Sources<\/a><\/li>\n<\/ul>\n<\/nav>\n<\/div>\nDirect answer: how do you choose a power supply for a robot?<\/h2>\n Start with the duty cycle, not the wattage.<\/strong> A robot that works a shift and returns to a dock has a completely different power problem from one that runs near-continuously with opportunity charging. That single question decides your architecture:<\/p>\n\nShift-based, returns to dock:<\/strong> an external adapter sized to recharge the pack inside the off-shift window, plus an onboard charger. The adapter rating follows from pack capacity divided by available charge time, plus charge efficiency losses.<\/li>\nNear-continuous duty:<\/strong> opportunity charging at higher current, battery swap, or contact\/wireless charging built into the floor. The adapter becomes part of the infrastructure, and thermal design at the installation site matters more than the adapter’s own size.<\/li>\nTethered or fixed-base robots (cobots, arms):<\/strong> a mains-powered supply sized for peak, not average, load, with allowance for motor inrush and regenerative energy.<\/li>\n<\/ul>\nThe number most people get wrong is peak, not average.<\/strong> A drive train that averages 150W can pull three to five times that when a motor stalls against an obstruction or reverses under load. If you size the supply to the average, the adapter hits over-current protection exactly when the robot most needs to recover.<\/p>\nMatch the power architecture to the robot class<\/h2>\n \n\n\nRobot class<\/th>\n Typical bus<\/th>\n Power behaviour<\/th>\n What the supply must tolerate<\/th>\n<\/tr>\n<\/thead>\n \n\nAGV \/ AMR (warehouse)<\/td>\n 24V or 48V pack<\/td>\n Long steady draw, short high peaks<\/td>\n 2–3x peak headroom, inrush limiting<\/td>\n<\/tr>\n \nCollaborative arm (cobot)<\/td>\n 48V DC or mains AC<\/td>\n Highly pulsed, regenerative on decel<\/td>\n Reverse-energy handling, tight hold-up<\/td>\n<\/tr>\n \nService \/ cleaning robot<\/td>\n 24V pack<\/td>\n Moderate draw, overnight dock charge<\/td>\n Unattended-charging safety, low standby draw<\/td>\n<\/tr>\n \nDelivery \/ last-mile robot<\/td>\n 24–36V pack<\/td>\n Intermittent, weather-exposed docking<\/td>\n IP-rated connectors, wide ambient range<\/td>\n<\/tr>\n \nHumanoid \/ legged<\/td>\n 48V and above<\/td>\n Extreme transient peaks, high regen<\/td>\n Bulk capacitance, brake energy absorption<\/td>\n<\/tr>\n \nInspection \/ rail robot<\/td>\n 24–48V, often PoE for payload<\/td>\n Low draw, sensitive instrumentation<\/td>\n Low ripple, clean ground, EMC discipline<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\nMobile robots draw steady current for most of a shift and short, sharp peaks when a drive stalls or reverses. The supply has to be sized for the peak.<\/figcaption><\/figure>\nSizing rule: charge window, not wattage<\/h2>\n For any robot that returns to a dock, the adapter rating falls out of one calculation:<\/p>\n
Required output ≈ (pack capacity in Wh × charge efficiency losses) ÷ available charge window in hours.<\/strong><\/p>\nA 480Wh pack that must be recharged in four hours needs roughly 120–135W at the pack, because the charge chain is not lossless. Halve the window to two hours and you need roughly 250W, at which point the conversation shifts to whether the pack can accept that current at all, and whether the dock connector and cable can carry it continuously without heating.<\/p>\n
This is why robotics programs often end up with a larger adapter than the robot’s average consumption suggests, and why it is worth specifying the adapter, the dock connector and the cable as one system rather than as three separate purchases.<\/p>\n
Opportunity charging changes the failure mode<\/h3>\n Short, frequent high-current charges keep robots in service but raise pack temperature and accelerate ageing. If your model depends on opportunity charging, specify the adapter with a genuine continuous rating at the installation ambient, not a peak rating measured on a bench at 25°C. A supply that folds back mid-shift pushes the robot into a low-battery state that is far more expensive than an oversized adapter.<\/p>\n
24V or 48V: why the bus voltage decision comes first<\/h2>\n Mobile robotics has been migrating from 24V to 48V, and the reason is simple arithmetic: for the same power, doubling the voltage halves the current, and resistive losses scale with the square of current. Halving the current cuts I²R losses in the harness and connectors by roughly four times.<\/p>\n
The practical ceiling is the SELV limit. Under IEC 62368-1, a DC bus that stays at or below 60V is treated as a safety extra-low voltage circuit, which dramatically simplifies insulation, creepage and clearance requirements across the whole machine. A 48V nominal lithium pack sits comfortably inside that boundary even at full charge, which is exactly why 48V has become the default for new AMR and cobot platforms.<\/p>\n
Choose the bus before you choose the supply. Every downstream decision — cable gauge, connector rating, protection thresholds, adapter output — follows from it.<\/p>\n
What the adapter actually has to survive<\/h2>\n Robotics environments are harder on power supplies than office or consumer environments in four specific ways. Specify against all four.<\/p>\n
