Seahorse ROV MVP 1 Underwater Scout

Underwater Fishing Scout ROV — MVP 1

A biomimetic seahorse-shaped underwater ROV that scouts ahead for anglers. Dual fisheye stereo cameras stream immersive VR180 video to your headset or phone — look around freely through the seahorse's eyes. A parallel depth pipeline maps underwater structure and locates fish. Designed for both freshwater and saltwater, boat and shore deployment.

Body & Propulsion Reference →

The Pitch

You're at the bank of a river. Pull the seahorse from your pack, power it on, toss it in. It rights itself and you're underwater — 180° VR vision through the seahorse's eyes on your headset or phone. Look around freely: bottom structure, drop-offs, submerged logs, weed beds, and actual fish. Meanwhile the depth pipeline maps everything in 3D behind the scenes. Steer it upstream, find a deep pool behind a boulder with fish holding, mark the GPS spot, recall the seahorse, and start casting to that exact position. Works from a boat too — drop it over the side and let it scout the area before you anchor up.

SALTWATER FRESHWATER BOAT DEPLOY SHORE DEPLOY 1–10M DEPTH VR180 IMMERSIVE

3D Model Preview

Interactive — drag to rotate, scroll to zoom, right-click to pan.

One Platform, Five Configurations

The segmented body is a modular platform. The L-shaped plates, spine channel, joint system, and servo fin mount are identical across Mini/Scout/Pro. You're just snapping together more or fewer segments. Sprint takes the same seahorse body and adds a rigid hydrodynamic fairing + jet drive. Torpedo is a separate product line — rigid torpedo body, shared electronics DNA.

Upgrade path: A Mini buyer who gets hooked can literally buy more segments, a bigger battery, and additional fin rays to upgrade to a Scout. That's a retention model DJI doesn't have — their products are separate platforms. Yours are the same platform at different scales.

NIR stealth vision across all tiers. Every tier ships with 940nm IR LEDs + NoIR cameras. Fish can't see 940nm light — the seahorse observes in complete stealth. Night fishing, murky water, dawn/dusk — NIR sees through it all. "The seahorse sees what fish can't see."

SEAHORSE MINI

Size: 15–20cm, 12–15 segments

Fins: 4 dorsal rays. Fin-only propulsion. ~0.1 m/s

Battery: 1S 1000mAh. ~30-45 min swim

Range: 20-30m tether. Close-range look tool

Tail: Passive (no prehensile grip)

—

Harvest: None

—

Vision: 1x IMX219 NoIR fisheye · 940nm LED ring (4x LEDs) · Stealth Mode

Use: Wade fisherman. Peek under a cut bank, check a pool. Single fisheye — 360° mono VR look-around on your phone. NIR stealth mode for night/murky water. Fits in a vest pocket. The gateway product.

SEAHORSE SCOUT

Size: 25–35cm, 25–30 segments

Fins: 8 dorsal rays. Fin-only. ~0.2 m/s

Battery: 2S 2200mAh. ~1.5hr swim, 8-12hr patrol

Range: 50m tether. Area scouting

Tail: Prehensile (tendon-driven curl)

—

Harvest: Solar + turbine. All-day patrol capable

—

Vision: 2x IMX219 NoIR fisheye · 940nm LED ring per eye · Visible/NIR switchable

Use: Boat + shore angler. Dual fisheye VR180 stereo — immersive underwater vision + depth mapping. NIR stealth mode sees fish without spooking them. Map structure, anchor and observe, mark waypoints. This is MVP 1.

SEAHORSE PRO

Size: 40–50cm, 35+ segments

Fins: 12 dorsal rays + tail thruster. Hybrid. ~1+ m/s transit, ~0.4 m/s silent

Battery: 3S 4400mAh. ~1hr transit, 12hr+ patrol

Range: 100m tether or autonomous (future). Multi-spot scouting

Tail: Advanced prehensile (2-axis curl, larger grip)

—

Harvest: Solar + turbine + larger panels. All-day+

—

Vision: 2x IMX219 NoIR fisheye · 940nm high-power array (8x LEDs/eye) · 4-5m NIR range · Visible/NIR/Dual mode

Use: Guide, tournament angler. Dual fisheye VR180 + ML fish detection. NIR stealth gives unbiased fish observation day or night. Transit a kilometer, scout multiple spots, long-duration patrol. The serious tool.

High-Speed Tiers — Burst Escape + Rapid Repositioning

Real seahorses are the most effective predators in the ocean — 90%+ strike success rate. They don't chase prey. They hover, blend in, wait, and strike. Speed isn't for cruising — it's for escape and repositioning. Sprint and Torpedo follow the same philosophy: stealth-first, burst when needed.

SEAHORSE SPRINT

FAIRED SEAHORSE — BURST ESCAPE + RAPID REPOSITION

Size: 35–45cm, rigid hydrodynamic fairing over seahorse segments

Burst speed: ~10 m/s (36 km/h) for 5-10s — dodge boats, escape current, reposition 50m. Patrols on silent fins like Pro

Drive: 24mm water jet (burst mode). Silent dorsal fin propulsion (patrol mode). Jet fires only when needed

—

Vision: 2x IMX296 global shutter (120fps during burst) · 940nm flood + spot · Forward sonar (50m obstacle avoidance)

—

Compute: Jetson Orin Nano — ML stereo depth, sonar fusion during burst, standard pipeline during patrol

—

Nav: Tethered during patrol. Autonomous burst escape (INS/DVL + RF buoy relay for reconnect)

Battery: 3S-4S 4400+ mAh. High-discharge cells for burst jet — most energy still goes to patrol

Use: Same stealth patrol as Pro — but with an escape button. Anchored on tail, NIR stealth, silent fins. When a boat approaches, current shifts, or you need to reposition 50m upstream: jet fires, 10 m/s burst for seconds, then back to stealth. Like a real seahorse — 90% hover, 10% dart. Global shutter + sonar only activate during burst to avoid obstacles.

