64. Embodied AI

Mini-Project: Embodied AI: Grid Navigation with BFS Pathfinding and Dynamic Replanning

A sense-plan-act loop that navigates a grid world using BFS pathfinding, handles dynamic obstacles appearing mid-execution by replanning from the current position, and escalates to human assistance when no viable path exists.

View on GitHub

---

Description

Embodied AI refers to AI agents that interact with the physical world through sensors and actuators -- robots, drones, autonomous vehicles, or smart devices. Unlike purely digital agents, embodied agents must handle continuous state spaces, real-time constraints, physical safety, and sensor noise. They combine perception (vision, LIDAR, touch), reasoning (planning, navigation), and action (motor control) in a closed loop.

Architecture Diagram

flowchart TD
    A[Physical Environment] --> B[Sensors]
    B --> C[Perception Module]
    C --> D[World State Estimate]
    D --> E[Reasoning / Planning]
    E --> F[Action Selection]
    F --> G[Motor Control]
    G --> H[Actuators]
    H --> A

    I[Language Interface] --> E

    style A fill:#4CAF50,color:#fff
    style C fill:#2196F3,color:#fff
    style E fill:#FF5722,color:#fff
    style G fill:#9C27B0,color:#fff

Key Models/Systems

System	Domain	Description
Boston Dynamics Spot	Quadruped Robot	Autonomous navigation and manipulation
Tesla FSD	Autonomous Driving	Vision-based self-driving
Google Aloha	Manipulation	Bimanual robotic manipulation
Figure 01	Humanoid	General-purpose humanoid robot