Junhyeok Lee

Driven by vision. Defined by persistence.

B.S. in Automotive EngineeringKookmin University, Senior

LinkedInInstagram
© 2026 Junhyeok Lee · Seoul
Junhyeok Lee
Autonomous Driving
Back to Stack
04

Planning

IN PROGRESS

From global route finding to local trajectory optimization — generating safe, smooth, and dynamically feasible paths that account for obstacles, road geometry, and traffic rules.

A*RRT*Behavior TreeTrajectory OptimizationOccupancy GridROS2
A*
Global
RRT*
Local
BT
Decision
10Hz
Replan Rate

Global Path Planning

A*-based search on a precomputed road network graph generates the high-level route from start to goal. Waypoints are extracted and smoothed using cubic spline interpolation to create a drivable reference path for the local planner.

Local Trajectory Planning

RRT*-based sampling generates candidate trajectories that respect vehicle kinematics and avoid detected obstacles. Trajectories are scored by a cost function balancing safety margin, path smoothness, and deviation from the global reference path.

Behavior Decision-Making

A behavior tree architecture handles high-level driving decisions — lane following, lane changing, intersection handling, and emergency stops. The behavior layer sits between perception and planning, translating environmental context into planning objectives.

© 2026 Junhyeok LeeSeoul, Korea