TORPEDO

SEPARATE PRODUCT — BURST TRANSIT + OBSERVE AT DESTINATION

Size: 50–80cm rigid torpedo body — not seahorse-shaped. Supercavitating nose cone

Burst transit: ~72 m/s (260 km/h) to reach target zone fast. Decelerates to conventional speed for observation

Drive: Gas turbine / rocket (burst transit). Conventional thruster (observation mode). Two-phase mission

—

Vision: Sonar-only during burst transit (100m+). Camera + 940nm NIR activate after deceleration for target ID

—

Compute: FPGA + Jetson embedded — sonar nav during transit, vision pipeline during observation

—

Nav: INS/DVL + wire-guide or autonomous. Pre-programmed waypoints. Burst to location, slow down, observe

Battery: 6S+ high-discharge or chemical propellant. Transit budget + observation budget

Use: Research, defense, extreme survey. Burst transit at 72 m/s to reach a distant location fast (inside a supercavitation gas bubble — sonar-only nav). Decelerates to conventional speed at destination, camera + NIR activate for target identification and observation. Not sustained speed — it's "get there fast, then look." Born from the Seahorse platform — same vision stack heritage, same compute architecture.

Shared Platform Components

Identical Across All Tiers

L-plate geometry, spine channel diameter, joint interface, servo fin ray mount slot, eye socket design, head/tail module interface, TPU joint spec, fastener standard (316 SS M2.5).

Varies By Tier

Segment count, battery capacity (1S–6S+), fin ray count (4/8/12), tail complexity, thruster/jet type, NIR LED power (4–20 LEDs), camera sensor (IMX219 NoIR → IMX296 global shutter), compute (Pi Zero 2W → Pi 5 → Jetson → FPGA), navigation (tether → autonomous → INS/DVL).

Software — Fully Shared

Same Pi image, same app, same OpenCV pipeline, same fin wave controller. Software auto-detects config (segment count, servo count) and adapts. One codebase.

Upgrade Path

Mini → Scout → Pro: buy more segments, fin rays, batteries. Same platform. Sprint: adds rigid fairing + jet drive over seahorse body. Torpedo: separate product line, but inherits vision stack, software, and compute architecture from the Seahorse family.

VISION — Fisheye Stereo Eyes (Head Segment)

LEFT EYE — IMX219 NoIR

8MP · CSI-2 · Fisheye 190°+
Dome viewport · No IR filter

            ← baseline = head width →

            190°+ fisheye FOV each

            VR180 stereo · panoramic immersion

RIGHT EYE — IMX219 NoIR

8MP · CSI-2 · Fisheye 190°+
Dome viewport · No IR filter

▼ MIPI CSI-2 (dual)

DUAL PIPELINE — Body Segment

Raspberry Pi 5 — BCM2712

Dual CSI-2 · Dual pipeline fork

VR Pipeline

Fisheye → equirectangular
PiCam360 1/50× compression
VR180 stream → headset/phone

Depth Pipeline

Fisheye dewarp → stereo region
OpenCV disparity → depth map
Structure + fish detection

▼

NAV + SENSORS — Body Segment

IMU (9-DOF)

BNO085 / ICM-20948

Depth / Pressure

MS5837-30BA

LED + NIR Array

Visible LEDs + 940nm IR
Stealth Mode (fish invisible)

Water Temp

DS18B20 (fish behavior)

▼

PROPULSION — Mid Segment

Dorsal Fin Array

Servo-driven undulating fin rays
Silent propulsion · low turbidity

Pectoral Fins (×2)

Yaw + fine positioning
Micro servos, head segment

Buoyancy Engine

Syringe/bladder ballast
Salt ↔ fresh compensation

▼

PREHENSILE TAIL — Tail Segment

Tendon Curl Actuator

Single servo + dual cable
Curls tail to grip structure

Grip Surface

Ridged TPU inner pads
Last 5–6 segments

Contact Sensor

FSR on tail tip
Touch-triggered grip

▼

ENERGY — Harvest + Storage

Battery

2S LiPo · 2200mAh+

PDB + BEC

5V for Pi · 6V servo rail

Solar Cells

Dorsal surface · 0.5–1W
Between fin rays

Micro Turbine

Tail segment · 100–200mW
Current-powered

▼

COMMS + TOPSIDE

Tether (MVP 1)

Thin cable · video + control
30–50m · spine channel routed

VR Headset / Phone VR

PiCam360 viewer · VR180 immersive
Look around freely underwater

Phone / Tablet App

Depth map overlay · structure marking
GPS waypoints · fish heatmap

Vision

Compute

Nav/Sensors

Propulsion

Power

Comms/Output

Vision Scales With Speed — Burst Mode

This diagram shows the Mini/Scout/Pro pipeline (0.1–1 m/s) — the primary operating mode for all tiers. Higher tiers add burst escape capability. Speed isn't sustained — it's for repositioning and escape, then back to stealth patrol:

Patrol (all tiers) — Camera-First, Stealth

IMX219 NoIR at 30fps. 940nm NIR stealth. Dual pipeline: VR180 + depth. Silent fin propulsion. Tail-anchored observation. This is where the seahorse spends 90%+ of its time — like the real animal.

