Sim-to-real gap
The documented tendency for methods, particularly control policies trained through reinforcement learning or optimization, to perform differently, often worse, when transferred from simulation to physical hardware, due to unmodeled dynamics, sensor noise, and other real-world factors absent from the simulator. (see Topic 1.2) for first-principles reasoning about where a specific simulated result is likely to break on hardware.
Defined in 2 GAGE programs, which carry 11 distinct definitions of it. The wording above is taught in Engineering Judgment and Professional Formation.
How each discipline defines it
The same term does different work depending on who is using it. These are the definitions as each program teaches them, unedited.
The documented tendency for methods, particularly control policies trained through reinforcement learning or optimization, to perform differently, often worse, when transferred from simulation to physical hardware, due to unmodeled dynamics, sensor noise, and other real-world factors absent from the simulator. (see Topic 1.2) for first-principles reasoning about where a specific simulated result is likely to break on hardware.
The difference between how a robot behavior or learned policy performs in simulation and how it performs on physical hardware. Caused primarily by unmodeled friction and mechanical asymmetry, sensor noise characteristics that differ from simulation, and timing and latency differences between a simulated clock and a real system clock. Managing the sim-to-real gap is a central problem in Program 3's robot learning content.
The difference between robot behavior in simulation and behavior on real hardware. Arises from the approximations in any physical model: friction, inertia, sensor noise distributions, and communication timing are approximated rather than exactly replicated. Managed through domain randomization, calibrated simulation models, and targeted Phase 3 hardware tests.
The class of behavioral discrepancies that arise because a simulation approximates physical reality with simplified models of physics, timing, and sensor behavior. Not a hardware fault category. The diagnostic question is which physical property the algorithm depends on that differs between the simulated and real environments.
The specific instance of model mismatch that arises when a robot system validated in simulation is deployed in the real world and encounters physical dynamics, sensor behaviors, and environmental properties that differ from the simulation model.
Where it is taught
The exact lessons this term appears in. The first module of every program is free with a free account.
- Simulate first: the practice loop that elite robotics programs run on · Your Robotics Command Center, Foundations of Robotics and Physical AI
- Your map of the robotics path: the Launchpad you are in now and the road ahead · Your Robotics Command Center, Foundations of Robotics and Physical AI
- The debugging mindset and the safety habits to build from day one · Your Robotics Command Center, Foundations of Robotics and Physical AI
- Why robotics is hard: the world is noisy, partial, and unforgiving · What a Robot Actually Is, Foundations of Robotics and Physical AI
- Why simulation is where you learn, break things, and iterate safely · Your First Simulator to Practice In, Foundations of Robotics and Physical AI
- What changes the moment code leaves the simulator · From Code to Real Hardware, Foundations of Robotics and Physical AI
- Demonstrate, document, and defend: your first portfolio piece · Capstone: Build and Program a Simulated Mobile Robot, Foundations of Robotics and Physical AI
- When to trust a learned policy and when a classical controller must own the loop · Mechanical Intelligence, the Art of Removing Motors, Engineering Judgment and Professional Formation
- Reading a research paper and extracting what matters · Research, Benchmarking, and Staying at the Frontier, Engineering Judgment and Professional Formation
- Reproducing a result and learning why it works · Research, Benchmarking, and Staying at the Frontier, Engineering Judgment and Professional Formation
Where it shows up at work
A definition is where Sim-to-real gap starts. The lessons above teach it inside the competencies below, which are what job descriptions ask for and what the GAGE credential records. Every link here is drawn from the same careers map that builds a role's readiness check.
Explains training, tokens, context windows, embeddings, retrieval and fine-tuning well enough to ask an engineer a precise question and spot weak evidence.
2 lessons on this page teach it as a primary outcome.
Understands AI that acts in the physical world: perception stacks, vision language action models, sim to real transfer, sensor fusion, and where each of them fails.
One lesson on this page teach it as a primary outcome.
Reads bills and rules like a practitioner, compares options with costs and unintended effects, and writes briefs a decision-maker can act on.
One lesson on this page teach it as a primary outcome.
Roles that hire for it
Each role page shows the full competency set, what the seat decides, and a readiness check against your own evidence.
- Chief Technology Officer, AI engineering editionneeds it at depth
- AI Security Architectneeds it at depth · 8 open postings this week
- Robotics AI Engineer, embodied systems editionneeds it at depth
- AI Evaluation Specialistneeds it at depth · 6 open postings this week
- AI Governance Engineerneeds it at depth · 7 open postings this week
- AI Model Validatorneeds it at depth
How job descriptions say it
Postings rarely use the textbook word. These are the phrases the careers map reads as asking for the competencies above.
One interview question about it
Explain how a large language model produces an answer, at the depth a governance decision needs and no deeper.
Asked of candidates for how models work, at a governance depth. Answering it in your own words is exactly how a GAGE topic is passed, and the credential records the pass where an employer can check it.
If you come from another field
Some backgrounds already carry this competency, and the readiness check gives credit for it.
- Data analysis, stewardship and business intelligence carries it as a claim to back up: SQL and analytical fluency make the technical layer approachable.
- Model risk, credit risk and quantitative analysis carries it as work you can show: Quantitative fluency is demonstrated in your work.
- A computer science or software engineering degree carries it as work you can show: You have trained or integrated a model in coursework or a project.
Terms it appears with
Not an alphabetical neighbourhood: these are the terms taught in the same lessons, ranked by how often they appear together.