Overcooked Plus: A Comprehensive Cooking Scenario TestBed for Enhancing the Evaluation of Autonomous Planning Algorithms

This paper introduces Overcooked Plus, a testbed designed to evaluate the performance of autonomous systems in complex and dynamic cooking scenarios. In these scenarios, single or multiple-agent systems must complete various cooking tasks such as fetching ingredients, handling kitchen utensils, chop-ping, cooking, and dishwashing. The testbed aims to offer a customizable platform that supports compre-hensive evaluation of planning algorithms and adaptive behaviors within autonomous systems. Key features of Overcooked Plus include controllable task difficulty, complex task constraints, multitask planning, dynamic environments, a graphical interface for human-on-the-loop, and a communication interface for multiagent systems. The paper details the system architecture and implementation of Overcooked Plus, and presents experimental setups to demonstrate its usage. GitHub: github.com/545659928/Overcooked-Plus