Autonomous Embodied Agents: When Robotics Meets Deep Learning Reasoning

TL;DR

This work surveys and advances Embodied AI by addressing representation learning, intrinsic motivation, and sim-to-real transfer for autonomous agents. It introduces an impact-based exploration framework with a neural mapper and pose estimator, validated on Gibson and MP3D, and demonstrates real-world deployment on LoCoBot. It further extends the paradigm with Explore and Explain, a Transformer-based captioning module, and a unified pipeline for efficient exploration and scene description, evaluated via a novel ED S metric. The Spot the Difference dataset and AG3D gallery environment broaden benchmarks for changing environments and long-horizon navigation. Collectively, the thesis delivers methodologies, benchmarks, and real-world insights that support practical deployment of embodied agents in complex indoor spaces, including museums and office environments.

Abstract

The increase in available computing power and the Deep Learning revolution have allowed the exploration of new topics and frontiers in Artificial Intelligence research. A new field called Embodied Artificial Intelligence, which places at the intersection of Computer Vision, Robotics, and Decision Making, has been gaining importance during the last few years, as it aims to foster the development of smart autonomous robots and their deployment in society. The recent availability of large collections of 3D models for photorealistic robotic simulation has allowed faster and safe training of learning-based agents for millions of frames and a careful evaluation of their behavior before deploying the models on real robotic platforms. These intelligent agents are intended to perform a certain task in a possibly unknown environment. To this end, during the training in simulation, the agents learn to perform continuous interactions with the surroundings, such as gathering information from the environment, encoding and extracting useful cues for the task, and performing actions towards the final goal; where every action of the agent influences the interactions. This dissertation follows the complete creation process of embodied agents for indoor environments, from their concept to their implementation and deployment. We aim to contribute to research in Embodied AI and autonomous agents, in order to foster future work in this field. We present a detailed analysis of the procedure behind implementing an intelligent embodied agent, comprehending a thorough description of the current state-of-the-art in literature, technical explanations of the proposed methods, and accurate experimental studies on relevant robotic tasks.

Figures (25)

• Figure 1: We propose an impact-based reward for robot exploration of continuous indoor spaces. The robot is encouraged to take actions that maximize the difference between two consecutive observations.
• Figure 2: Our modular exploration architecture consists of a Mapper that iteratively builds a top-down occupancy map of the environment, a Pose Estimator that predicts the pose of the robot at every step, and a hierarchical self-supervised Navigation Module in charge of sequentially setting exploration goals and predicting actions to navigate towards it. We exploit the impact-based reward to guide the exploration and adapt it for continuous environments, using an Observation Encoder to extract observation features and depending on the method, a Density Model or a Grid to compute the pseudo-count.
• Figure 3: Qualitative results. For each model, we report three exploration episodes on Gibson and mp3d datasets for $T=500$. The exploration capabilities of the Impact-based models are higher than the baselines, in particular in larger environments.
• Figure 4: We propose a novel setting in which an embodied agent performs joint curiosity-driven exploration and explanation in unseen environments. While navigating the environment, the agent must produce informative descriptions of what it sees, providing a means of interpreting its internal state.
• Figure 5: Overview of our ex2 framework for navigation and captioning. Our model is composed of three main components: a navigation module which is in charge of exploring the environment, a captioning module that produces a textual sentence describing the agent point of view, and a speaker policy that connects the previous modules and activates the captioning component based on the information collected during the navigation.