A standalone browser tool that uses Multi-Objective Particle Swarm Optimization (MOPSO) and Fibonacci sphere seeding to find optimal LiDAR sensor arrangements for drone platforms. Designed specifically for use with PUP — Parameter Uplink Spectagraph, a PUP simulator inspired by Prometheus (2012). Results can be exported directly as parameter tables for input into parametric CAD assemblies.

Given a sensor count, drone body radius, and LiDAR hardware specifications, the optimizer searches for sensor orientations (pitch + yaw) that simultaneously maximize volumetric coverage and minimize inter-sensor overlap. The search space is the surface of a sphere, constrained to a user-selected orientation zone. A live Plotly 3D visualization updates in real time as the swarm evolves and resizes correctly with the browser window.

• MOPSO — Multi-Objective Particle Swarm Optimization with a capped Pareto archive (max 100 solutions). Each PSO step runs 3 sub-iterations per animation frame via requestAnimationFrame, keeping the UI responsive throughout
• Fibonacci sphere initialization — The swarm is seeded using the golden-angle Fibonacci spiral, giving near-uniform initial sensor spacing across the sphere before perturbation
• Two-objective fitness evaluation — Each particle is evaluated against two objectives: blind spot ratio (fraction of 2,000 reference sphere points not covered by any sensor) and maximum pairwise sensor overlap (cosine similarity of facing normals). The Pareto front tracks all non-dominated solutions
• Body occlusion test — For every reference point, a ray-sphere intersection check (checkOcclusion) determines whether the drone body blocks the line of sight from the sensor to the evaluation point, excluding self-occluded coverage from the fitness score
• Cosine-normalized yaw velocity — Yaw PSO updates are divided by cos(pitch) to prevent particles from spinning excessively near the poles, stabilizing convergence at high and low latitudes
• Patience-based auto-stop — Optimization halts automatically after 120 consecutive iterations with improvement below 1e-6, preventing unnecessary compute after convergence
• Angular resolution auto-calculator — H. and V. angular resolutions are computed live from scan frequency, ranging frequency, beam divergence (converted from mrad), channels, and FOV — accounting for beam divergence as a floor on the achievable resolution
• Responsive 3D plot — The Plotly visualization fills its container at any window size and calls Plotly.Plots.resize() on window resize to stay correctly fitted
• CAD export matrix — Produces a formatted plain-text report of all optimized sensor positions (pitch °, yaw °) alongside full hardware specs, PSO solution quality, and Pareto front size. Suitable for pasting directly into parametric assembly environments
• Controls lockout during optimization — All inputs are disabled while the swarm runs; only the Stop button remains interactive

PUP-optimizer-v1.0.1.html   # Main entry point and UI layout
script.js                   # MOPSO engine, fitness evaluation, Plotly visualization, export
styles.css                  # Dark-mode UI styling

• Plotly.js 2.27.0 — 3D scatter visualization

No build step required. Open PUP-optimizer-v1.0.1.html directly in any modern browser. Internet connection required on first load.

H. and V. angular resolutions are computed automatically and displayed as read-only fields.

Click Optimize to start. The 3D plot updates every 15 iterations showing current sensor positions, scanning arcs, the reference coverage sphere, and the drone body. The status log reports iteration count, coverage %, max overlap %, and Pareto front size.

Click Stop at any time to halt and unlock controls. The optimizer also stops automatically on convergence.

Click Export Optimized Positions to open the CAD Export Matrix modal, containing a full formatted report of sensor positions and hardware specs.

Fitness objectives (both minimized):

• obj[0] — blind spot ratio: 1 - coveredPoints / totalRefPoints
• obj[1] — max overlap: highest absolute dot product between any two sensor normal vectors

Best archive member is selected by weighted scalarization: 0.8 × obj[0] + 0.2 × obj[1], prioritizing coverage over overlap reduction.

Velocity clamping — pitch and yaw velocities are clamped to ±0.35 rad/step. Yaw velocity is further normalized by cos(pitch) to maintain uniform angular step size across latitudes.

Pareto dominance — solution A dominates B if A is better or equal on both objectives and strictly better on at least one. The archive is capped at 100 members; oldest excess members are pruned when the cap is reached.

Any modern browser supporting ES6 and WebGL (required by Plotly 3D). Tested in Chrome, Firefox, and Edge.