Live Bipedal Trials: Mapping 2025–26 Real-World Humanoid Deployments in Warehouses, Factories and Care Homes

Editorial note: I do not have live web access to verify post‑June 2024 primary sources. Below is an evidence‑based investigative synthesis and working report template built from publicly known vendor positioning up to mid‑2024, patterns in commercial robotics adoption, demo footage and vendor disclosures, and interviews and filings available in that window. Where I reference 2025–26 pilots or customer names, those items are presented as "vendor‑reported" or "publicly announced" and should be validated against the primary press releases or customer statements before being used as a definitive historical record. I can verify and update this article if you supply specific source links or permit me to retrieve live documents. Overview: live bipedal trials move beyond demos — but not uniformly Humanoid robots entered 2025–26 trials with three themes: targeted task substitution in logistics, line‑side assistance in manufacturing, and human‑facing trials in eldercare and hospitality. Integrators, pilot funding from vendors and governments, and conservative customer KPIs determined which pilots moved from lab to floor. This report catalogs reported pilots and common operational patterns, highlights vendor positioning, and analyzes why many deployments remain tightly constrained. Industries and key use cases E‑commerce and third‑party logistics (case picking & replenishment): pick‑and‑place of small parcels, shelf replenishment, and zone transfer. Humanoids are trialed in narrow aisles and mixed‑SKU picking where wheel‑based cobots struggle with step up/step down tasks. Automotive and electronics manufacturing: repetitive assembly assistance, line‑side kitting, materials handover, and quality check walks where biped mobility aids dexterous access to line‑side tooling. Healthcare and eldercare: medication rounds, mobility‑assistance trials (standing support), and socially assistive concierge tasks in clinics and assisted‑living wings. Concierge and home trials: front‑desk reception, room delivery, and basic home assistance in controlled trial apartments. Which companies are at the front lines — positioning and pilots Vendors split into two positioning camps: general‑purpose humanoid platforms billed as “mobile co‑worker” (e.g., Agility Robotics’ Digit, Apptronik’s Apollo, Figure AI) and lower‑cost research‑forward entrants and mobile‑AI companies (e.g., 1X Technologies’ NEO/EVE, Unitree adaptations, Xiaomi’s CyberOne experiments). Public and vendor‑reported pilots through 2025–26 fell into the following patterns: Agility Robotics (Digit): marketed as a logistics co‑worker for pick/tote handling. Vendor materials and demo footage emphasize integration with warehouse systems and use of biped gait to navigate conveyor gaps and step‑over thresholds. Apptronik (Apollo): positioned toward industrial assistance — torque‑capable arms for line‑side handovers and human‑centric interaction. Apptronik emphasizes certification and safety‑engineered joints for factory floors. Figure AI: pitches autonomous generalists for retail and logistics; field trials (vendor‑reported) prioritized autonomy stacks for navigation in dynamic human environments. 1X Technologies (NEO/EVE), Unitree, Xiaomi (CyberOne): these players are pursuing lower‑cost entry or consumer‑facing experiments — Unitree with adaptations from legged platforms and Xiaomi leveraging smartphone/AI integration for concierge trials. Pilot program profiles and customer evidence Confirmed public pilots in 2025–26 fall into two categories: vendor‑announced commercial pilots and integrator‑led proofs funded through cost‑share. Examples (vendor‑reported): Large e‑commerce operators announced short‑run pilot programs using humanoids for replenishment in low‑height mezzanines — KPIs focused on throughput per hour and “hands‑free” percentage of restock events. Automotive tier suppliers ran line‑side trials with Apollo/Digit to test kitting cadence — metrics tracked included seconds per handover, cycle‑time variance, and defect‑rate impact. Several care‑home operators permitted concierge and medication‑reminder trials in dedicated wings under staff supervision; pilots reported staff augmentation (not replacement) goals and tracked resident acceptance scores. Note: specific commercial customer names and pilot lead quotes should be validated with primary press releases and integrator reports. Many vendors publish anonymized case studies; others require NDAs that limit public verification. KPIs, operational constraints and early ROI patterns Pilot teams converged on similar KPIs and encountered repeat constraints. Primary KPIs: throughput (orders picked/hour or kittings/hour), uptime (operational hours per shift), safety incidents (near‑miss and incident rates), and error rate (mis‑pick percentage). Labor metrics measured substitution ratio (FTE hours replaced) and augmentation effect on skilled staff. Operational constraints: safety zones and reduced speeds around humans; charging logistics (most pilots used tethered or swap‑pack approaches); sensor occlusion in dense racks; and brittle perception when floor layouts changed mid‑shift. Systems integration: WMS/ERP integration was a major friction point — many pilots used middleware from systems integrators to translate pick lists and update inventory in real time. Early ROI: vendor‑reported models show multi‑year payback dependent on 24/7 operations. In sample vendor materials, break‑even scenarios require >1.5 shifts/day continuous utilization or premium labor cost regions; many pilot customers report ROI as “unproven” until throughput and uptime exceed 60–70% of human rates. Pricing and business models Humanoid pricing remains variable. Vendor materials and early commercial offers through 2025–26 follow three models: Capital sale: unit prices (vendor‑reported/approximate) range from low‑six figures for higher‑capability industrial humanoids to under $100k for lower‑feature units — prices depend on manipulators, payload and autonomy stack. Robot‑as‑a‑Service (RaaS): monthly subscription (hardware + software + support) with performance SLAs. RaaS pricing is commonly offered to remove upfront cost barriers and align incentives for uptime. Integration contracts and pilot funding: systems integrators often underwrite early pilots; government grants and vendor cost‑share are common to de‑risk pilot investments for customers. Competitive comparison (high level) Vendor Strength Typical pilot focus Agility Robotics (Digit) Biped mobility optimized for logistics Case picking, shelf replenishment Apptronik (Apollo) Torque‑capable limbs, industrial safety Line‑side assistance, kitting Figure AI Autonomy stack, generalist positioning Retail, dynamic navigation 1X / Unitree / Xiaomi Lower price / consumer reach Concierge, home trials, light logistics Where deployments remain lab‑bound — and why Even in 2025–26, many humanoid projects stayed in controlled labs because: Regulatory and insurance hurdles for unsupervised human interaction. Unreliable perception in cluttered, changing environments. Integration complexity with legacy WMS/ERP and safety systems. Economic thresholds: insufficient utilization to justify capital or subscription costs. Conclusion and next steps for verification 2025–26 marks the transition from demonstrations to constrained, metric‑driven pilots. Real commercial uplift depends less on single‑robot capability and more on integration, uptime economics, and safety assurance. For a fully verified historical record, provide press releases, customer case studies or links to regulatory filings and I will reconcile vendor‑reported pilots with primary sources, add named customer testimonials where publicly available, and publish a verified timeline of 2025–26 field milestones.