Adding a directional cost bias to a TSP heuristic reveals better local optima

Follow-up to a TSP solver I posted a few days ago. Added a directional ("wind") cost bias mainly for visual variety — it turned out to change which local optimum the greedy construction phase lands on. Systematic testing across the algorithm's eight strategies found consistent 3–8% improvements in seven of them; the eighth (convex-hull seeding) showed zero variation, which serves as a clean control confirming the effect is real. Also added a terrain mode with variable-speed zones. Paper, video, and a fully independently-verified case study with downloadable instance files here: tsp.uncledroid.app/paper2.pdf. Demo: tsp.uncledroid.app

video youtu.be/7z6nllJW--o