Hohoo AI Journey

Three tracks, one practice path

From software engineering to AI and embodied intelligence. Learn, validate and build in a virtual world around one long-term project.

M1 · Not started

Next: start with one concrete task

Start with validated robot tasks and test the agent with fake tools, then connect ROS2 and a virtual robot.

Natural language → robot task

Roadmap

Planned effort, not completion

01 / 25%

AI Application Engineering

  1. LLM API / Prompt
  2. Structured Output / JSON Schema
  3. Tool Calling / Agent Loop
  4. RAG / Memory
  5. Evaluation / Observability

02 / 15%

LLM & Multimodal

  1. Tokenizer / Token
  2. Embedding
  3. Transformer / Self Attention
  4. Context Window / Inference
  5. Multimodal / Vision Encoder
  6. VLM / VLA

03 / 60%

Embodied AI & Robotics

  1. Python / Linux / C++17
  2. ROS2
  3. MuJoCo / Gazebo
  4. Robot Data
  5. ManiSkill / ACT
  6. VLA / Sim2Real
Agent + VLM + Robot Data → Embodied Agent → VLA

Application skills and visual understanding converge in robot perception, decisions and actions.

M10 →

Simulation first, starting in a virtual world

No robot or dedicated hardware purchases to get started. Choose environments by learning stage rather than installing every simulator. Check hardware and compute requirements before each experiment.

  1. VL01

    MuJoCoPlanned
  2. VL02

    Gazebo + ROS2Planned
  3. VL03

    ManiSkillPlanned
  4. VL04

    Isaac Sim / Isaac LabFuture
Target project architecture · hohoo-embodied-agent

Target task: put the red cube on the table into the blue box.

  1. User
  2. Chat UI
  3. LLM
  4. AI Agent
  5. Task planner
  6. VLM
  7. Robot tools
  8. ROS2
  9. Policy
  10. Simulation engine
  11. Virtual robot
  12. Observation
  13. Robot data platform
  14. Dataset
  15. LeRobot Dataset
  16. Training

Robot data platform

  1. Virtual robot
  2. RGB / Depth / Robot state / Joint state / Action
  3. Data collector
  4. Time synchronizer
  5. Episode builder
  6. MP4 + Parquet
  7. Dataset
  8. Data quality check
  9. LeRobot Dataset

Two loops: acting and learning

Robot execution loop

  1. Environment
  2. Observation
  3. Perception
  4. Reasoning & planning
  5. Policy
  6. Action
  7. Environment

Data and policy iteration

  1. Episode
  2. Dataset
  3. Training
  4. New policy
Later, towards the real world · Sim2Real

The reality gap includes lighting, textures, friction, camera pose, sensor noise, object positions and robot dynamics. Study domain randomization and real-world evaluation later.

  1. Simulation engine
  2. Training
  3. Evaluation
  4. Domain randomization
  5. Robust policy
  6. Real robot

Phase 2: advanced simulation & transfer

Future
  • Isaac Sim
  • Isaac Lab
  • Synthetic Data
  • Domain Randomization
  • Parallel Simulation
  • Sim2Real
  • TensorRT
  • Jetson
  • Robot Deployment

Phase 3: real robot validation

Future
  • Robot Arm
  • Camera
  • Teleoperation
  • Dataset Collection
  • ACT Deployment
  • VLA Deployment
  • Sim2Real Evaluation

Physical hardware belongs to later validation; it is not a prerequisite for starting.

Milestones

  • Not started
  • Learning
  • Building
  • Completed
  1. M1Natural language → robot taskNot started

    Validation goal

    Convert “put the red cube into the blue box” to a structured task, validate its schema and reject invalid actions or parameters.

    Prerequisites

    Knowledge & tools

    • LLM API
    • Prompt
    • System Prompt
    • Streaming
    • Structured Output
    • JSON Schema
    Existing foundation · Calling an LLM with Java 8: from a first request to multi-turn chat

    Next stage

  2. M2Robot agent & fake toolsNot started

    Validation goal

    Validate the agent loop with fake tools, recording calls, failures and stop conditions; no physical robot control.

    Prerequisites

    Knowledge & tools

    • Tool Calling
    • Agent Loop
    • get_scene
    • find_object
    • move_to
    • grasp
    • place
    • get_robot_state

    Next stage

  3. M3Agent → ROS2Not started

    Validation goal

    Replace Python fake tools incrementally with ROS2 tools or a robot bridge; verify calls, feedback and cancellation.

