Orangutan Canopy Planning Lab (3D)

Fly through a rainforest canopy: orangutans pick fruit trees, plan A* routes along lianas and sway trees across gaps. Every hop is priced in joules.

Switch to 2D

Presets

Playback

Time and troop

Speed up time or add orangutans; they all share the same fruit.

Forest and fruit

More trees and a smaller gap limit mean a denser web of branches.

Orangutan and weather

Mass, effort, memory and wind change the price of every hop.

Camera

Layers

Quality

Model checks

The science

Orangutans are the largest tree-living apes. They rarely come to the ground, so a day is a route through the canopy: which fruit tree next, and by which chain of branches? This lab turns that into a graph problem with an energy price on every link. Trees are nodes; an edge exists where crowns touch or the gap is small enough to bridge.

Cost of one hop (joules)

J = c m d + m g max(0, dz) + 0.25 m g max(0, -dz) + Prest m d / v + Jsway / p

  • c m d: clambering cost, c joules per kg per metre over the centre-to-centre distance d (slider; the default 10 J/kg/m is an illustrative value, arboreal travel is several times costlier than walking).
  • m g dz: work against gravity when climbing; a quarter of that is paid for a controlled descent.
  • Prest m d / v: the resting metabolism (1.8 W/kg) keeps burning while moving at v = 0.55 m/s.
  • Gap crossing by tree sway. If the crown gap exceeds the arm reach (0.9 m), the orangutan pumps the tree at its natural frequency until the tip swings far enough: amplitude a = gap - 0.9 m. The tree is treated as a pendulum of effective length L (6-12 m), so the lift is m g (L - sqrt(L2 - a2)), the period is T = 2 pi sqrt(L/g) (1 + th2/16 + 11 th4/3072) with th = asin(a/L), and the number of pumping cycles is ceil(a / 0.4 m). Muscle efficiency is 25 %, so Jsway = lift / 0.25 + resting power x pumping time.
  • Wind. A gust scatters the tip by sigma = 0.22 m per m/s of wind (a little more for big swings). The grab succeeds with p = erf(0.9 / (sigma sqrt 2)). A failed swing is simply repeated (the orangutan re-grips; no injuries are modelled), so the expected cost is Jsway / p, a geometric series of retries.

Planning

Each orangutan scores every tree it believes has fruit by (Efruit Fest - Jroute) / (1 + Jroute / 20 kJ), with Efruit = 2.8 MJ per kg, and walks the cheapest route found by A* over the canopy graph. The heuristic is the straight-line distance times the cheapest possible cost per metre (locomotion plus resting cost). No real edge is cheaper than that, so the heuristic is admissible and A* returns the same optimum as an exhaustive Dijkstra search; the check panel verifies this on random pairs.

Memory and fruit

Fruit regrows as dF/dt = (Fmax x abundance - F) / tau with tau = 8 h / regrowth. An orangutan only knows a tree's true stock when it is there; between visits its estimate decays toward a prior of 0.9 kg with time constant 15 min / (1 - memory), so low memory means stale or wrong beliefs and wasted trips to empty trees. Eating takes about 0.7 kg per feeding hour.

What is simplified

Real canopies are three-dimensional tangles of lianas; here every tree is one node and one effective pendulum. Trees are cantilever beams rather than point pendulums, sway amplitude pumping per cycle is an assumption, and all metabolic coefficients are illustrative round numbers (they set the shape of the trade-off, not field-measured energetics). Real orangutans also use social information, fruit scent, season and weather forecasts that are not modelled here. Orangutans are critically endangered: this is an educational model of how they might think about a route, not a tool for conservation planning.

What to try

  • Click two trees. The cyan route is the cheapest A* path; compare it with the greedy "nearest to the goal" route in the readout. When does greedy lose?
  • Press Storm: strong gusts: wide sways fail often, the retries multiply their cost, and orangutans start taking long detours over touching crowns.
  • Raise the body mass to 78 kg: climbing and sway cost rise in proportion, and the cheapest route shifts toward flat, touching crowns.
  • Set memory to 0, then to 100 %, and tick "Memory beliefs" in the layers: yellow rings are what orangutans believe, orange rings are the truth. Watch the energy graph.
  • Drop fruit abundance to 0.3: net energy turns negative and the troop must range further.
Loading model...
Loading Blender model...
Cost of one hop versus crown gap
Energy balance and fruit stock
Tree sway: pendulum period (check)
A* route cost versus straight-line bound
Made with 🟠 Blender · Three.js