Burst Escape (Sprint) — Camera + Sonar, 5-10s

Jet fires, 10 m/s for seconds. IMX296 global shutter at 120fps activates (rolling shutter = 330mm blur/frame at speed). Sonar for 50m obstacle detection. Dodge a boat, escape current, reposition 50m. Then back to stealth patrol.

Burst Transit (Torpedo) — Sonar-Only, Then Observe

72 m/s supercavitation burst to reach target zone. Inside gas bubble — sonar-only nav (100m+). Decelerates to conventional speed at destination. Camera + NIR activate for observation. "Get there fast, then look."

Biomimetic: The Ambush Predator

Seahorses have 90%+ strike success — highest in the ocean. They don't chase. They hover, wait, strike. Our ROV follows the same pattern: patient stealth observation with burst capability when needed. Speed is defensive, not offensive.

Undulating Dorsal Fin — Primary Thrust

The seahorse's dorsal fin oscillates at up to 35Hz in nature. We replicate this with a series of servo-driven fin rays running along the dorsal (top) ridge of the body. A sine wave propagates down the array — the frequency controls speed, amplitude controls power, and wave direction controls forward/reverse.

This is genuinely silent propulsion. Traditional thrusters create pressure waves that fish detect through their lateral line from meters away. An undulating fin mimics the movement of actual marine life and produces minimal turbulence. This is arguably the product's biggest competitive advantage over conventional ROVs.

Fin Ray Configuration

6-8

Dorsal fin rays — spaced evenly along segments 4–20. Each driven by one micro servo (MG90S). Rays are rigid spars (3D printed or carbon rod) with flexible silicone/TPU membrane connecting them.

Pectoral fins — one each side of the head segment. Handle yaw (turning) and fine positioning. Oscillating or flapping motion. Could also be simple micro thrusters for MVP if fin control is too complex.

PCA9685 servo driver — 16 channel I2C PWM. Generates the phased sine wave across all dorsal servos simultaneously. Pi sends wave parameters, driver handles the PWM timing.

Wave Parameters

Frequency (0.5–5 Hz) — controls swim speed. Low = cruise, high = burst. Higher than 5Hz is beyond servo capability.

Amplitude (10°–45°) — controls thrust power. Small oscillations for hovering, large for pushing through current.

Wavelength (body segments) — typically 1–2 full waves across the fin array. Shorter wavelength = more agile, longer = more efficient.

Direction — wave propagates head→tail = forward, tail→head = reverse. Asymmetric wave = turning assist.

Buoyancy Control

MVP 1: Manual adjustable ballast weights. Velcro-mounted lead/steel weights on the underside. Swap weight amounts when switching salt ↔ fresh. Simple, reliable, no moving parts.

Future: Active buoyancy engine. Small syringe pump in the mid segment — push water into a bladder to sink, push it out to rise. Automated depth hold. Auto-compensates for salt vs fresh density difference (~2.5% buoyancy shift).

Prehensile Tail — Zero-Power Station Keeping

The seahorse's tail is its anchor. It wraps around coral, seaweed, or structure and holds position with zero energy expenditure. We replicate this with a tendon-driven curling mechanism — one servo, two cables, and the segmented tail curls like the real thing.

This solves the biggest energy problem in patrol mode: holding position in current. Instead of burning battery fighting drift, the tail grips structure and the entire propulsion system shuts down. Combined with energy harvesting, the ROV can potentially run energy-positive in patrol mode — harvesting more than it consumes.

Tail Mechanism

Tendon cables — two braided Dyneema lines run through the tail spine channel on opposite sides. Curl cable on inner face, uncurl cable on outer face.

Single servo — mounted at tail base (segment 20–21). Pulls one cable to curl, other to uncurl. Simple, lightweight, one failure point.

Graduated stiffness — joints near tip are looser (fine grip), joints near base are stiffer (holding force). Matches biological seahorse mechanics.

Grip pads — ridged TPU on inner face of last 5–6 segments. Micro-ribbed texture for wet grip on bark, rock, coral, pilings.

Behavior Loop

→

Transit — Fins active (or thruster on Pro). Tail straight. Fast movement to target area.

Approach — Fin-only, silent. Cameras scan for grip-worthy structure. Slow creep toward target.

—

Anchor — Tail tip contacts structure (FSR trigger). Servo curls tail around grip point. Fins stop. Zero propulsion power.

—

Patrol — Anchored observation. Low-power mode. Cameras at reduced frame rate. Harvesting solar + turbine energy. Logging fish activity.

—

Release — Command or auto-trigger. Servo uncurls tail. Fins kick in. Transit to next spot or return. Fail-safe: power loss = slack tendon = auto-release + float to surface.

MVP 1 — Validation BOM Strategy

Phase 1 focuses on validating stereo underwater vision. Waterproofing is minimal — acrylic tube or dry bag enclosure. No custom PCB. Off-the-shelf components on a dev platform. Get it in the water fast, prove the cameras work, iterate from there.

