RAIN

Articles

Building autonomous robots where power and parts are scarce

Unreliable power and hard-to-find components change how you design robots, not whether you can. Practical principles for energy budgets, local sourcing, lean onboard intelligence and field testing.

The RAIN Team · 19 November 2026 · 5 min read

Many robotics tutorials assume a reliable mains supply, next-day delivery and a budget for specialist components. Much of the world, including most of Nigeria, has none of those. That does not make autonomous robots impossible. It changes how you design them. In fact, scarce power and parts tend to produce machines that are simpler, tougher and easier to repair, which are qualities every robot needs. This article sets out practical principles for building under those conditions.

Design for the power you actually have

Power shapes every other decision, so start with an energy budget before choosing a single component. List every load on the robot: drive motors, steering, sensors, computers, radios, lights. Estimate the current each draws and for how long, and add them up for a working day. The total tells you how much battery you need, and usually shows that one or two items consume most of the energy.

Drive motors are almost always the largest load, so mechanical efficiency is worth more than any software trick. Keep the robot light, choose wheels and tyres that roll easily on the actual ground, gear the motors so they run near their efficient speed, and make sure bearings and shafts are aligned. A robot that wastes energy fighting its own friction will never have enough battery.

Electronics come next. Microcontrollers can sleep most of the time and wake on a timer or a sensor event, drawing very little in between. Computers that run vision should be switched on only when they are needed, not left idling. Radios, especially cellular modems, draw power in bursts, so sending data in batches rather than continuously saves a surprising amount.

Choose battery chemistry with care. Lithium iron phosphate cells are heavier than some alternatives but are widely regarded as safer and longer-lasting, which matters when a pack is charged daily in hot conditions. Reused cells from old laptops or power banks can work for small projects, but only after each cell has been tested for capacity and internal resistance, matched with similar cells and protected by a proper battery management system. Badly assembled packs of salvaged cells are a fire risk.

Where possible, plan for charging from more than one source. Solar charging makes sense in much of Africa, and a robot that works by day and charges at a solar station, or from a generator when the grid is off, is far more useful than one that sits waiting for the grid to come back.

Work with the parts you can get

Imported components are expensive, slow to arrive and sometimes counterfeit. Designing around what you can buy locally, or salvage, makes projects faster and makes them repairable.

Nigeria has deep fabrication skills that robotics builders often overlook. Welders, machinists, panel beaters and auto electricians can produce chassis, brackets, shafts and wiring harnesses at a fraction of the cost of imported equivalents. Markets such as Ladipo in Lagos are full of motors, gearboxes, bearings and wheels from vehicles and machinery, and electronics markets such as Computer Village in Ikeja stock common modules. End-of-life parts can be excellent: car wiper motors, wheelchair drives and electric scooter hubs are robust, geared and designed for daily use.

RAIN has shown how far this approach can go. Its industrial mobile robot prototype, an unmanned ground vehicle, carries a 250 kg payload with 15 hours of endurance and was built from locally fabricated and end-of-life parts. The point is not that every team should build the same machine, but that local sourcing need not mean low performance.

A few habits make salvaged and local parts easier to work with:

Keep the intelligence lean

Autonomy does not require a powerful computer in the cloud. Where connectivity is patchy and data costs money, it is usually better to do the thinking on the robot itself.

A sensible split puts time-critical control on a microcontroller, which is cheap, reliable and frugal with power, and puts heavier tasks such as vision on a small edge computer. Machine learning models can be shrunk for this purpose by quantising them and choosing compact architectures; the field known as TinyML specialises in running models on microcontrollers that draw only milliwatts. A smaller model that runs reliably on the robot is worth more than a large one that depends on a connection.

Communication should assume the network will fail. Store data on the robot and send it when a signal is available. For alerts, the mobile network is often the best channel, and even where mobile data is weak, a text message may get through. One of RAIN's drone prototypes uses this pattern, analysing footage on an edge computer in real time and sending alerts over the mobile network rather than streaming video.

Above all, make sure the robot fails safe. If a sensor dies or the link drops, the robot should stop or return to a known state, not carry on blind.

Build, test and maintain for real conditions

Test where the robot will work, not just in the workshop. Heat, dust, rain, mud and uneven ground find weaknesses no bench test reveals. Electronics enclosures need ventilation without letting dust in. Connectors need strain relief and protection from water. Cables should be routed away from wheels and hot motors.

Plan for maintenance by the people who will actually own the robot. A farmer, a factory technician or a security team should be able to swap a battery, replace a wheel and read a simple fault indicator without calling an engineer. Clear labels, a printed wiring diagram inside the enclosure and a short maintenance sheet do more for a robot's working life than another sensor.

Think about spares from the first design review. If a part fails often, make it cheap, easy to reach and quick to change. Keep a small stock of the parts most likely to wear, such as wheels, fuses, connectors and motor brushes, close to where the robot works.

Finally, keep a log. Record every failure, its cause and the fix. Over months, that log becomes the most valuable design document you own, because it tells you what to change in the next version.

Constraints are not a reason to wait, so start your next robot by writing its energy budget and a list of parts you can source within a day's travel, then design around both.

More articles

Learn with RAIN from anywhere

Live classes in your own time zone, online labs and Professional Certificates.

Study from abroad