01
Choose the right robot
Compare wheeled, bipedal and legged platforms on your tasks instead of their spec sheets.
Work with us
Build your use case in simulation, with us
Bring us a task and a site. We work with your team to build the simulation assets for your use case: a model of your building, your fixtures and products, and the robot doing the job. You see what a humanoid can do on your floor before you buy one.
The problems that stop a robot deployment are physical and specific
An arm that cannot reach the far side of the table, or a fire door the robot cannot push. Simulation finds these while they are still cheap to fix.
Why start in simulation
02
Size the cell
Find reach limits, clearances and cycle times before anyone moves a conveyor or orders a fixture.
03
Find the hard parts early
Doors, lifts, shiny floors and people in corridors go into the model, and we test the robot against them.
04
Start training before delivery
Task software, demonstration data and policies from the study carry into the pilot, so the robot does not start from zero.
What we build
A working model of your site
We build the model from your drawings, photos and a day on site. One layout file generates the photoreal scene in NVIDIA Isaac Sim, the fast-physics world in MuJoCo and the maps, so all of them agree. A build check stops the model if furniture blocks a door or two objects overlap.
Then we add your products, your fixtures and the candidate robots, each running the same software it would run on site.
What we test
Your tasks, with every run recorded
We run reach studies, cycle-time runs and navigation through your doors and lifts. We inject failures, such as a dropped item, a blocked corridor or a lost signal, and check that the robot recovers or stops safely.
Every run is recorded, so each number in the report links back to the run that produced it.
How we work together
Five phases from first visit to pilot plan
The scope and timing depend on the site and the tasks. We agree both after the discovery phase, before the modelling starts.
- 01Discovery
We visit, walk the process with your team and agree the tasks, the success measures and the constraints on safety, space and shifts.
- 02Site model
We model the building, the fixtures, your products and the candidate robots.
- 03Trials
Reach studies, cycle-time runs, navigation, failure cases and recovery, all recorded.
- 04Architecture
Robot choice, cell layout, safety zones, links to your line control and warehouse systems, and the people around the robot.
- 05Pilot plan
A plan for the first robot on your floor, with acceptance tests taken from the simulation.
What you get
Evidence you can take to a budget meeting
A study ends with a clear recommendation: which robot, in which layout, doing which tasks, with the numbers behind it. Sometimes the recommendation is a different kind of automation, and we say so.
- Simulation assets for your use case: the site, fixtures, products and tasks
- Reach, clearance and cycle-time results for each candidate robot
- A robot shortlist with the reasons for each choice
- Cell layout drawings
- An integration and safety architecture
- Risks that simulation cannot settle, with a plan to test each one on hardware
- A pilot plan with acceptance tests
- Video of the robot doing your task in simulation, for your stakeholders
What simulation can and cannot tell you
Simulation answers reach, layout, cycle-time and navigation questions well. It is weaker on contact-rich handling of soft goods, such as bags of salad, and on how people behave around a robot in a busy space.
We flag those risks in the report and plan hardware tests for them early, so they do not surprise you in the pilot.
Questions
Before you start
Do we need to have chosen a robot?
No. Choosing the robot is part of the study. We work with off-the-shelf humanoids from several manufacturers, including Unitree and AgiBot, and compare them on your tasks.
What do you need from us?
Floor plans or drawings, photos of the area, samples or specifications of the products the robot will handle, a description of the process and a day on site with the people who run it.
Can the study lead straight into a deployment?
Yes. Our deployment software carries the site model, the task software and the training data straight into the pilot, all built on the robot’s real interface from the start.
What if the answer is that a humanoid is the wrong choice?
Then the report says so, and says what would work better. A clear no early on is a good outcome for a study.
How do you handle our drawings and site data?
We treat them as confidential and agree terms with you before the study starts.
Work with us on your use case
Tell us the task and the site. We will tell you which simulation assets to build first and what they will show.