队及开源同行交流与合作。
针对 Roo Code 的多模型本地代理适配层。基于 FastAPI 构建本地服务拦截与重构 HTTP 请求,动态注入缺失参数,解决 Roo Code 与 DeepSeek 等模型间的 API 格式不兼容及上下文截断问题。包含跨平台自动化安装与测试脚本。
A local proxy adapter for Roo Code based on FastAPI. Intercepts and restructures HTTP requests, dynamically injecting missing fields to resolve API incompatibilities with models like DeepSeek. Includes cross-platform automation scripts.
针对达妙 19 自由度双足机器人的 RL 步态优化框架,包含 Isaac Gym 仿真环境搭建、策略训练,以及 ONNX 模型在物理硬件上的 Sim-to-Real 部署。
RL-based gait optimization framework for the Damiao 19-DOF bipedal humanoid robot — Sim-to-Real deployment and high-DOF policy control.
针对 RoboMaster 偏置并联机构的 C++ 底层控制框架,实现了闭链运动学正逆解算器与基于 LQR/VMC 的倒立摆平衡控制器。
C++ motion control framework for RoboMaster biased parallel infantry chassis — Kinematics solvers & LQR/VMC balance control.
摩擦轮飞镖发射时序与转速控制器,通过 PID/ADRC 实现双电机同步闭环,以有限状态机 (FSM) 管理发射流程。
Friction-wheel dart launch controller — PID/ADRC dual-motor sync & FSM-based sequencing.
自主检索机器人系统,集成了 OpenCV 视觉检测、机械臂抓取控制与底盘驱动。
Full-stack autonomous retrieval robot — OpenCV vision pipeline, robotic arm control & chassis drive integration.
I study Robotics Engineering with a focus on Embodied AI. My work covers high-speed combat robotic systems and high-DOF bipedal locomotion, with current emphasis on Reinforcement Learning (RL) for sim-to-real transfer of articulated robots and robust dynamic gait generation.
Open to collaborations on open-source robotics, RL algorithms, and related technical work.
A local proxy adapter for Roo Code, built on FastAPI to intercept and restructure HTTP requests. Dynamically injects missing fields to resolve API format incompatibilities and context truncation between Roo Code and models like DeepSeek. Includes cross-platform automation scripts for installation and testing.
RL-based gait optimization framework for the Damiao 19-DOF bipedal humanoid robot. Covers Sim-to-Real deployment and high-DOF policy control on NVIDIA Orin platforms.
C++ motion control framework for RoboMaster biased parallel infantry chassis. Implements non-linear kinematics solvers and LQR/VMC balance control.
Launch sequence controller for friction-wheel dart systems, with dual-motor speed synchronization (PID/ADRC) and finite-state-machine sequencing.
Full-stack autonomous retrieval system integrating OpenCV-based object detection with STM32 chassis control.
A workspace for daily research logs, debugging notes, and math derivations for embodied AI.