COMPUTE + VISION

Component	Qty	Est.
Raspberry Pi 5 (4GB) Dual CSI-2, Cortex-A76 quad, native stereo camera support	1	$60
RPi Camera Module v2 NoIR (IMX219) 8MP, no IR filter. Side-mounted in eye sockets with dome viewports. NoIR variant enables 940nm stealth vision.	2	$50
Fisheye Lens Adapter (M12 mount, 190°+) Wide-angle fisheye for VR180 panoramic capture. Replaces stock lens on IMX219 board.	2	$15
CSI-2 Adapter Cable (22→15 pin) Pi 5 uses 22-pin, camera modules are 15-pin	2	$6
940nm IR LED Ring 4-8x 940nm LEDs per eye socket, surrounding dome viewport. Invisible to fish (vision cuts off ~800nm). 2-5m stealth illumination range. Based on iFO open-source fish observation project.	2	$5
MicroSD Card (64GB, A2/U3) Fast read for stereo capture + video logging	1	$12

NAVIGATION SENSORS

Component	Qty	Est.
IMU — BNO085 or ICM-20948 9-DOF, orientation + heading. I2C. Critical for attitude underwater (no GPS)	1	$20
Depth Sensor — MS5837-30BA Pressure-based depth. I2C. Gel-filled, designed for ROVs. 0–30m range	1	$25
Water Temp — DS18B20 (waterproof) Stainless probe. Useful data for fish behavior. OneWire bus	1	$4

PROPULSION (MVP 1 — Simplified)

Component	Qty	Est.
Micro Servos — MG90S or equivalent Dorsal fin rays (6–8 servos), pectoral fins (2 servos). Metal gear, waterproofed with silicone or conformal coat	10	$30
Servo Driver — PCA9685 16-channel PWM via I2C. Drives all fin servos from one bus	1	$6
Fin Material — Silicone / TPU Sheet Flexible membrane connecting fin rays. Cut to profile, glue to servo horns	1	$10

PREHENSILE TAIL

Component	Qty	Est.
Tendon Curl Servo — MG90S Single servo at tail base. Pulls dual braided cables through spine channel to curl/uncurl tail. Fail-safe: power off = slack = release.	1	$3
Tendon Cable — Braided Dyneema/Spectra 0.5mm braided line. Two runs (curl + uncurl) through tail spine channel. High strength, near-zero stretch.	2m	$4
Grip Pads — Ridged TPU Textured inner surface on last 5–6 tail segments. Micro-ribbed for wet grip on wood, rock, coral.	1 set	$3
Contact Sensor — FSR (Force Sensing Resistor) Tail tip. Detects contact with structure to trigger grip tightening. Analog input to Pi GPIO.	1	$3

ENERGY HARVESTING (Scout+ / Pro tier)

Component	Qty	Est.
Flexible Solar Cells — 5V 200mA panels Thin-film flex cells on dorsal surface between fin rays. ~0.5–1W total in good sun at 1–2m depth. Wired to charge controller.	2–3	$15
Micro Hydro Turbine — 20–30mm Integrated in tail segment. Water funnels through tapered tail, spins turbine. ~100–200mW in 0.3+ m/s current. DC generator output.	1	$12
MPPT Charge Controller (micro) Manages solar + turbine input, trickle charges LiPo. BQ25570 or similar ultra-low-power harvester IC.	1	$8

POWER + COMMS

Component	Qty	Est.
2S LiPo Battery (7.4V, 2200mAh) Compact, good energy density for servos + Pi	1	$18
BEC — 5V 3A + 6V Servo Rail Dual output: 5V for Pi, 6V for servo rail	1	$8
Tether Cable — Cat5e or USB3 (thin) 30–50m. Ethernet for video stream + control. Routes through spine channel. Neutral buoyancy preferred.	1	$25

ENCLOSURE (MVP 1 — Quick & Dirty)

Component	Qty	Est.
Acrylic Tube — 90mm OD × 300mm Clear tube, end caps with o-ring seals. Houses Pi + battery + sensors. BlueRobotics or similar.	1	$40
Dome Viewports — Acrylic Hemisphere Clear dome inserts for eye sockets. Houses fisheye lens FOV. O-ring sealed to head plate.	2	$15
Cable Penetrators / Potted Bulkheads Waterproof pass-throughs for camera ribbons, servo wires, tether. Epoxy-potted.	1 set	$20
Ballast Weights (lead/steel) Adjustable trim. Different amounts for salt vs fresh. Velcro-mount to tube.	1 set	$10

MINERAL ARMOR — EMA Setup (one-time)

Component	Qty	Est.
DC Power Supply (6V, 5A adjustable) Bench supply or dedicated unit. Constant voltage mode at 6V, current limited to 2 A/m².	1	$50
Titanium Mesh Anode Corrosion-resistant anode. MMO-coated titanium mesh. Reusable across all plate batches.	1	$20
Conductive Spray (graphite/nickel) Applied to PETG plate surfaces before EMA deposition. Creates cathode conductivity layer (~20μm).	1 can	$15
Marine Salt Mix + CaCl₂ + NaHCO₃ Electrolyte components. Marine salt for base ions, calcium chloride for aragonite, sodium bicarbonate for pH buffer.	1 kit	$26
Soaking Tank + Misc (clips, wiring) Plastic tank large enough for plate batches. Alligator clips, wire, thermometer.	1 set	$20
EMA SETUP TOTAL (one-time) Per-plate mineral cost is effectively $0 (just electricity + soak time)		~$131

