Autoware in Construction: Gap Analysis and LiDAR Perception Toward Off-Road Autonomous Driving
Organizations: Department of Civil and Environmental Engineering, University of California, Irvine (UCI), Irvine, CA, USA · TIER IV, Inc., Shinagawa, Tokyo, Japan
Abstract
Autoware is an open-source autonomous driving software platform widely adopted by researchers and industry developers. Originally developed primarily for public-road applications, including passenger vehicles, taxis, and buses, Autoware is increasingly being extended to off-road environments such as construction and agricultural sites. Construction sites, however, differ fundamentally from public roads and challenge many assumptions underlying conventional autonomous driving systems. They are characterized by unstructured terrain, airborne dust, continuously evolving site conditions, and construction-specific objects. This paper presents lessons learned from ongoing efforts to extend an Autoware-based autonomous driving system to large dump trucks operating at construction sites. We identify gaps between public-road and off-road construction applications across the sensing, mapping, localization, perception, planning, and control modules of the Autoware stack. We then discuss potential solutions for addressing these gaps, with particular emphasis on ongoing LiDAR-based perception pipeline development and findings from field testing. Finally, we propose a roadmap toward end-to-end autonomous driving for construction vehicles.
Figures & tables
| Modules | Public-road capabilities available in Autoware | Off-road construction challenges and requirements |
|---|---|---|
| Sensing | Supports LiDAR, camera, radar, GNSS, and IMU sensors. Standard sensor drivers and sensor calibration routines are also provided. | Dust can cause occlusion. Mud can contaminate sensor surfaces. Large vehicle bodies can create blind spots. IMU measurements are degraded by strong vibrations. Severe vertical vibration requires 3D LiDAR distortion correction. |
| Mapping | Lane geometry and surrounding structures are static, so high-definition maps can be maintained offline. Environments are geometrically feature-rich, with abundant vertical structures. | Construction sites evolve continuously through excavation, grading, and material placement, requiring frequent map updates. |
| Localization | GNSS/RTK can be unreliable in tunnels, urban canyons, and beneath elevated roads. Map-based localization performs well with feature-rich high-definition maps. Planar localization in , , and yaw is often sufficient. | GNSS/RTK is generally reliable under open skies. Map-based localization is challenged by feature-poor, changing environments. Full 6-DOF pose estimation is often required due to slopes and rough terrain. |
| Perception | Supports detection, tracking, and sensor fusion using both non-ML and ML-based approaches. Its ML detectors support objects such as cars, trucks, buses, bicycles, and pedestrians. | Relevant objects include heavy construction machinery, passenger vehicles, construction workers, cones, rocks, berms, and temporary structures. The drivable space is unstructured and must be identified. |
| Planning | Scenario-based, such as lane driving and parking on predefined road networks, governed by traffic rules and dominated by forward driving. | Haulage roads are stable over days to weeks and can be given as waypoints, but loading and dumping areas change continuously and require coordination with excavators and dozers. Traversability must be evaluated to determine optimal trajectories off the haul road. |
| Control and Vehicle Interface | Tracks the planned trajectory by regulating steering and acceleration, typically assuming Ackermann-steered vehicles. | Must account for uneven terrain, wheel slip, steep slopes, and payload changes that alter mass and center of gravity, as well as articulated steering where applicable; hydraulic actuation introduces larger delay and stronger non-linearity than road-vehicle actuators. Additional machine-specific work equipment, such as the dump bed. |