Hai Wiki
home / knowledge / 07-energy.md

Layer 7 — Power and energy management

Supporting depth. Battery design is not a solutions-architecture problem, but why one deployment needs 18 charging stations for 53 robots is — that is a capacity question dressed as an electrical one.


Battery chemistry selection

What it is. Choosing a cell chemistry to match the robot's duty cycle.

Why it matters here. Chemistry determines charge time, cycle life and safety — and therefore how many robots are in service at any moment, which is a throughput input.

Fundamentals.

Chemistry Charge speed Cycle life Energy density Typical use
LiFePO4 (lithium iron phosphate) Moderate ~2,000–5,000 Moderate Steady-duty robots; thermally very stable
LTO (lithium titanate) Very fast ~10,000–20,000 Low High-cycle, opportunity-charged robots
NMC / Li-ion Moderate ~1,000–2,000 High Where range matters more than cycles
  • The design logic is duty cycle. A robot that works two shifts and charges in a gap wants cycle life and fast charge, not energy density — it is never far from a charger. That is why warehouse robotics favours LiFePO4 and LTO despite their lower energy density.
  • LTO's low energy density is acceptable precisely because the robot is indoors, always near a charger, and charges in minutes. Ten thousand cycles at 15 minutes each is a machine that is essentially always available.
  • Thermal safety matters in a building full of flammable inventory: LiFePO4 and LTO are both far more thermally stable than NMC. In a warehouse fire-risk assessment this is a real argument.
  • Cycle life sets the replacement interval, which is a total-cost-of-ownership line the customer will ask about.

How Hai applies it.

"The HaiPick system uses lithium-iron phosphate batteries in all its models." — /robots/haipick-a3

"Fast-Transit Companion AMRs use LTO batteries with up to 10,000 charge cycles." — /solutions/haipick-system-3

"LTO batteries offer high safety, fast charging, and wide temperature tolerance." — /solutions/haipick-system-3

"10-year battery lifespan in typical working temperatures." — /robots/fast-transit-amr-k50

The chemistry split maps exactly onto the role split. ACRs (LiFePO4) do slow, heavy vertical work with long dwell times. Fast-transit AMRs (LTO) run constantly at 4 m/s and need charge-and-go. That is a coherent engineering decision rather than an accident of sourcing.

Tradeoffs and pitfalls. Wide temperature tolerance matters for cold-chain and grocery sites — both are in Hai's industry list. Battery performance in a chilled warehouse is a legitimate design question.

Key terms. LiFePO4 · LTO · NMC · cycle life · energy density · thermal stability · depth of discharge · TCO.


Autonomous charging

What it is. Robots taking themselves out of service to charge, unattended.

Why it matters here. Charging-station count is sized in every single case study in the corpus. It is a first-class layout parameter competing for floor space with storage.

Fundamentals.

  • Recap the duty-cycle arithmetic from algorithms: steady-state chargers ≈ N × C / (R + C).
  • Real provisioning always exceeds the steady-state figure because demand is not smooth — robots deplete in clusters after a peak, and you must not enter the next peak with a queue at the chargers.
  • Fast charge changes the topology of the problem. At 10 minutes per charge, a charger serves many robots per hour, and charging can hide entirely inside natural idle gaps. At 2 hours it cannot, and you need a bigger fleet to cover the absent robots.
  • Charging bays consume floor area that could hold inventory — an explicit density tradeoff to raise in a layout review.

Observed ratios in Hai's own deployments (from data/catalog.json):

Site Robots Chargers Ratio
Anta Jinjiang 53 18 2.9 : 1
Best Supply Chain footwear 40 14 2.9 : 1
Bosideng HPS1 87 24 3.6 : 1
Chicheng Group 33 11 3.0 : 1
Automotive electronics 7 4 1.8 : 1

Roughly one charger per 3 robots, tightening for small fleets where there is less statistical smoothing. This is a genuinely useful first-pass planning heuristic.

How Hai applies it.

"it automatically moves to a charging station to get fully charged" — /robots/haipick-a3

"Built-in protection circuits ensure safe charging throughout the process. After charging, the robot stops charging automatically." — /robots/haipick-a3

"Fast Charge ability: 10 minutes of charging powers the AMR for 2 hours." — /robots/fast-transit-amr-k50

"Each charge takes ~15 minutes and provides roughly 3 hours of operation." — /solutions/haipick-system-3

Hai also sells the charger as a named product, HaiCharger.

A 15-minute charge for 3 hours of operation is a ~7.7% duty overhead — meaning steady-state maths alone would suggest ~4 chargers for 53 robots, not 18. The gap between 4 and 18 is entirely peak coincidence and queue avoidance, and being able to explain that gap is a very good answer.

Tradeoffs and pitfalls.

  • Under-provisioning creates a charger queue, which removes robots from service for far longer than the charge itself.
  • Charging during peak steals exactly the capacity you need; charge scheduling should anticipate the peak, not react to it.

Key terms. opportunity charging · duty overhead · charger provisioning ratio · peak coincidence · state of charge · HaiCharger.