Skip to content
SDB
Fundamentals of Robotics and Automation

2 hours

System-level Architecture of Robots

Levels of Autonomy and Control Hierarchies

Subhendu Datta BhowmikRobotics Tutorials

Robot System Architecture Overview

A robotic system is a complex integration of hardware and software subsystems working in concert. Understanding the architecture helps engineers design, debug, and optimize robot behavior. The architecture defines how sensing, processing, decision-making, and actuation are organized and interconnected.

Primary Subsystems:

  1. Sensing/Perception: Gathering information about the robot and environment
  2. Cognition/Processing: Interpreting sensor data, planning actions
  3. Actuation/Execution: Physically executing planned motions
  4. Communication: Interfacing with operators, other robots, and infrastructure
  5. Power: Powering all subsystems reliably

Key Sensors in Robotic Systems

Sensor TypeMeasurementTechnologyApplicationRange/Accuracy
EncoderJoint angle/positionOptical/MagneticJoint feedback< 0.001° resolution
IMUAcceleration, rotationMEMSNavigation, balance< 0.1° drift/hr
Camera (RGB)2D visual dataCMOSObject detection, vision1-12 MP, 30-120 fps
Depth Camera3D point cloudStructured light/ToF3D perception0.3-8m, ±2mm
Lidar3D environment scanLaser rangingSLAM, obstacle avoidance0.1-200m, ±2cm
Force/TorqueContact forcesStrain gaugeManipulation, assembly< 0.1N resolution
GPS/RTKGlobal positionSatellite signalOutdoor navigation±2cm (RTK)
UltrasonicProximitySound echoCollision avoidance2cm-4m, ±1cm

Control Architectures

1. Hierarchical (Deliberative) Architecture Classical AI approach: Sense → Plan → Act. A top-level planner receives world state, computes optimal plan, sends commands down through layers. Clean and predictable but slow for dynamic environments. Used in early space rovers and factory robots.

2. Reactive (Behavior-Based) Architecture Rodney Brooks' Subsumption Architecture (1986): No central world model. Multiple layers of simple behaviors run in parallel; higher layers can inhibit/subsume lower ones. Fast response to environment changes. Used in early autonomous vehicles and Roomba.

3. Hybrid Architecture Combines deliberative planning for high-level goals with reactive behaviors for low-level obstacle avoidance. Most modern robots use this approach. Example: autonomous cars plan routes using maps (deliberative) while reacting to pedestrians in real-time (reactive).

4. Three-Layer Architecture (Common Standard)

  • Deliberative Layer: Long-term planning, task sequencing, goal management
  • Executive Layer: Translates plans into action sequences, monitors execution
  • Reactive Layer: Real-time sensor-motor loops, safety behaviors

5. ROS (Robot Operating System) ROS is the de-facto middleware for robotics. It provides a node-based communication framework (topics, services, actions), hardware abstraction, device drivers, and a rich ecosystem of packages. ROS 2 (2017+) adds real-time support and improved security.

Levels of Robot Autonomy (SAE-inspired framework)

LevelNameHuman RoleRobot RoleExample
0No AutonomyFull control, robot is a toolExecute exact commandsTeleoperated surgical robot
1Driver AssistanceMonitor, correct robotAssist with specific subtasksIndustrial robot with safety override
2Partial AutomationSupervise, handle exceptionsAutomate specific tasksWarehouse sorting robot with human oversight
3Conditional AutomationAvailable for takeover when askedFull task execution in defined conditionsSelf-driving car on highway
4High AutomationOptional, can disengageHandle all scenarios in operational domainAutonomous taxi in geo-fenced area
5Full AutonomyNo intervention neededFully autonomous in all conditionsMars rover (delayed communication)

Key Architectural Components

  1. 01

    World Model / Environment Map: Internal representation of the robot's surroundings — occupancy grids, semantic maps, scene graphs

  2. 02

    Localization Module: Determining the robot's position using SLAM, GPS, or visual odometry

  3. 03

    Path Planner: Computing collision-free paths from start to goal — A*, Dijkstra, RRT, PRM algorithms

  4. 04

    Task Planner: High-level sequencing of subtasks to achieve goals — STRIPS, HTN planning, PDDL

  5. 05

    Trajectory Generator: Converting discrete paths to smooth, physically executable trajectories with velocity profiles

  6. 06

    Controller: Closed-loop feedback control — PID, computed torque, impedance control

  7. 07

    Safety Monitor: Real-time collision detection, force limiting, emergency stop — safety-critical real-time systems

  8. 08

    Communication Interface: ROS topics/services, OPC-UA for industrial systems, WebSocket for cloud connectivity

Fundamentals of Robotics and Automation