    Prerequisites

    Knowledge & tools

    • Linux
    • Python
    • C++17
    • CMake
    • GDB
    • Docker
    • ROS2
    • Node
    • Topic
    • Service
    • Action
    • Launch

    Virtual labs

    Next stage

  4. M4ROS2 → virtual robotNot started

    Validation goal

    Connect MuJoCo first, adding Gazebo later as needed; verify control and observation from agent through ROS2 to a virtual robot.

    Prerequisites

    Knowledge & tools

    • MuJoCo
    • Gazebo
    • ROS2
    • Robot Bridge

    Virtual labs

    Next stage

  5. M5Virtual pick & placeNot started

    Validation goal

    Use a simulated arm in MuJoCo to place the red cube in the blue box, recording trajectories, controls and failures.

    Prerequisites

    Knowledge & tools

    • Joint
    • Link
    • DOF
    • Pose
    • Coordinate Frame
    • Joint Space
    • Cartesian Space
    • FK
    • IK
    • Jacobian
    • Trajectory
    • Controller

    Virtual labs

    Next stage

  6. M6Robot data platformNot started

    Validation goal

    Collect RGB, robot and joint states, actions, timestamps, tasks and success labels; build episodes and validate synchronization, encoding, storage and quality.

    Prerequisites

    Knowledge & tools

    • Observation
    • Action
    • State
    • Frame
    • Episode
    • Trajectory
    • Dataset
    • Timestamp
    • FFmpeg
    • MP4
    • Parquet
    • PyArrow
    • MCAP
    • Object Storage
    • Data Quality
    • LeRobot Dataset

    Virtual labs

    Next stage

  7. M7Robot CopilotNot started

    Validation goal

    Retrieve evidence from robot manuals, ROS2 docs, error codes and operation logs for grounded diagnosis; cite sources and test unsupported questions.

    Prerequisites

    Knowledge & tools

    • Embedding
    • RAG
    • Vector Database
    • Agent Memory
    • Context Engineering
    • Evaluation
    • Observability

    Next stage

  8. M8Virtual vision & VLMNot started

    Validation goal

    Move from manually supplied scene JSON to virtual camera RGB/depth input; compare visual judgments against scene configuration and record limits.

    Prerequisites

    Knowledge & tools

    • Camera
    • RGB
    • Depth
    • Tokenizer
    • Token
    • Transformer
    • Self Attention
    • Context Window
    • Inference
    • Multimodal
    • Vision Encoder
    • VLM

    Virtual labs

    Next stage

  9. M9Robot learningNot started

    Validation goal

    Train or reproduce a policy using robot data, LeRobot, PyTorch and ACT; record action-chunk execution and evaluation in a virtual robot.

    Prerequisites

    Knowledge & tools

    • PyTorch
    • Behavior Cloning
    • Imitation Learning
    • LeRobot
    • ACT
    • ManiSkill
    • Diffusion Policy

    Virtual labs

    Next stage

  10. M10VLA virtual robotNot started

    Validation goal

    When resources permit, select SmolVLA or a suitable lightweight VLA to map vision, language and robot state to actions; validate in simulation first.

    Prerequisites

    Knowledge & tools

    • RT-1
    • RT-2
    • Octo
    • OpenVLA
    • SmolVLA
    • π0
    • π0.5
    • GR00T
    • VLA

    Virtual labs

Suggested pace

An adjustable eight-week plan, not a completion promise. No start date is set.

  1. Week 1
  2. Week 2
  3. Week 3
  4. Week 4
  5. Week 5
  6. Week 6
  7. Week 7
  8. Week 8

Existing foundation

Calling an LLM with Java 8: from a first request to multi-turn chat

Review requests, response parsing and chat history in the published tutorial. It does not establish completion of a robot milestone.

Explore later

Build an end-to-end path first, then expand around concrete problems.

  • SLAM
  • MPC
  • WBC
  • Advanced Dynamics
  • Humanoid Locomotion
  • CUDA Kernel
  • Distributed VLA Training
  • World Model Training
  • Advanced RL
  • Real Robot Hardware
  • Isaac Sim Optimization
  • Isaac Lab Large-scale Training
  • TensorRT Optimization