As we deploy autonomous agents in safety-critical domains, it becomes important to develop an understanding of their internal mechanisms and representations. We outline an approach to imitation learning for reverse-engineering black box agent policies in MDP environments, yielding simplified, interpretable models in the form of decision trees. As part of this process, we explicitly model and learn agents’ latent state representations by selecting from a large space of candidate features constructed from the Markov state. We present initial promising results from an implementation in a multi-agent traffic environment.
|Publication status||Accepted/In press - 4 Sept 2020|
|Event||1st TAILOR Workshop at ECAI 2020 - Santiago de Compostela, Spain|
Duration: 4 Sept 2020 → 5 Sept 2020
|Workshop||1st TAILOR Workshop at ECAI 2020|
|Abbreviated title||TAILOR 2020|
|City||Santiago de Compostela|
|Period||4/09/20 → 5/09/20|