ESTIMATED TOTAL — MVP 1 (core + NIR)		~$362
With energy harvesting (Scout+ tier)		~$397

SPRINT TIER — Key Upgrades

Component	Qty	Est.
Jetson Orin Nano (8GB) GPU-accelerated ML stereo depth + sonar fusion at 120fps. Replaces Pi 5.	1	$250
IMX296 Global Shutter Camera 1.6MP, 60fps native (120fps binned). No rolling shutter distortion at 10 m/s. Pi-compatible CSI-2.	2	$100
Forward-Looking Sonar Ping360 or similar. Obstacle detection at 20-50m. Camera useless in turbid water at speed.	1	$300
24mm Water Jet Drive Compact jet unit. Reversible. 10 m/s max. Fins lock/retract at speed.	1	$80
INS/DVL Navigation Module Inertial + Doppler velocity log. GPS-denied underwater navigation. Autonomous waypoints.	1	$400
Rigid Hydrodynamic Fairing Carbon fiber/fiberglass shell over seahorse segments. Faired flush dome viewport. Smooth profile for speed.	1	$200
940nm NIR Flood + Spot Array High-power NIR for observation during deceleration phase. Extended range.	1	$15

TORPEDO TIER — Key Components

Component	Qty	Est.
FPGA + Jetson Embedded Compute Hardened for extreme environment. Real-time sonar processing. Same software stack heritage as Seahorse family.	1	$800
Forward-Looking Active Sonar Array 100m+ range. Primary navigation at supercavitation speed. Multi-beam for terrain mapping.	1	$2,000
Supercavitating Nose Cone (Cavitator) Precision-machined cavitator disc. Generates gas envelope at 72+ m/s. Titanium or hardened steel.	1	$1,500
Gas Turbine / Rocket Propulsion High-thrust propulsion for supercavitation regime. Chemical propellant or high-discharge electric.	1	$3,000
INS/DVL + Wire-Guide Navigation Military-grade inertial navigation. Optional fiber-optic wire-guide for real-time control.	1	$2,500
IMX296 Global Shutter (500fps burst) Terminal identification camera. Activates on deceleration. 940nm NIR burst array for target ID.	1	$50
Rigid Torpedo Body (50-80cm) Carbon fiber pressure hull. Not seahorse-shaped. Hydrodynamic torpedo form. Pressure-rated.	1	$2,000

Why This Stack?

The Raspberry Pi 5 gives us dual CSI-2 for synchronized stereo capture, enough compute for real-time depth mapping and VR streaming, and a full Linux stack for rapid iteration. No custom PCB needed for validation.

Side-mounted fisheye cameras mimic real seahorse eye placement. Each eye sits in a socket on the side of the head with a dome viewport housing a 190°+ fisheye lens. The wide FOV captures a VR180 stereo panorama — the user puts on a headset and is underwater, looking around freely through the seahorse's eyes. Meanwhile, the depth pipeline dewarps the central stereo region and computes disparity maps for structure identification. Dual pipeline: immersive VR experience + computed depth data. Based on PiCam360's VR streaming approach with 1/50× bandwidth compression.

The tether-first approach eliminates the hardest underwater problem — wireless comms. RF dies in water, acoustic is slow, optical is line-of-sight. A thin cable gives us full-bandwidth video + bidirectional control with zero latency. The tether runs through the spine channel, protected by the segmented body.

Undulating fin propulsion is silent and low-turbidity. Fish detect traditional thrusters through their lateral line organ from meters away. A wave-driven fin mimics natural fish movement and won't spook your targets.

Software Stack

libcamera + picamera2

Synchronized dual fisheye capture. Hardware ISP debayer. Underwater color correction (red channel boost).

PiCam360 VR Server

Fisheye → equirectangular. Proprietary 1/50× compression (vs raw VR180). Live stream to headset/phone viewer apps. Open source.

Fisheye Dewarp + OpenCV Stereo

Undistort central stereo region from fisheye frames. SGBM disparity → depth map. ~15-30fps at 640×480. Runs parallel to VR pipeline.

Fisheye Stereo Calibration

Underwater calibration with fisheye distortion model. Must account for: (1) refraction through dome viewports + water, (2) fisheye lens distortion (equidistant/equisolid model), (3) convergent side-mounted geometry.

Depth Map → Structure Map

Convert disparity to bottom contour. Identify structure types: rock, wood, vegetation, drop-offs.

Fin Wave Generator

Sine wave propagation across servo array. Frequency = speed, amplitude = power, direction = steering.

Tether Comms Protocol

VR stream + depth data upstream, joystick commands downstream. PiCam360 compression keeps bandwidth under 1.5 MBps.

NIR Stealth Mode (940nm)

Single toggle in app: "Stealth Mode". Switches from visible LEDs to 940nm IR array. Same camera pipeline (NoIR sensor passes 940nm). VR view shows grayscale NIR — like underwater night vision goggles. Fish can't see 940nm (vision cuts off ~800nm). Unbiased behavior observation.

Burst Mode Vision Pipeline (Sprint / Torpedo)

Sprint — Burst Escape Pipeline (5-10s)

During patrol: standard IMX219 NoIR + Pi 5 pipeline. When jet fires for burst escape: switches to IMX296 global shutter at 120fps on Jetson Orin Nano + forward sonar (50m obstacle avoidance). Burst lasts seconds — dodge, reposition, then back to stealth patrol pipeline. Two modes, one vehicle.

