Hardware Atlas · Special
Component-to-Intelligence Map
Every hardware component plays a specific role in the AI pipeline. This map shows how electronics map to the perception–processing–action loop that every embodied AI system must implement.
Sense
Sensors
→
Process
MCU / Edge
→
Act
Actuators
→
Communicate
Wi-Fi / Serial
→
Power
Battery / PSU
Perception
Sensory input to the policySensors convert physical phenomena into data that an AI policy can reason about. Without sensors, the robot is operating open-loop — blind to the state of the world.
IMU Sensor (MPU-6050)
Orientation, acceleration, angular velocity — the robot's vestibular system
Ultrasonic Sensor
Distance measurement — obstacle detection for navigation
Load Cell + HX711
Force and weight — contact detection, payload measurement
Rotary Encoder
Motor shaft position and velocity — proprioception for closed-loop control
Camera (ESP32-CAM)
Visual input — object detection, visual servoing, scene understanding
Processing
Policy computation and state estimationMicrocontrollers run the control policy — translating sensor observations into actuator commands. The more capable the MCU, the more sophisticated the policy it can execute locally.
Arduino Uno R3
PID control loops, sensor reading, servo PWM generation — simple deterministic policies
ESP32 DevKit V1
Dual-core: real-time control on Core 1, Wi-Fi comms on Core 0. TFLite Micro for edge inference.
Raspberry Pi (external)
Full Linux: PyTorch inference, OpenCV vision, ROS nodes — the brain of complex systems
Action
Policy output executionActuators translate policy outputs (target angles, velocity commands, force setpoints) into physical motion. The fidelity of the action layer determines how precisely a robot can execute a learned behaviour.
Servo Motor (SG90)
Joint angle execution — each DOF in a robot arm maps to one servo position command
DC Motor + Motor Driver
Wheel velocity execution — differential drive translates velocity commands to left/right PWM
Stepper Motor
Open-loop position control — linear actuators, CNC stages where encoder is unaffordable
L298N Motor Driver
Current amplifier — bridges MCU logic (40 mA) to motor current demand (1–3 A)
Communication
Data collection, remote control, multi-agent coordinationCommunication components connect the embedded system to the broader AI pipeline — streaming data for learning, receiving policy updates, and coordinating multi-robot systems.
ESP32 Wi-Fi
HTTP/MQTT: stream sensor data to Python training scripts; receive policy commands
ESP32 BLE
Bluetooth HID for wireless teleoperation; BLE beacons for indoor positioning
Arduino Serial (USB)
Bridge between a Python AI process on a laptop and Arduino actuator control
I2C Bus
Connect multiple sensors (IMU, ADC, display) to one MCU with two wires
Power
Energy budget and system reliabilityPower components determine how long a robot runs, how stable its logic operates, and whether actuators have enough current to execute the commanded actions without browning out the MCU.
LiPo Battery
Energy source — cell voltage, capacity (mAh), and C-rating determine robot runtime and peak current
Voltage Regulator (LDO)
Stable 3.3V or 5V for logic — prevents MCU resets when motor current spikes
Capacitors (decoupling)
100nF across VCC/GND absorbs high-frequency noise from motors — prevents ADC corruption
Flyback Diodes
Protect MCU from back-EMF voltage spikes when motors are switched off
Full System Example — 2-DOF Robotic Arm
Sensors
- MPU-6050 IMU (wrist)
- 2× Rotary encoders (joints)
MCU
- Arduino Uno R3
- Runs PID at 50 Hz
Actuators
- 2× SG90 servo motors
- 1× DC gripper motor
Driver
- L298N for gripper DC motor
- Servo library for joints
Comms
- USB Serial to Python
- Streams joint states at 20 Hz