Abstract:Humans achieve cognitive development through continuous interaction with their environment, enhancing both perception and behavior. However, current robots lack the capacity for human-like action and evolution, posing a bottleneck to improving robotic intelligence. Existing research predominantly models robots as one-way, static mappings from observations to actions, neglecting the dynamic processes of perception and behavior. This paper introduces a novel approach to robot cognitive learning by considering physical properties. We propose a theoretical framework wherein a robot is conceptualized as a three-body physical system comprising a perception-body (P-body), a cognition-body (C-body), and a behavior-body (B-body). Each body engages in physical dynamics and operates within a closed-loop interaction. Significantly, three crucial interactions connect these bodies. The C-body relies on the P-body’s extracted states and reciprocally offers long-term rewards, optimizing the P-body’s perception policy. In addition, the C-body directs the B-body’s actions through sub-goals, and subsequent P-body-derived states facilitate the C-body’s cognition dynamics learning. At last, the B-body would follow the sub-goal generated by the C-body and perform actions conditioned on the perceptive state from the P-body, which leads to the next interactive step. These interactions foster the joint evolution of each body, culminating in optimal design. To validate our approach, we employ a navigation task using a four-legged robot, D’Kitty, equipped with a movable global camera. Navigational prowess demands intricate coordination of sensing, planning, and D’Kitty’s motion. Leveraging our framework yields superior task performance compared with conventional methodologies. In conclusion, this paper establishes a paradigm shift in robot cognitive learning by integrating physical interactions across the P-body, C-body, and B-body, while considering physical properties. Our framework’s successful application to a navigation task underscores its efficacy in enhancing robotic intelligence.

Curiosity-driven Intuitive Physics Learning

Intuitive physics learning in a deep-learning model inspired by developmental psychology

Learning Intuitive Physics and One-Shot Imitation Using State-Action-Prediction Self-Organizing Maps

Learning physical parameters from dynamic scenes

Probing Physics Knowledge Using Tools from Developmental Psychology

Intrinsic Physical Concepts Discovery with Object-Centric Predictive Models

Emergence of Structured Behaviors from Curiosity-Based Intrinsic Motivation

Robot Cognitive Learning by Considering Physical Properties

Dynamic Neural Curiosity Enhances Learning Flexibility for Autonomous Goal Discovery

Toward an AI Physicist for Unsupervised Learning

Intrinsic Motivation Driven Intuitive Physics Learning using Deep Reinforcement Learning with Intrinsic Reward Normalization

Humanoid infers Archimedes' principle: understanding physical relations and object affordances through cumulative learning experiences

Intrinsically Motivated Acquisition of Modular Slow Features for Humanoids in Continuous and Non-Stationary Environments

Quick and Accurate Affordance Learning

Learning to Play Video Games with Intuitive Physics Priors

Active Inference, Curiosity and Insight

Discovering Latent States for Model Learning: Applying Sensorimotor Contingencies Theory and Predictive Processing to Model Context

Computational Theories of Curiosity-Driven Learning

Learning dynamical behaviors in physical systems

A dedicated mental resource for intuitive physics

Rapid trial-and-error learning with simulation supports flexible tool use and physical reasoning