Skip to content

Find Your Reference Design#

Use the decision flowchart to identify which reference configuration matches your deployment scenario, then explore real-world examples that demonstrate each configuration.

Decision Flowchart#

flowchart TD
    START[Start] --> Q1{Purpose?}

    Q1 -->|Production| Q2{Environment?}
    Q1 -->|Development/Education| DEV[See Development Platforms]

    Q2 -->|Indoor only| INDOOR[Indoor Configuration]
    Q2 -->|Outdoor| Q3{Budget?}
    Q2 -->|Both indoor/outdoor| Q6{Size constraint?}

    Q3 -->|Limited| BUDGET[Budget Configuration]
    Q3 -->|Moderate| CAMPUS[Campus Configuration]
    Q3 -->|Flexible| Q4{Max capability needed?}

    Q4 -->|Yes| HIGH[High-Performance Configuration]
    Q4 -->|No| CAMPUS

    Q6 -->|Yes| COMPACT[Compact Configuration]
    Q6 -->|No| HIGH

Reference Configurations#

These are conceptual configurations optimized for different deployment scenarios. Each defines recommended components, performance targets, and ODD coverage.

Configuration Best For Environment Localization Key Feature
Campus Standard outdoor deployment Paved, GPS available RTK GNSS Balanced cost/capability
Indoor Warehouse, factory GPS-denied LiDAR SLAM + markers No GNSS dependency
High-Performance Research, complex scenarios Any GNSS + SLAM fusion Maximum capability
Budget Prototyping, simple routes Paved, GPS available Single GNSS Cost-optimized
Compact Small vehicles Indoor/outdoor Camera SLAM Space-constrained

For detailed component specifications, see Design Choices by Example.

Real World Examples by Configuration#

Campus Configuration#

These production deployments demonstrate the Campus configuration in action:

TalTech iseAuto#

Campus shuttle at Tallinn University of Technology, Estonia.

Aspect Details
Platform Custom electric shuttle
Sensors 3D LiDARs, automotive cameras, GNSS/IMU
Connectivity Private 5G network
Status Regular public transport service on campus

Software Stack:

Component Details
Autoware Autoware.universe
Architecture ISEAUTO Paper (PDF)

Highlights: First self-driving vehicle in Estonia; Level 4 shuttle built in one year using Autoware; V2X and teleoperation research platform.

Links: iseAuto Project | TalTech Autoware Foundation

NC A&T Aggie Auto#

Autonomous GEM shuttles at North Carolina A&T State University.

Aspect Details
Platform GEM e6 electric vehicles
Sensors Multi-sensor LiDAR suite, NovAtel GNSS with dual antennas
Drive-by-wire AutonomouStuff PACMod
Status Public pilot program completed

Software Stack:

Component Details
Autoware Autoware (open source)
Integration AutonomouStuff Speed and Steering Control (SSC)

Highlights: SAE Level 4 autonomy; connected autonomous vehicle (CAV) testbed; 2-mile rural test track; public service connecting campus to downtown Greensboro.

Links: Aggie Auto Project | AutonomouStuff Case Study

KingWayTek Micro LSV#

Micro self-driving vehicles deployed in Taiwan for passenger transport and cargo delivery.

Aspect Details
Platform Custom micro EV (3380×1350×1850mm)
Max Speed ≤15 km/h
Autonomy Level L4
Capacity 4-5 passengers or 350kg cargo
Range 50km

Technology Stack:

Component Details
Sensors LiDAR, radar, cameras
Maps HD Maps (centimeter-level precision)
Communication C-V2X (4G/5G)

Highlights: First Taiwan Lantern Festival self-driving vehicle (2024); 400 trips serving 1,000+ passengers in 16 days; deployed at 13+ locations including TSMC Southern Taiwan Science Park.

Links: KingWayTek | Self-Driving Solutions | Introduction (PDF)

Development Platforms#

These platforms are designed for algorithm development, education, and prototyping:

Go-Kart (1/3 Scale)#

Human-rideable platform for software development and testing, developed by the Autoware Center of Excellence at University of Pennsylvania.

Component Choice
Platform TopKart chassis
ECU x86 laptop + NVIDIA GPU
LiDAR Ouster OS1
Camera OAK-D (depth + on-device AI)
GNSS RTK-GNSS

Software Stack:

Component Details
ROS 2 Foxy / Humble
Control Pure Pursuit, MPC
Sensor Code gokart-sensor
MCU Code gokart-mechatronics

Best For: Algorithm development, SAE Level 0-3 testing

Documentation: Go-Kart Details | Project Docs | GitHub

RoboRacer (1/10 Scale)#

Lowest-cost entry point for learning autonomy.

Component Choice
Platform Traxxas Slash 4x4
ECU Jetson Xavier NX
LiDAR Hokuyo UTM-30LX (2D)
Camera ZED 2 or RealSense (optional)
Motor Control VESC 6 MK III

Software Stack:

Component Details
ROS 2 Foxy
Simulator RoboRacer Gym
Source Code GitHub

Best For: Education, racing competitions, algorithm prototyping

Documentation: RoboRacer Details | RoboRacer Portal | RoboRacer Learn

AutoSDV#

Home-buildable 1/10 scale platform with comprehensive documentation, developed by NEWSLab at National Taiwan University.

Model LiDAR Localization Connectivity
Base None (camera only) Vision only Standard
360° LiDAR Velodyne VLP-32C NDT (full autonomous) Standard
Solid-State Seyond Robin-W (150m) Development needed Standard
Connected Seyond Robin-W Development needed 5G (Ataya/MOXA)

Core Platform: Jetson AGX Orin 64GB, ZED X Mini stereo camera, Tekno TKR9500 chassis

Software Stack:

Component Details
Autoware Autoware.universe
ROS 2 Humble
Source Code GitHub

Best For: Self-learners, home projects, university courses

Documentation: AutoSDV Details | AutoSDV Book | GitHub

Decision Criteria Reference#

Use these tables to refine your configuration choice based on specific requirements.

By Environment#

Your Environment Recommended Configuration
Outdoor, paved, GPS available Campus or Budget
Indoor, structured, GPS-denied Indoor
Mixed indoor/outdoor Compact or High-Performance

By Speed Requirement#

Target Speed Recommended Configuration
<5 kph Indoor, Compact
5-10 kph Budget, Campus
10-15 kph Campus, High-Performance
>15 kph High-Performance (research only)

By Budget#

Budget Level Recommended Configuration
Limited Budget
Moderate Campus
Flexible High-Performance

Other Example Designs#

For detailed documentation on specific platforms:

Next Steps#