Sim2Real Reinforcement learning (RL)

Reinforcement learning (RL) has shown capabilities to deal with complex systems, including autonomous driving. However, most results are only in simulation or game environments as training/testing in real life is unsafe and expensive.

Sim2Real Reinforcement learning (RL)

Few is trying to make it work in simulation then transfer to the real system (sim2real) but using a simple and unstructured simulation environment causing large, challenging domain gaps.

We try to explore it furthur… in a structured way like standard testing process in automotive industry, and obtained some interesting results. You can see in the video an interesting demo after training RL using the X-in-the-Loop testing environments (where X is high-fidelity model, software, hardware, and vehicle), and combining with sim2real techniques like domain randomization and adaptation. Not super performance as model-based optimal control yet but it works from the first shot, safe, and fun… Hope you also enjoy it :).

If you are interested, see the paper draft here for details:  https://lnkd.in/ebATSaXr (with Kevin Voogd, Jean Pierre Allamaa, and Javier Alonso-Mora)

The demo uses the Siemens Simcenter Simrod vehicle. Check here for more innovative technologies that Siemens Digital Industries Software is developing on this platform: https://lnkd.in/eKCd5ear