Torpedo — Burst Transit → Observe

Two-phase mission: (1) Burst transit at 72 m/s — sonar-only nav (100m+) + INS/DVL, camera off, inside supercavitation bubble. (2) Deceleration at target zone — camera + 940nm NIR activate for observation and identification. FPGA handles sonar during transit, Jetson handles vision during observation. Not sustained speed — it's rapid deployment.

Underwater Vision Challenges

Refraction

Light bends at the dome viewport/water interface. Dome shape reduces refraction distortion vs flat ports. Stereo calibration must be done in-water with fisheye distortion model.

Color Absorption

Red drops off fast underwater. Need white-balance correction. Visible LEDs help at depth. In NIR stealth mode, 940nm illumination provides consistent monochrome image regardless of depth color shift.

Particulate Scatter

Sediment and plankton create noise in stereo matching. May need confidence filtering on depth map.

Dual Water Design Considerations

Freshwater: Lower density (~1.0 g/cm³). ROV will be less buoyant — may need less ballast weight. No corrosion concern on most materials. Better visibility in lakes, worse in rivers with sediment.

Saltwater: Higher density (~1.025 g/cm³). More buoyant — add ballast to compensate. Corrosion on anything non-marine-grade. Use 316 SS or titanium fasteners, conformal coat all electronics, rinse after every use. Better visibility nearshore in calm conditions.

Ballast strategy: Adjustable ballast weights with Velcro mount for MVP 1. Swap weights when switching between salt and fresh. Future: active buoyancy engine (syringe pump) auto-compensates.

Biomimetic Armor — Seahorse Plate Structure

Inspired by McKittrick & Meyers' research at UC San Diego. The seahorse tail uses 36 square segments, each made of 4 L-shaped interlocking plates. The structure compresses to ~50% before permanent damage, protecting the spine. We apply the same principles to create a modular, impact-resistant, field-replaceable enclosure for the ROV electronics.

Ref: UCSD — Seahorse Armor Research · Acta Biomaterialia, 2013

Bio → ROV Mapping

36 Segments → Modular Shell

Tapering segments from head (cameras) to tail. Each independently replaceable. Corroded saltwater plate? Pop and swap.

4 L-Plates → Quad-Panel Segments

Each segment = 4 interlocking panels around a central spine channel. Snap together, no fasteners on outer shell.

Seahorse Eyes → Fisheye VR Cameras + NIR

Cameras sit in eye sockets on the sides of the head — exactly like a real seahorse. Dome viewports house 190°+ fisheye lenses (IMX219 NoIR — no IR filter). 940nm IR LED ring surrounds each dome for stealth illumination — fish can't see 940nm. Dual pipeline: immersive VR stream + computed depth maps. "Stealth Mode" toggle in app switches visible → NIR (grayscale night-vision view in headset).

Gliding Joints → Impact Flex

Plates slide on impact — survives being tossed into water from shore, bumping into rocks, fish strikes. O-ring seals maintain waterproofing under compression.

Vertebral Column → Wire Spine

Central sealed tube through all segments. Carries flex cables, power, and the tether. Waterproof core.

Hard Ridges / Soft Grooves → Dual Material

ASA/PETG ridges for impact and UV resistance. TPU grooves and joints for flex and waterproof sealing.

Dorsal Fin Ridge → Propulsion Spine

Fin ray servos mount along the dorsal ridge. Fin membrane stretches between rays, forming a continuous undulating surface.

Electronics Placement by Zone

Head (Seg 1–3) — Side-mounted fisheye stereo cameras in eye sockets, like real seahorse eyes. Each eye has a dome viewport housing a 190°+ fisheye lens (IMX219 NoIR). 940nm IR LED ring surrounds each dome — stealth illumination invisible to fish. Pectoral fin servos. Widest segment — baseline = head width. VR180 stereo + NIR stealth mode.

Body (Seg 4–10) — Main PCB: Pi 5 (future: CM5 carrier), IMU, depth sensor, servo driver. This is the sealed rigid core — minimum flex, maximum protection. Tether breakout here.

Mid (Seg 11–20) — Battery, BEC, ballast weights / buoyancy engine. Weight centered for neutral trim. Dorsal fin ray servos mount along this zone. Most of the propulsion happens here.

Tail (Seg 21–30) — Prehensile gripping tail. Tendon curl servo at base, braided Dyneema cables run through spine. Ridged TPU grip pads on inner surface of last 5–6 segments. FSR contact sensor at tip. Micro hydro turbine in final segments (water funnels through tapered tail). Tether exit point. Fail-safe: power loss = slack tendon = auto-release + positive buoyancy float to surface.

Material Selection — Dual Water Rated

Rigid Plates — ASA

UV resistant, saltwater tolerant, good impact strength. Better than PLA/PETG for marine use. Print at 0.2mm.

Flex Joints — TPU 95A

Living hinges between segments. Waterproof gasket properties. Excellent fatigue resistance for repeated flex.

Fasteners — 316 Stainless Steel

Marine grade. All internal fasteners, o-ring screws for sealed compartments. Titanium for premium build.

