Reward Function Design for Industrial Manipulation Tasks
Closing the gap between what robots are rewarded to do and what you actually want them to do.
Daniel Ferreira-Lopes
Section
Robot learning, sim-to-real and on-the-job adaptation
7 stories in Robot learning, sim-to-real and on-the-job adaptation.
Continual Learning for Robots Handling New SKUs
Robots must learn new products without forgetting old ones to keep warehouses running.
Ingrid Solheim
Simulation Fidelity Requirements for Manipulation Policy Transfer
Researchers must identify which simulation parameters matter most for real-world robot grasping.
Daniel Ferreira-Lopes
Imitation Learning From Operator Demonstrations on the Line
Robots learn factory tasks directly from human demonstrations, avoiding hand-coded programming.
Daniel Ferreira-Lopes
Domain Randomization Strategies for Industrial Robot Training
Picking the right uncertainties to randomize matters more than randomizing everything.
Ingrid Solheim
Sim-to-Real Transfer Failures in Production Deployments
Specific simulator gaps cause real-world failures, not generic noise.
Daniel Ferreira-Lopes
On-Robot Fine-Tuning After Factory Deployment
Robots need real-world fine-tuning after deployment to handle novel conditions.
Theo Osei-Bonsu