Shielding Atari Games with Bounded Prescience

We present the first explicit-state method for analysing and ensuring the safety of DRL agents for Atari games. Our method only requires access to the emulator. We give a suite of 42 properties that characterise “safe behaviour” for 31 games. We evaluate the safety of the best available DRL agents which, as our experiments show, violate most of our properties. We propose a countermeasure that implements shielding using bounded explicit-state exploration. Our method improved their overall safety, producing the safest DRL agents for Atari games currently available.

Authors

• Mirco Giacobbe University of Birmingham
• Mohammadhosein Hasanbeig
• Daniel Kroening
• Hjalmar Wijk

Keywords

• Safe AI
• Deep Reinforcement Learning
• Atari Games