Optimize association algorithm with better assignment (post-MVP)

[jonnyspicer]

Overview

Replace greedy nearest neighbor association with optimal assignment algorithm.

Priority Revision

DOWNGRADED from MVP Critical → Post-MVP Enhancement

Reason: The current greedy association may be adequate for MVP. The real problem is having any radar-only association (#20), not optimizing the assignment algorithm.

Current State

• Simple greedy nearest neighbor association in Stone Soup tracker
• Basic cost matrix with distance-based costs
• Works adequately with current ADS-B data

Target State (Post-MVP)

• Optimal assignment algorithm (Hungarian algorithm)
• Better cost metrics incorporating uncertainty
• Improved association accuracy

Dependencies

• BLOCKED by Implement RadarOnlyAssociator for pure radar tracking #20 - Need basic radar-only association working first
• Should be evaluated after radar-only operation is proven

Reality Check

The flowchart shows "Nearest Neighbors" as part of the association process, suggesting the current approach may be acceptable. The issue isn't the assignment algorithm, it's the lack of radar-only association capability.

Benefits (Post-MVP)

• Optimal assignments instead of greedy selection
• Better overall association quality
• Foundation for complex multi-target scenarios

Note: This is an optimization of association that doesn't exist yet for radar-only mode. Get basic radar association working first, then optimize.