@article {pub3878,
	title = {Reasoning about Uncertain Parameters and Agent Behaviors through Encoded Experiences and Belief Planning},
	author = {Akinobu Hayashi AND Dirk Ruiken AND Christian Goerick AND Tadaaki Hasegawa},
	year = {2019},
	month = {December},
	abstract = {Robots are expected to handle increasingly complex tasks. Such tasks often include interaction with novel objects or collaboration with unfamiliar agents. One of the key challenges for reasoning in such situations is the lack of accurate models that hinders the effectiveness of planners. We propose a system for online model adaptation that corrects hypothetical types and parameters of the models based on observations and encoded prior experiences. We encode prior experiences into a recurrent neural network to generate possible types and parameters of the models. An online POMDP solver is used to plan actions to complete the task while progressively validating and improving the models. We perform experiments both in collaborative tasks with multiple unknown agents and in object manipulation tasks with unknown objects and environments. We compare against state-of-the-art methods, and the results show a better success rate of the proposed approach. Additionally, the approach achieves a better estimation of types and parameters of other agents and is capable of handling dynamics with discontinuities.},
	publisher = {Elsevier},
	journal = {Artificial Intelligence, Special Issue on Autonomous Agents Modelling Other Agents}
}