Moose Simulation Project - Modeling Course 2024

This project simulates moose movements and behaviors in a forested area, and models how effectively a drone-based surveying method can estimate the number of moose within that area. The simulation ties into a scenario where a drone equipped with LiDAR and thermal cameras is used to count moose in a given forest region near Tampere and Valkeakoski (east side of highway E12 in Finland). The primary goal is to propose a cost-effective and accurate moose counting method using drones, supported by a simulation model that estimates detection accuracy under varying moose populations.

Original Assignment (in Finnish):
"Laadi tarjous hirvien laskennasta drooneilla Tampereen ja Valkeakosken välisellä, tien E12 itäpuolisella metsäalueella.
Kriteerit onnistumiselle:

• halpa hinta
• laskennan tarkkuus

Laske hinta droonin ja kameran hankintakustannuksille ja 10 tutkimuskerralle. Voit olettaa, että työvoimakustannus on 100 € / tunti.
Perustele simulaatiomallilla, kuinka tarkka mittauksenne on erilaisilla hirvimäärillä."

---

Features

• Moose Behavior Modeling:

  • Moose move mostly at dawn and dusk.
  • Moose eat and rest for several hours a day.
  • Moose wander around throughout the day, moving alone or as a cow-calf pair.
  • Movement patterns reflect traveling between feeding and watering areas, often in a zig-zag pattern.
  • Moose prefer being near water sources.

• Drone Surveying Setup:

  • Drone equipped with LiDAR and thermal cameras to detect moose and distinguish them from false positives
  • Detects if a cow moose is accompanied by a calf using thermal imagery.
  • The drone follows a planned path over the area and records how many moose are detected within its "view cone."
    • The flight path is based on the bar scan sweep pattern

• Simulation Model Details:

  • Simulates moose locations, hunger cycles, sleeping and resting periods, wandering around, and their movements towards the nearest feeding/watering "bars" (areas).
  • Models drone path and scanning patterns over the forest, and calculates the number of moose detected.
  • Allows for multiple simulation runs to assess average detection accuracy and variability.

Code Structure

• Main.lua (LOVE2D-based visual simulation):
  Runs a single simulation visually. Displays moose movements, drone coverage, and detection counts over time. Used for demonstration and visualization.

• crunch.lua (LuaJIT-based model):
  Runs the core simulation logic without visualization, designed for fast execution and statistical analysis.

  • Parameters: area size, moose density, feeding site density, start time, drone speed, view cone size, and flight duration.
  • Runs the simulation 100 times for a given parameter combination
  • Outputs summary statistics (detected moose vs. actual moose population, smallest & largest amount of detected moose population).

• run.lua:
  Batch runner that executes crunch.lua with various parameter sets, collecting data for aggregate analysis.

How to Run

1. Requirements:

   • LOVE2D for visual simulation (Main.lua).
   • LuaJIT for the CLI simulations (crunch.lua and run.lua).
   • A working Lua environment.

2. Visual Simulation (Main.lua):

   • Install LOVE2D.
   • Run: love . inside the project directory.
   • A window will open showing moose, feeding sites, and the drone's path.

3. Command-Line Simulation (crunch.lua):

   • Run: luajit crunch.lua [parameters]
   • Example parameters: luajit crunch.lua 80000 3.1 12 0 1200 10 10 2
     • These parameters represent area, moose density, feeding site density, start time, drone speed, view width/height, and simulation duration.
   • The script prints out detection statistics. Use different sets of parameters to compare accuracy.

4. Batch Runs (run.lua):

   • Create a parameter file listing various parameter sets line by line.
   • Run: luajit run.lua parameters.txt crunch.lua output.txt
   • Each line in parameters.txt should have a set of parameters, seperated by whitespace.
   • Collects results for multiple parameter sets, useful for tuning parameters and analyzing accuracy over many simulations.

Parameter Units of Measure

• alue (Area):

  • Unit: hectares (ha)
  • 1 hectare (ha) = 10,000 m²
    In this simulation, 1 ha is represented as a 1 hm × 1 hm area (1 hm = 100 m). Thus, 1 ha corresponds to 1 hm² in the simulation’s coordinate system.

• hirviTiheys (Moose Density):

  • Unit: moose per 1000 hectares (moose/1000 ha)

• baariTiheys (Feeding Site Density):

  • Unit: feeding sites (bars) per 1000 hectares (bars/1000 ha)

• vasaTiheys (Calf Density):

  • Unit: calves per moose

• droneSpeed (Drone Speed):

  • Unit: hectometers per hour (hm/h)
    (1 hm = 100 m, so drone speed is given in hm/h)

• Hirvi (Moose) Movement Speeds:

  • Wandering speed: hm/h
  • Searching speed: hm/h

• Drone "View Cone" Dimensions (w, h):

  • Unit: hectometers (hm)

Because the simulation uses a coordinate system where 1 unit of area = 1 ha and 1 unit of length = 1 hm, these units allow straightforward calculations. Distances and drone coverage are in hm, while area-related densities are conveniently expressed in ha.

Interpreting Results

• Drone Counts:
  The simulation prints how many moose were detected and the estimated density based on the surveyed area.

• Accuracy & Cost Analysis:
  Combine simulation results with cost assumptions (e.g., drone purchase, camera cost, operator wages at 100 €/hour, and number of survey runs) to estimate overall cost-efficiency and accuracy.

• Parameter Variations:
  By adjusting moose density, feeding site density, and drone flight parameters, you can see how detection accuracy changes. This helps justify and refine the proposed drone counting strategy.

---

License

This project is licensed under the MIT License.
Please see the LICENSE file for the full terms.

Acknowledgments

• LOVE2D for the 2D game framework.
• LuaJIT for efficient scripting.

This project serves as a conceptual and academic exercise in wildlife surveying simulations.