Fin Membrane — Silicone Sheet

Flexible, waterproof, UV resistant. Bonds to ASA fin rays with silicone adhesive. 0.5–1mm thickness.

Electrochemical Mineral Armor — Biorock-Inspired

Low-voltage DC current through conductive-coated PETG plates in mineral electrolyte deposits aragonite/brucite ceramic armor — the same mineral that builds coral reefs. This transforms the body from bare plastic to a bio-composite: tough PETG core + hard mineral shell, mimicking actual seahorse bony armor.

EMA Process

Conductive-coated PETG plates act as cathode in mineral electrolyte. 6V DC at 2 A/m² deposits aragonite (CaCO₃) / brucite (Mg(OH)₂) ceramic crust. 48–168hr soak = 0.5–2.5mm mineral shell.

Impact Armor

Mineral crust (Mohs 3.5–4) absorbs and distributes impact force. PETG core holds structural integrity. Compression-resistant composite — bumping rocks, structure, or shore launch.

Integrated Ballast

Mineral density 2.9 g/cm³. Thicker crust for saltwater, thinner for freshwater. Same plates, different soak time — replaces external ballast weights.

Sacrificial + Self-Healing

Cracked or chipped mineral re-grows by re-soaking damaged plates. No stripping, no repainting. Field repair with a portable tank.

Acoustic Dampening

Dense mineral surface reduces plastic resonance — quieter near fish. Complements the silent fin propulsion stealth approach.

Natural Camouflage

Mineral texture + optional biofilm top coat = natural appearance underwater. The seahorse looks like a piece of reef structure, not a plastic gadget.

Form Factor Advantage — Fishing Stealth

The seahorse shape isn't cosmetic — it's functional camouflage. Side-mounted eyes are the key detail. Fish have evolved to recognize eye placement as friend-or-foe. Forward-facing eyes (cats, sharks, eagles) signal predator. Side-facing eyes (most fish, seahorses, herbivores) signal non-threat. By placing cameras on the sides of the head like a real seahorse, the ROV registers as fauna, not hunter. It gets closer to fish without triggering flight response — exactly what you need for a scouting tool.

The compact profile also means no protruding motor arms to snag on weeds, branches, or fishing line — common hazards in the exact structures where fish hold. The dome viewports for the fisheye lenses sit flush in the curved eye sockets — smooth hydrodynamic profile with minimal drag.

SVG Reference Diagrams — Body & Propulsion →

The Endurance Equation

The combination of prehensile tail anchoring, solar harvesting, and water turbine generation fundamentally changes what this device is. It's not an ROV you use for 20 minutes. It's an autonomous underwater observation platform that can run for hours — potentially all day in the right conditions.

The key insight: anchored patrol mode draws dramatically less power than swimming. Propulsion is 60-70% of total energy budget. Eliminate it and you extend runtime 3-4x from battery alone. Add harvesting on top and the math gets very interesting.

Power Budget — By Mode

—

Transit (thruster — Pro only): ~15-20W total. Pi at full load (7W), thruster (8-12W), sensors (1W). Highest draw. Battery-only. ~45min runtime on 2200mAh.

—

Swimming (fin array): ~10-12W total. Pi at full load (7W), 8 servos oscillating (3-5W), sensors (1W). ~1-1.5hr runtime on 2200mAh.

—

Anchored patrol: ~2-3W total. Pi in low-power duty cycle (wake every 5s for frame capture: avg 1.5W), sensors (0.5W), tail servo holding (0.2W). ~5-7hr runtime on battery alone.

—

Anchored patrol + harvesting: ~2-3W draw minus 0.5-1.2W harvest = net 1-2W. Runtime extends to 8-12+ hours. In ideal conditions (sunny, shallow, good current), near energy-neutral.

Harvesting Sources

—

Solar — 0.5-1W: Thin-film flex cells on dorsal surface between fin rays. At 1-2m depth in clear water, 40-60% of surface irradiance reaches the panels. Useless at night, murky water, or below 3m. But in the common case (shallow, daytime, clear lake) it's meaningful.

—

Water turbine — 100-200mW: Micro turbine in tail segments. Water funnels through tapered tail, spins generator. Works in any current >0.2 m/s. Best in rivers. When anchored in current, tail simultaneously grips structure AND funnels water through the turbine. Dual function.

—

MPPT controller — BQ25570: Ultra-low-power harvesting IC. Manages both solar and turbine inputs. Trickle charges LiPo during patrol. Prevents overcharge. Auto-switches between sources based on availability.

All-Day Patrol Scenario

6:00 AM — Deploy seahorse at a channel swing on a river. Thruster transit to submerged log jam 40m upstream. Tail anchors to a branch on the current-facing side.

6:05 AM–12:00 PM — Anchored patrol mode. Cameras at low frame rate duty cycle. Solar charging from morning sun. Turbine harvesting from river current. Logging water temp, depth, and video frames. When fish activity detected, switches to full-rate video and flags timestamp.

12:00 PM — Retrieve. You have 6 hours of data: fish activity heatmap (most active 7:15–8:30 AM), water temp curve (peaked at 18.2°C at 10 AM), and video clips of every fish that passed the structure. You now know exactly when and where to fish tomorrow.

No fish finder on the market provides this dataset.

Phase 1 — Fisheye Stereo Vision + VR Streaming + NIR

WE ARE HERE

Bench test — Pi 5 + two IMX219 NoIR with fisheye lens adapters + 940nm LED rings. Fisheye stereo calibration in air. Verify synchronized capture. Test PiCam360 VR streaming to phone. Validate NIR stealth mode toggle (visible → 940nm).

