Each prey animal emits a scent field that falls off as a Gaussian with spread σ. A marten senses the analytic gradient of the sum of every active prey's field and steers toward it — this is the same biased-random-walk (chemotaxis) model used to describe scent-tracking foraging in real predators:
∇C(p) = Σ_i (x_i − p) · exp(−|x_i − p|² / 2σ²) / σ²
heading → blend(current heading, atan2(∇C), scent sensitivity)
Movement costs energy quadratically in speed (E' = −(metabolism + k·v²)), so sprinting toward a scent is expensive — a real optimal-foraging trade-off. A marten below its energy floor stops hunting and rests until it recovers. When several martens reach the same prey within one step, interference competition decides the outcome: catch probability falls as 1 / (1 + competition·contenders), and every loser still pays the metabolic cost of the chase, exactly as site-defense and kleptoparasitism reduce net intake in crowded real hunting grounds.
- Scent sensitivity — how strongly the gradient overrides random exploration; low values make search nearly random, high values make it a direct chase.
- Competition pressure — how much a shared target degrades every contender's catch odds and energy budget.
- Population — more martens sharing the same prey pool raises the contest rate and lowers average intake per hunter, a direct density-dependence effect.