Skip to content

Design Choices by Example#

This page provides detailed component specifications for each reference configuration. Use these as starting points for your own build.

Configuration Overview#

Configuration ECU Primary LiDAR Localization
Campus AGX Orin 32/64GB Ouster OS1-64 RTK GNSS
Indoor AGX Orin Ouster OSDome LiDAR SLAM + markers
High-Performance x86 + RTX 4090 OS2 + OSDome GNSS + SLAM fusion
Budget Orin NX 16GB Ouster OS0-32 Single GNSS
Compact Orin Nano/NX Ouster OSDome Camera SLAM

Campus Configuration#

Balanced configuration for outdoor paved environments with pedestrian interaction.

Component Choice Rationale
ECU NVIDIA AGX Orin 32/64GB 275 TOPS, 15-60W, proven Autoware support
LiDAR Ouster OS1-64 120m range, 64 channels, reliable ROS 2 driver
Cameras 2-4 wide-angle Object classification, sign detection
GNSS Dual-antenna RTK Cm-level accuracy, heading estimation
IMU Integrated or standalone Motion estimation, sensor fusion

Performance Targets:

Metric Value
LiDAR detection 10 Hz
Camera detection 15 Hz
Planning 10 Hz
Control 50 Hz
End-to-end latency <200 ms
Power (ECU) <50W

ODD Coverage: LSA-CAM-0001 to 0040 (normal driving through intersections)

Real World Examples: TalTech iseAuto, NC A&T Aggie Auto

For detailed ECU specifications, see ARM-based ECUs.

Indoor Configuration#

Optimized for GPS-denied warehouse and indoor structured environments.

Component Choice Rationale
ECU AGX Orin Same compute as Campus
LiDAR Ouster OSDome 180 deg FOV for narrow aisles, 45m range
Cameras Wide-angle + depth SLAM features, close-range detection
GNSS None GPS-denied environment
Localization LiDAR SLAM + markers AprilTags or reflectors for accuracy

Key Differences from Campus:

  • No GNSS dependency
  • Wide-FOV sensors for tight maneuvering
  • Marker infrastructure for localization accuracy
  • Lower speed limits (5-10 kph typical)

For indoor scenarios, see Indoor ODD.

High-Performance Configuration#

Maximum capability for research and complex scenarios.

Component Choice Benefit
ECU x86 Xeon/Core + RTX 4090 Maximum compute, 24GB VRAM
LiDAR Ouster OS2 (200m) + OSDome Long range + wide FOV coverage
Cameras 6-8 surround cameras 360 deg visual coverage
GNSS Multi-constellation RTK Maximum accuracy
Localization GNSS + LiDAR SLAM fusion Maximum robustness

ODD Coverage: Full LSA-CAM-0001 to 0082

Trade-offs: Higher cost, power consumption (150-450W), complexity

Budget Configuration#

Cost-optimized with reduced capabilities.

Component Choice Trade-off vs Campus
ECU Jetson Orin NX 16GB Lower compute headroom
LiDAR Ouster OS0-32 or OS1-32 Shorter range (35-90m), fewer channels
Cameras 1-2 cameras Reduced coverage
GNSS Single-antenna Lower heading accuracy

ODD Limitations: LSA-CAM-0001 to 0020 (basic scenarios only)

Best For: Simple fixed routes, low-traffic environments

Compact Configuration#

For small vehicles with space and weight constraints.

Component Choice Rationale
ECU Orin Nano or NX Small form factor, 7-25W
LiDAR Ouster OSDome Compact, wide FOV
Carrier Integrated board Minimizes size and cabling
Cameras 1-2 compact modules Essential coverage
Localization Camera SLAM No external infrastructure

Best For: Sidewalk delivery robots, small indoor AGVs

Component Selection Reference#

ECU Selection by Criteria#

Criteria Recommendation
Balanced cost/performance AGX Orin 32GB
Minimum cost Orin NX 16GB
Maximum compute x86 + RTX 4090
Minimum size/power Orin Nano
Industrial environment Neousys 9160-GC
Wide temperature range Crystal Rugged AVC series

For full ECU specifications:

LiDAR Selection by Application#

Application Recommendation Range Channels
Outdoor campus Ouster OS1-64 90m 64
Long-range outdoor Ouster OS2-128 200m 128
Indoor/warehouse Ouster OSDome 20m 128 (180 deg FOV)
Budget outdoor Ouster OS0-32 35m 32
Compact vehicles Ouster OSDome 20m 128

For sensor options, see Sensors and Actuators.

Localization Method by Environment#

Environment Primary Method Backup/Enhancement
Outdoor, open sky RTK GNSS NDT matching
Outdoor, urban canyon GNSS + LiDAR SLAM Visual odometry
Indoor, structured LiDAR SLAM Marker-based (AprilTags)
Indoor, featureless Marker-based Wheel odometry

Middleware Selection#

Scenario Recommendation Rationale
Wireless operation Zenoh Better packet handling
Wired, lowest latency Cyclone DDS Minimal overhead
Enterprise integration RTI Connext Commercial support

For middleware setup, see Middleware Configuration.

Next Steps#