In recent years, humanoid robotics have entered the news. From logistic warehouses to industrial factories, to even domestic settings, humanoid robots are slowly integrating into environments designed for people. And that is their primary feature.

Notably, humanoids are starting to take on industrial tasks, such as assembly and material handling. For example, Hyundai’s new U.S. auto plant incorporates advanced robotics and is preparing to integrate humanoids like Boston Dynamics’ Atlas. In domestic settings, Tesla’s Optimus robot has been demonstrated cooking, cleaning and vacuuming.

However, a critical limitation remains an obstacle to humanoid expansion: power. Battery science, overall system weight, thermal management needs and actuator efficiency all contribute to a power-hungry android with limited energy reserves. To overcome this trade-off, increasing runtime and enhancing power deployment potential is key.

Humanoids are power hungry

One of the main drawbacks of current lithium (Li)-ion batteries is their energy density, which tends to fall far short of useful runtimes for humanoid robots, depending on factors such as the cell chemistry and form factor. In the latest models of humanoids, such as Figure 01 and UBTech’s Walker S2, these batteries limit the runtime to just two to four hours per charge, before it must return to a charging station or swap batteries.

Bigger batteries aren’t a simple answer. The limitation to power is partly attributed to the humanoid's physical form. With human-like proportions, the internal capacity available for battery packs is generally smaller, and the center of mass must be balanced on two legs. As compared to mobile robots, with a large chassis, wheels and a low center of mass. The constraints in size and balance makes it challenging to meet the power demands of a humanoid.

The high energy demands stem from operational performance, such as rapid actuation cycles, load bearing and prolonged locomotion. Boston Dynamics’ Atlas, for example, in this video, demonstrates high-torque joints, hydraulically actuated movements that generate significant heat.

While Boston Dynamics has not disclosed the exact power draw of their system, external sources report a runtime of approximately 30 to 60 minutes per charge, suggesting that Atlas requires significant energy during operation. And such a short runtime limits overall utility.

Humanoid robots hinge on battery power

A bigger Li-ion battery might not be the answer humanoid robots needs, but a different chemistry could be.

Advancements in battery technology, driven by larger trends in electrification, have brought an alternative to traditional Li-ion batteries in solid state and silicon-anode Li-ion batteries. These options offer improved thermal stability, a longer life cycle and a faster charging rate, with research showcasing energy levels exceeding 350 Wh/kg.

Silicon-Li batteries, which are more common in mobile robots, have high compliance with international safety and transport standards (UN38.3 and UL2271), which makes them an attractive alternative in humanoid robots as well.

Recent studies have shown promising results into zinc-air and quantum-dot-based batteries, with theoretical energy densities nearing 400 Wh/kg in energy densities, up to 1079 Wh/kg using advanced catalysts. In theory, this would extend runtime by a great margin without increasing the overall humanoid system mass. However, these batteries are still prone to issues such as humidity sensitivity, limited charge-discharge durability and challenges in large-scale manufacturing, so other engineering challenges remain.

A majority of the chemistry alternatives remain in the pre-commercial stages, so lithium-ion batteries will likely remain the first choice for humanoid robots due to their maturity and cost-effectiveness

Beyond the batteries themselves, battery management systems (BMS) play an increasingly critical role in extending runtime and mitigating thermal overhead by actively monitoring and managing battery pack performance.

Advanced BMS software tracks real-time data, predicts charge-discharge patterns and optimizes charging currents across individual cells, helping maximize energy efficiency and the battery’s life cycle. This allows humanoids to draw more usable energy per cycle while also protecting cells from premature degradation. During humanoid locomotion, manipulation or even in idle states, a BMS can dynamically balance loads, prevent overheating from occurring as well as adapt to varying power demands.

Humanoids such as Figure 01 and Sanctuary AI’s Phoenix already integrate BMS algorithms that optimize recharge timing and regulate thermal output during active operation, allowing for more frequent and safer cycling.

Extending operational time without increasing onboard energy storage is also key. Battery-swapping systems could be one answer.

UBTech’s Walker S2 contains an integrated 48 V battery system that autonomously removes and replaces its depleted pack. These battery packs can usually support between 800-1,200 charge cycles, and with fast-charging they can potentially reach 80% capacity in approximately 45 minutes.

Shedding kilograms and increasing efficiency

Optimizing humanoid robot efficiency is another pathway to extending runtimes.

Operating temperatures during high-performance tasks can range up to 50° C, where there is decreased efficiency due to temperature derating, and decreased performance and reliability. The compact nature of humanoids further limits surface area for heat dissipation, making active cooling systems – another power draw – essential in many designs.

These thermal limitations are closely tied to actuator choice. Actuators are the robot’s muscles and can have a significant impact on locomotion quality, energy consumed and mechanical integration. Series elastic actuators (SEAs) are often favored for their impact resilience, force control accuracy as well as their lower-limb application ability, as noted in research on humanoid actuator design.

However, SEAs tend to be bulkier, in comparison to other configurations, which means these actuators exhibit lower control bandwidth in comparison, which directly limits the suitability for applications that require rapid and precise movements.

One alternative being explored is the quasi-direct drive BLDC motor. These use limited gearing, thereby offering a higher torque and a faster response through the joints, which allows for greater dexterity. For applications that require fine motor control in a humanoid's upper limbs and manipulators, BLDC actuators are ideal. However, these actuators consume high amounts of torque, with outputs exceeding 20 Nm and a weight that often exceeds 1.5 kg per joint, which poses integration challenges, especially in compact and weight-sensitive areas.

To address these trade-offs, engineers are exploring hybrid actuators, such as variable stiffness actuators, that balance torque control speed, energy efficiency and thermal management that could be an ideal alternative for wider humanoid robot use.

Reducing the overall weight of a humanoid is another potential way to improve energy efficiency. Most weigh between 40 – 60 kg, but a research bipedal robot from Duke demonstrated a 31% reduction in energy use through the integration of lightweight structural materials and dynamic gait optimization.

Mechanical efficiency can also be enhanced by hybrid actuation systems that combine electric and pneumatic elements. This enhances output-to-weight ratios while also preserving responsiveness. Further improvements could be achieved through optimizing gear ratios, variable stiffness actuators (VSAs) and parallel elastic actuators (PEAs). These changes could ideally reduce inertial loads and improve torque efficiency during operations.

By collectively reducing the energy taxed through thermal overhead, this allows for more available energy dedicated to locomotion and manipulation, ultimately extending the runtime per charge.

Humanoids are ready to walk, not run

Power remains one of the most significant bottlenecks to the real-world deployment of humanoids across potentially all industries. Limitations in battery energy density, compounded by the humanoid form, thermal dissipation demands, and actuator inefficiencies, all restrict current runtimes to between two to four hours in most humanoids.

The autonomy, task complexity, and total deployment cost of the robot are all directly impacted by these limitations. And therefore, use cases today are largely limited to niche, fringe or novel applications – not disruptive humanoid robot workforces.

However, the robotics industry is advancing on multiple fronts and eventually, the technology will eventually develop to a point where robots can work an entire 8 hour shift – coffee or bathroom breaks not required.

About the author

Isaac Sofiyan is a Malaysian content strategist who makes complex robotics and automation topics clear and engaging. He loves turning technical ideas into stories that connect with real people. Read more of his work on Qviro.