Frenet Optimal Trajectory

Problem Statement

In corridor-like navigation, curvilinear coordinates simplify trajectory generation relative to a reference path. Frenet planning separates longitudinal and lateral behavior, enabling efficient candidate generation and selection.

Model and Formulation

State in Frenet frame:

$$s:\text{arc length along reference},d:\text{lateral offset}$$

Candidates are built from polynomial models:

• longitudinal s(t) via quartic/quintic forms
• lateral d(t) via quintic forms

Total cost aggregates smoothness, offset, speed-tracking, and collision penalties:

$$J=w_d{J}_d+w_s{J}_s+w_c{J}_{collision}$$

Algorithm Procedure

1. Generate candidate lateral and longitudinal profiles.
2. Transform each candidate back to Cartesian coordinates.
3. Evaluate dynamic, collision, and comfort constraints.
4. Select minimum-cost feasible trajectory.

Tuning Guidance

• Increase collision term when operating near dense obstacles.
• Penalize lateral offset more heavily for narrow corridors.
• Short horizons improve responsiveness but can increase myopic behavior.

Failure Modes and Diagnostics

• Poor reference-path quality leads to unstable Frenet projections.
• Candidate set too small misses globally better maneuvers.
• Fast curvature changes can violate dynamic feasibility.

Implementation and Execution

python -m uav_sim.simulations.trajectory_planning.frenet_optimal

Evidence

References

• Werling et al., Optimal Trajectory Generation for Dynamic Street Scenarios in a Frenet Frame (2010)
• Bender et al., The Urban Challenge: Integrating Planning and Control

Related Algorithms

• A* 3D
• Quintic Polynomial