1. Motor inrush and stall current<\/h3>\n A stalled or reversing drive can draw several times nominal current for hundreds of milliseconds. An adapter with fast, non-latching over-current protection will trip; one sized generously but without current limiting will sag the bus and brown out the compute board. The right answer is a supply with a defined overload curve — capable of, say, 150% for a defined period before folding back — and you should ask for that curve in writing.<\/p>\n
2. Regenerative energy<\/h3>\n When a motor decelerates a load, it becomes a generator and pushes energy back onto the DC bus. If nothing absorbs it, bus voltage rises until something fails. In larger machines this is handled by a brake chopper and resistor; in smaller ones it is absorbed by capacitance and the supply itself. Either way, ask your supplier directly: what happens to this adapter if the output is driven above its set voltage?<\/em> A supply with no reverse-current tolerance will be destroyed by a robot that brakes hard downhill.<\/p>\n3. Electrical noise in both directions<\/h3>\n Robots contain both the noisiest and the most sensitive electronics on the same bus: PWM motor drives on one side, lidar, cameras and encoder interfaces on the other. A switching adapter adds its own noise on top. Specify output ripple and noise in millivolts peak-to-peak under your actual load, not at no load, and insist on conducted and radiated EMC testing with the real drive train running. Bench EMC results obtained with a resistive load are close to meaningless.<\/p>\n
4. Vibration, dust and washdown<\/h3>\n Mobile platforms vibrate continuously and, in food, pharma and outdoor applications, get washed or rained on. That drives three requirements: a connector with positive retention and a defined mating cycle life, an ingress rating matched to the cleaning method, and a housing with no unsecured internal parts that can resonate. Washdown installations should be looking at IP65 or better at the enclosure and a mating connector rated to the same standard; a cheap IP20 adapter inside a sealed cabinet is often the more robust and cheaper answer.<\/p>\nRobotics adapters are burned in at load and elevated ambient before shipment, so early-life failures stay in the factory.<\/figcaption><\/figure>\n\n\n\nMethod<\/th>\n Best for<\/th>\n Watch out for<\/th>\n<\/tr>\n<\/thead>\n \n\nManual plug-in<\/td>\n Low fleet counts, supervised sites<\/td>\n Labour cost, connector wear, human error<\/td>\n<\/tr>\n \nDocking contacts<\/td>\n AMR\/AGV fleets, repeated autonomous docking<\/td>\n Contact wear, contamination, alignment tolerance<\/td>\n<\/tr>\n \nInductive \/ wireless<\/td>\n Washdown, cleanroom, sealed enclosures<\/td>\n Lower efficiency, thermal management, higher cost<\/td>\n<\/tr>\n \nBattery swap<\/td>\n Near-continuous duty, no charge window<\/td>\n Pack inventory, mechanical wear, operator safety<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\nWhichever you choose, the external adapter is still in the chain, and its safety behaviour during unattended charging matters. For overnight or unattended docks, look for defined behaviour on output short, reverse polarity and overtemperature, and make sure the adapter recovers safely rather than latching into a fault state nobody is present to clear.<\/p>\n
Safety and standards: who owns which part<\/h2>\n Robot safety is a system property, and it is worth being precise about what the power supply is responsible for and what the machine builder is responsible for:<\/p>\n
\nThe adapter<\/strong> is normally approved to IEC\/EN 62368-1<\/strong> as information-technology equipment, with EMC declarations against the relevant IEC 61000-6<\/strong> generic standards for industrial environments. It is a component, not a safety system.<\/li>\nThe machine<\/strong> carries the functional-safety obligations — ISO 10218<\/strong> for industrial robots, ISO 13849<\/strong> or IEC 62061<\/strong> for control-system safety, and an emergency-stop function that removes motive power, not just logic power.<\/li>\nThe installer<\/strong> owns the fixed wiring, the dock supply and the local earthing arrangement.<\/li>\n<\/ul>\nThat division matters commercially. If a supplier claims an adapter is “robot-safety certified,” ask what exactly is certified. In most cases the honest answer is that the adapter is approved as a component, and the machine integrator carries the rest.<\/p>\n
Emergency stop and the power supply<\/h3>\n A well-designed robot cuts motive power at the drive, not by switching off the adapter input. Killing the adapter removes control power as well, which can leave brakes in an indeterminate state. If your safety concept does require removing input power, tell your supply vendor: it affects hold-up time, restart behaviour and whether the adapter needs to tolerate being switched on and off into a fully charged capacitive load thousands of times.<\/p>\n
For collaborative cells specifically, this stops being a design preference and becomes a documented requirement: the stop category decides whether the supply has to decay quickly after the contactor opens, or to stay inside regulation through a controlled deceleration.<\/p>\n