**Habitat: A Platform for Embodied AI Research** **Authors:** Manolis Savva, Abhishek Kadian, Oleksandr Maksymets, Yili Zhao, Erik Wijmans, Bhavana Jain, Julian Straub, Jia Liu, Vladlen Koltun, Jitendra Malik, Devi Parikh, Dhruv Batra **Abstract:** Habitat is a platform designed for research in embodied artificial intelligence (AI). It enables the training of virtual robots in highly efficient photorealistic 3D simulations. The platform consists of two main components: Habitat-Sim, a flexible and high-performance 3D simulator with configurable agents, sensors, and generic 3D dataset handling, and Habitat-API, a modular high-level library for end-to-end development of embodied AI algorithms. Habitat-Sim achieves several thousand frames per second (fps) when rendering scenes from Matterport3D datasets, and can reach over 10,000 fps multi-process on a single GPU. Habitat-API allows for defining tasks, configuring and training embodied agents, and benchmarking using standard metrics. **Key Contributions:** 1. **Performance:** Habitat-Sim outperforms existing simulators by orders of magnitude, achieving thousands of fps. 2. **Task Definition:** Habitat-API supports a wide range of tasks, including navigation, instruction following, and question answering. 3. **Generalization:** Experiments on point-goal navigation show that learning-based agents outperform classical SLAM approaches when scaled to a larger amount of experience. Additionally, agents equipped with depth sensors generalize well across different 3D datasets. **Habitat Challenge:** The Habitat Challenge is an autonomous navigation challenge that aims to benchmark and advance efforts in goal-directed visual navigation. It leverages the EvalAI platform for submission and evaluation, allowing participants to upload Docker containers with their agents. **Future Work:** Future developments will focus on incorporating physics simulation, enabling physics-based interaction between agents and objects, and procedural generation of 3D environments. Distributed simulation settings involving multiple agents will also be explored. **Conclusion:** Habitat provides a unified and flexible platform for embodied AI research, enabling significant advancements in the field through efficient simulation, task definition, and benchmarking.**Habitat: A Platform for Embodied AI Research** **Authors:** Manolis Savva, Abhishek Kadian, Oleksandr Maksymets, Yili Zhao, Erik Wijmans, Bhavana Jain, Julian Straub, Jia Liu, Vladlen Koltun, Jitendra Malik, Devi Parikh, Dhruv Batra **Abstract:** Habitat is a platform designed for research in embodied artificial intelligence (AI). It enables the training of virtual robots in highly efficient photorealistic 3D simulations. The platform consists of two main components: Habitat-Sim, a flexible and high-performance 3D simulator with configurable agents, sensors, and generic 3D dataset handling, and Habitat-API, a modular high-level library for end-to-end development of embodied AI algorithms. Habitat-Sim achieves several thousand frames per second (fps) when rendering scenes from Matterport3D datasets, and can reach over 10,000 fps multi-process on a single GPU. Habitat-API allows for defining tasks, configuring and training embodied agents, and benchmarking using standard metrics. **Key Contributions:** 1. **Performance:** Habitat-Sim outperforms existing simulators by orders of magnitude, achieving thousands of fps. 2. **Task Definition:** Habitat-API supports a wide range of tasks, including navigation, instruction following, and question answering. 3. **Generalization:** Experiments on point-goal navigation show that learning-based agents outperform classical SLAM approaches when scaled to a larger amount of experience. Additionally, agents equipped with depth sensors generalize well across different 3D datasets. **Habitat Challenge:** The Habitat Challenge is an autonomous navigation challenge that aims to benchmark and advance efforts in goal-directed visual navigation. It leverages the EvalAI platform for submission and evaluation, allowing participants to upload Docker containers with their agents. **Future Work:** Future developments will focus on incorporating physics simulation, enabling physics-based interaction between agents and objects, and procedural generation of 3D environments. Distributed simulation settings involving multiple agents will also be explored. **Conclusion:** Habitat provides a unified and flexible platform for embodied AI research, enabling significant advancements in the field through efficient simulation, task definition, and benchmarking.