Generate and track reference lines and lane topologies based on various methods such as lane line perception, semantic maps, and traffic flow clustering, as well as reconstruct real-time local topological relationships of roads and understand the full-scene road environment.
Understand and infer the regulatory boundaries of the road environment, realize drivable area reasoning, and model intersection spaces.
Design and develop a matching and fusion framework based on real-time perception and prior semantic maps to enhance the spatiotemporal understanding of the road environment by the autonomous driving system in complex scenarios.
Identify occluded areas and calculate risk speed limits.
Perform fusion matching and recognition of lane types, and complete the matching between traffic lights and lanes.
Calculate lane static navigation based on SD navigation information.
Job requirements
Major in computer science, automation, mathematics, automotive engineering or other related fields.
Have a foundation in computational graphics, be familiar with 3D set element processing, and be able to model and reason about complex environments.
Be familiar with optimal road network matching algorithms (KM, VF2, PDA) and have algorithm experience in optimal matching of lanes, traffic lights, etc.
Be familiar with multi-frame smoothing and optimization algorithms for reference lines.
Proficient in C++, and familiar with various optimization algorithms and algorithm libraries (Eigen, OSQP).
Having a knowledge background in NN models for local map road topology generation such as MapTr is a plus.