DC11 - Retrieval Augmented Scenario Generation for Safety Validation of Human-Robot-Collaboration

Objectives

1. To endow LLMs with structured safety and Human-Robot-Collaboration (HRC) risk knowledge from standards, enabling assisted specification of HRC validation scenarios.
2. To derive novel failure-inducing HRC scenarios from data (like execution traces in the field and simulations).

Expected Results

1. Open-source LLM which enables the downstream tasks of specifying and generating HRC scenarios.
2. An accessible repository of failure-inducing HRC scenarios, produced by the LLM-assisted scenario engine and corresponding field data.

Planned Secondment(s)

1. S21 LNE, Dr. Kalouguine, 2 months in M14-15
2. S22 UYork, Prof. Gerasimou, 2 months in M26-27

Required Skills

Essential

• Computer Science/Engineering degree
• Excellent programming skills (C++, Python etc.)

Desirable

• Knowledge in ROS and robot simulators (e.g. Gazebo, IsaacSim, etc.)
• Practical experience with LLMs and developing/testing complex robotic software systems