DC14 - Bridging the Simulation and Reality Gap in Testing Marine Robots

Objectives

1. To build a testbed for running controlled physical tests in a multirobot marine environment, which facilitates measuring metrics necessary for evaluating the reliability of such systems. The test bed shall be easy to deploy in small bodies of water and allow data collection for comparing the results of running scenarios in simulation and real world.
2. To develop a test-case selection/prioritization technique which can select test scenarios, based on simulation results, that are more likely to fail in physical tests. The technique will use a model learned from the collected and constructed dataset of test executions in simulation and the physical testbed.

Expected Results

1. A testbed that can be used for experimentation where test run transfer from simulation to reality.
2. An effective test selection/prioritization technique for underwater robotics. The technique selects test cases that are more likely to fail in physical environment given the results in simulation.

Planned Secondment(s)

1. S27 EIVA, Dr. Brodsky, 2 months in M14-15
2. S28 UBremen, Prof. Hochgeschwender, 2 months in M38-39

Required Skills

Essential

• Computer Science/Engineering related degree
• Solid programming skills

Desirable

• Background in some of the following fields: Knowledge about ROS and robotics simulators, Probabilistic programs and Bayesian statistics, Practical experience with robotic systems