Why we buy bare bodies and fit our own sensors

There are two easy ways to get into security robotics. One is to design and build your own robot from scratch. The other is to buy a finished, fully configured robot from a manufacturer and resell it with some software on top.

MoLogica does neither. We buy bare, mass-produced robot bodies, fit every one with the same sensor kit and the same brain, and run one software image across all of them. The rule is short: configured, never custom-built.

Where the value in a robot sits

A robot body is mostly chassis, motors and batteries. That is mature, commodity engineering. Manufacturers build these bodies in volume, and they keep getting better and less expensive as the market grows.

A configured unit is a different product. Most of what makes it expensive is the sensors and integration bolted onto that body, and those come as the manufacturer chose them, not as a security patrol needs them. A premium for a configuration you would not have picked is a poor use of a fleet budget.

So the split we chose is:

• Buy the part that is a commodity. The bare body.
• Standardise the part that varies. The sensor kit and the computer.
• Own the part that matters. The software brain that plans, supervises, inspects and learns.

One module set

Every MoLogica robot is designed to carry the same modules:

Module	Role	Standard or upgrade
Optical camera plus near-infrared with IR illuminators	Primary perception, day and night	Standard
Depth sensor, RTK GPS, IMU	Collision backstop and positioning	Standard
2D laser scanner	Indoor mapping	Standard on indoor bodies
Long-range thermal	Detects a person beyond 200 m; the robot drives over and identifies with the cameras	Upgrade
Industrial barcode scanner	Dense or small labels in logistics	Upgrade
Audio	Two-way audio at the robot	Standard

Expensive 3D spinning LiDAR is not fitted to patrol bodies. Cameras and depth sensors do the patrol job, and a 2D laser scanner handles indoor mapping.

The kit is night-first. Most site risk sits in the dark hours, so the primary cameras see in near-infrared with their own illuminators, and thermal is there for detection at range.

One brain

Every body carries an NVIDIA Jetson Orin. The entry board, the Orin Nano with 8 GB, runs perception, mapping, navigation and a small language model on the robot. Larger boards bring bigger on-robot AI and, later, robot arms. The same software stack runs on all of them.

One dock

A MoLogica dock is one geometry for every body, specced to the largest. Docking itself is software: the robot finds a printed visual marker on the charger and uses Nav2’s open-source docking server to line up and connect. Self-docking is proven in simulation, including 8 of 8 docks in the end-to-end proof gate.

One dock design means a site with a mixed fleet does not need a different charger for each body. Large sites rotate bodies through docks rather than waiting out long recharges.

One spares pool

Standardisation pays off most on a bad night. If every robot shares the same modules, a technician carries one set of spares and can service any robot in the fleet. A faulty camera on an indoor robot and a faulty camera on a yard rover are the same part.

It also changes how redundancy works. The service design holds spare bodies at the site, so a fault means a swap, not a gap. A spare is a large share of a small fleet and a small share of a big one: two robots plus one spare is 50% extra, twenty plus two is 10%. Standard modules and one software image mean a spare is ready to work as soon as it is swapped in.

The bodies

Wheels do most of the work, and legs only go where wheels cannot.

Role	Example body	Notes
Research and development	Hiwonder JetAuto	Small, indoor; the first real robot, on the bench
Indoor pilots	AgileX LIMO PRO	Indoor, about 2.5 hours per charge
Industrial wheeled	AgileX Scout Mini, Scout 2.0	Yards, depots, mine perimeters
Stairs and rough ground	Deep Robotics Lite3, X20, X30	Legged; the X30 is rated for harsh outdoor conditions

The Scout Mini class carries a 10 kg payload with about 3 hours per charge. The Scout 2.0 class carries 50 kg with double-wishbone suspension and all-weather cooling. These manufacturers supply bodies that MoLogica configures, and they do not endorse MoLogica.

Riding the hardware curve

The biggest reason to own the brain and buy the body is what happens next year. When a better or less expensive body ships, MoLogica points the same software at it. Checks, missions and history move with the fleet.

There is early evidence that this works. In simulation, the end-to-end proof gate passed on two different robot bodies with no rewrite: a standard research robot and a stand-in shaped like MoLogica’s first real robot, with omnidirectional wheels and different sensor heights. Two bodies is a start, and the roadmap adds a three-robot fleet and then a legged body on the same software.

A company that builds its own robots has to keep redesigning hardware to stay current. A company that resells configured robots is tied to one manufacturer’s roadmap. Buying bare bodies avoids both.

Mixed fleets as an asset

Standardising the kit and the brain has one more effect over time. A fleet of mixed bodies, all producing data in the same format from the same sensors, gives the learning loop more varied experience than a fleet of identical robots. Diverse data from mixed fleets across many sites is designed to train better models than volume from one. Every patrol is recorded in LeRobot format from day one, so that data stays usable as the fleet changes.

Security of the robot itself

A body on an exposed perimeter is a target. Every robot is designed to report position and status continuously, and one that leaves its route or loses its link is itself an alert. Standard bodies with standard modules also make a damaged robot quicker to replace from the spares pool.

Where this stands

Self-docking and the two-body proof gate are proven in simulation. The sensor kit, dock hardware, spares model, industrial bodies and tamper alerting are designed and not yet tested on real hardware. Legged bodies are on the roadmap. The first physical robot, a Hiwonder JetAuto, is on the bench.

Book a demo to see the same software running on two robot bodies in simulation.