Dual pipeline validation — Run VR stream (PiCam360) and depth pipeline (fisheye dewarp + OpenCV SGBM) simultaneously. Verify Pi 5 can handle both at acceptable framerates.

Underwater calibration — Seal cameras with dome viewports in dry bag enclosure. Fisheye stereo calibration in water (refraction through dome). Validate depth accuracy at 0.5–5m. Test VR180 stream quality underwater.

Pool test — Static placement in a pool. VR headset view — can you look around and see structure? Depth map overlay on phone — does it identify bottom features? Test color correction and LED illumination.

Open water test — Lake or calm shore. Tethered. VR experience in real conditions — particulate, current, variable light. Validate both pipelines work under field conditions.

Phase 2 — Propulsion + Tail Validation

Fin array bench test — 6-8 servos on a rail. Validate sine wave propagation, measure thrust in a water tank. Tune frequency/amplitude/wavelength.

Prehensile tail test — Tendon-driven curl mechanism on a test rig. Validate grip force on wood, rock, and rope. Test fail-safe release (power off = uncurl). Test FSR contact detection.

Integrate vision + propulsion + tail — Cameras + fin array + tail in acrylic tube. First self-propelled underwater run. Anchor to submerged structure. Tethered control from phone/laptop.

Buoyancy tuning — Test ballast in fresh and salt water. Establish weight profiles for each. Validate neutral trim and positive-buoyancy failsafe (float on power loss).

Phase 3 — Seahorse Body + Energy Harvesting

3D print segmented body — ASA plates + TPU joints. Test fit electronics from Phase 2. Validate flex/impact properties. Drop test and compression test.

Waterproof integration — Seal spine channel, eye socket viewports, tether penetrator. Test at 1m, 5m, 10m depth hold.

Dome viewport sealing — Acrylic dome inserts in eye sockets. O-ring seal testing. Validate fisheye FOV isn't clipped by socket geometry. Pressure test dome-to-plate seal at 10m.

Solar + turbine integration — Flex solar cells between dorsal fin rays. Micro turbine in tail segment. MPPT charge controller. Measure actual harvest rates at various depths, light conditions, and current speeds.

Mineral armor prototyping (EMA) — Electrochemical mineral accretion on PETG test coupons. Validate adhesion, impact resistance, o-ring seal integrity after deposition, and buoyancy delta between freshwater/saltwater crust thicknesses. Test self-healing by cracking and re-soaking.

Patrol mode endurance test — Anchor at structure, run low-power observation mode. Target: 8+ hours continuous operation with harvesting. Validate energy-positive conditions.

Phase 4 — Product

Custom PCB — CM5 carrier board shaped to body segment profile. Integrated servo driver, power management, MPPT harvester, tether interface. Replace all dev boards.

Phone app — Live stereo video with depth overlay. Structure marking + GPS waypoints. Joystick control. Fish activity heatmap over time. Water temp/depth trends.

Field testing — all environments — Freshwater lake, river, saltwater nearshore. Boat + shore deploy. Validate tail grip on natural structure. Validate energy harvest in real conditions.

Product line — Mini / Scout / Pro — Finalize segment counts, BOM, and pricing for each tier. Upgrade kits (extra segments, fin rays, harvesting modules). NIR stealth vision standard across all tiers. One codebase, three configs.

Future — Autonomy

∞

Autonomous scouting — release, auto-survey a grid, anchor at promising structure, observe, move on. Visual SLAM for GPS-denied underwater nav. Fish detection + behavior ML. Auto-recall on low battery. Cut the tether. True underwater trail camera for fish.

Phase 5 — Seahorse Sprint (Burst Escape)

Rigid fairing design — Hydrodynamic shell over seahorse segments. Faired flush dome viewport. CFD optimization for burst at 10 m/s. Still recognizably seahorse silhouette. Same body platform as Pro.

Dual-mode propulsion — 24mm water jet for burst escape (5-10s at 10 m/s). Silent dorsal fin for patrol. Fin lock/retract during jet burst. The seahorse is an ambush predator — 90% hover, 10% dart.

Burst vision pipeline — IMX296 global shutter + sonar activate only during burst escape. Standard IMX219 NoIR pipeline during patrol. Jetson Orin Nano handles both modes. Validate 1/1000s exposure eliminates motion blur at 10 m/s burst.

Burst escape nav — INS/DVL for autonomous burst repositioning. Tethered during patrol. RF surface buoy for reconnect after burst escape. Vehicle returns to stealth patrol after repositioning.

Phase 6 — Torpedo (Burst Transit) — Research

Supercavitation R&D — Cavitator disc design, gas bubble dynamics. Rigid torpedo body (50-80cm). Burst transit to target zone at 72 m/s (260 km/h). Not sustained — "get there fast, then observe."

Two-phase mission pipeline — Phase 1: sonar-only burst transit (100m+ range, INS/DVL). Phase 2: decelerate at target, switch to camera + 940nm NIR observation mode. FPGA for transit sonar, Jetson for observation vision.

Rapid deployment validation — Validate burst transit + deceleration + observation cycle. Gas bubble stability during transit. Smooth transition from sonar-nav to camera-observe. Shared vision stack heritage from Seahorse platform.