Each card separates stable infrastructure from the failure surfaces that actually produce useful deployment data.
Pre-contracted factory environments for robot deployment. Labs pick a site. Ashmyr handles the physical environment.
Each card separates stable infrastructure from the failure surfaces that actually produce useful deployment data.
Operator coordination, site access, capture workflow, and repeatable reporting. Labs still bring the model and task definition.
Variance, reset burden, human traffic, safety envelope, and how quickly a failed rollout becomes an informative correction.
6 environments shown
Addis Ababa, Ethiopia
A repeatable assembly cell with stable BOM windows, narrow tolerances, and enough operator cadence to stress precision manipulation instead of raw mobility.
Warm-start
Medium
Low
Moderate
Electronics assembly line
Indoor, climate-controlled, fluorescent + natural lighting, 400m² floor space
8hrs/day, 6 days/week
Conveyor belts, component bins, soldering stations, inspection fixtures
Single-arm workcell, precision insertion arm, inspection cobot
1 site lead, 2 line operators available for corrections, QA supervisor on request
Alignment misses, bin depletion, cable routing, fixture occlusion, and micro-reset events around part presentation.
ESD-sensitive stations, shared aisle access, low-force interaction rules, no live soldering handoff.
Best first site for teams that want repeatability, crisp failure labels, and short retraining loops before moving into higher-variance environments.
Addis Ababa, Ethiopia
A faster line with humidity drift, bag variance, and packaging changeovers that expose perception and grasping systems to real production noise.
Warm-start
High
Medium
High
Food packaging plant
Indoor, variable humidity, mixed lighting (industrial + natural), 600m² floor space
10hrs/day, 6 days/week
Packaging lines, conveyor systems, palletizing stations, weight scales
Dual-arm pack-out cell, palletizing robot, end-of-line QA manipulator
Dedicated shift lead, 2 pack operators, maintenance contact during daytime hours
Bag deformation, slip events, label skew, pallet pattern drift, and conveyor timing mismatches during high-throughput runs.
Food-contact adjacency, moisture-sensitive perception zones, shared forklift corridor near pallet outfeed.
Strong candidate for labs that need a contracted site where throughput pressure and packaging variability reveal brittle grasping and sequencing behavior quickly.
Hawassa Industrial Park, Ethiopia
Large-floor textile operations introduce deformable materials, long work surfaces, and broad layout variation that punish narrow demonstrations.
Pilot-ready
High
Medium
High
Garment and textile production
Indoor, large open floor, high ceilings, mixed artificial lighting, 800m² allocated space
8hrs/day, 5.5 days/week
Cutting tables, sewing stations, fabric rolls, sorting bins, inspection tables
Mobile manipulator, overhead assist arm, sorting and inspection cell
1 site coordinator, floor supervisor, operator pair for annotation during pilot windows
Fabric drape ambiguity, fold-state uncertainty, table-to-bin transfer errors, and long-horizon material flow breakdowns.
Shared walkways, hanging material, moving carts, and occasional visual clutter from stacked fabric rolls.
Useful when the goal is not just more sites, but more deformation, longer task horizons, and layout spread inside one environment class.
Addis Ababa, Ethiopia
This site trades fixture stability for aisle navigation, shelving variance, and dense human activity around loading and replenishment cycles.
Pilot-ready
High
High
High
Distribution warehouse
Indoor, ambient temperature, overhead industrial lighting, 1,200m² floor space
12hrs/day, 6 days/week
Racking systems, conveyor belts, sorting stations, loading docks, forklifts
Mobile manipulator, tote-moving base, scanning and sorting platform
1 warehouse supervisor, 1 replenishment lead, shared safety spotter during live aisle tests
Occluded picks, shelf pose drift, barcode misses, tote congestion, and navigation resets caused by dynamic aisle blockage.
Forklifts, dock traffic, mixed pallet zones, and dynamic no-go areas during loading windows.
Best fit for teams that need navigation plus manipulation failure data, not just fixed-cell grasping examples.
Addis Ababa, Ethiopia
A smaller but more unforgiving shop where reflective surfaces, machine cadence, and sharper safety boundaries expose brittleness fast.
Pilot-ready
Medium
Medium
Moderate
Small-batch metal parts fabrication
Indoor, variable lighting (welding flash, overhead fluorescent), 350m² floor space
8hrs/day, 5.5 days/week
CNC machines, press brakes, grinding stations, tool racks, parts bins
Machine-tending arm, metrology assist cell, part-transfer manipulator
Shop foreman, machine operator, and shared maintenance technician during pilot windows
Reflective metal perception failures, part orientation mistakes, queueing mismatches at machine unload, and tolerance-check exceptions.
No entry into active machining zones, mandatory guarding near grinders, reflective glare during welding-adjacent operations.
High-value site for labs that want precision handling under stricter safety constraints and less forgiving visual conditions.
Debre Birhan, Ethiopia
The most environmentally noisy site in the catalog, with seasonal variation, wet surfaces, and object heterogeneity across produce batches.
Exploratory
High
High
Moderate
Agricultural processing facility
Semi-indoor, variable temperature and humidity, natural + artificial lighting, 500m² floor space
8hrs/day, 6 days/week
Washing lines, sorting conveyors, grading stations, cold storage access, crate stackers
Sorting cell, crate-loading arm, conveyor tending platform
1 processing manager, 2 line operators, flexible annotation windows during batch transitions
Size and texture shift across produce lots, water glare, conveyor jams, grade-boundary ambiguity, and seasonal workflow changes.
Wet floors, chilled storage adjacency, wash-line splash zones, and temporary crate staging that changes aisle geometry.
Best when the objective is broad distributional shift, messy sensing conditions, and a clear test of whether a policy survives beyond climate-controlled floors.