Quantify Sim-to-Real Gap for a Warehouse Manipulation Policy
Overview
What this challenge is about.
You receive a trained pick-and-place policy (PyTorch), the simulation env (Isaac Lab), and access to a real-arm rig (or recorded teleop episodes if hardware is unavailable). Define a matched set of 20 bin scenes that exist in both sim and real. Run the policy 50 times per scene in each, log success/failure and failure mode. Use a causal-style breakdown (perception vs. dynamics vs. contact) to attribute the gap. Deliver a research memo with a ranked list of root causes and proposed next steps.
The Brief
What you'll do, and what you'll demonstrate.
Quantify the sim-to-real gap of a bin-picking policy on matched scenes and rank the top causes by impact.
Earning criteria — what you'll demonstrate
- Design a controlled sim-to-real comparison with matched conditions
- Attribute policy failures to perception, dynamics, or contact
- Run a statistically meaningful number of trials per condition
- Communicate research findings to engineering leadership
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
ML Researcher
Designing and running a controlled sim-to-real study with honest reporting is the daily reality of applied ML research in robotics.
This challenge sharpens
- sim-to-real
- experiment-design
- policy-evaluation
Research Scientist
Ablation-based attribution and statistical rigor mirror the standards expected of a junior research scientist's first project.
This challenge sharpens
- experiment-design
- policy-evaluation
- sim-to-real
Applied AI Scientist
Translating a research finding into a ranked, actionable memo for a CTO is the hallmark of applied-AI scientist work.
This challenge sharpens
- sim-to-real
- experiment-design
- manipulation