Autonomous Robotics and Deep Learning (SpringerBriefs in Computer Science)

This Springer Brief examines the combination of computer vision techniques and machine learning algorithms necessary for humanoid robots to develop “true consciousness.” It illustrates the critical first step towards reaching “deep learning,” long considered the holy grail for machine learning scientists worldwide. Using the example of the iCub, a humanoid robot which learns to solve 3D mazes, the book explores the challenges to create a robot that can perceive its own surroundings. Rather than relying solely on human programming, the robot uses physical touch to develop a neural map of its environment and learns to change the environment for its own benefit. These techniques allow the iCub to accurately solve any maze, if a solution exists, within a few iterations. With clear analysis of the iCub experiments and its results, this Springer Brief is ideal for advanced level students, researchers and professionals focused on computer vision, AI and machine learning.

這本Springer簡報探討了結合電腦視覺技術和機器學習演算法對於人形機器人發展「真正意識」所必需的。它展示了達到「深度學習」的關鍵第一步，這一直被全球機器學習科學家視為聖杯。本書以iCub為例，一個學習解決3D迷宮的人形機器人，探討了創造一個能感知自身環境的機器人所面臨的挑戰。機器人不僅依賴於人類編程，還利用物理觸摸來建立自己環境的神經地圖，並學會改變環境以獲得自身利益。這些技術使得iCub能夠在幾次迭代中準確解決任何迷宮問題（如果解存在的話）。通過對iCub實驗及其結果的清晰分析，這本Springer簡報非常適合專注於電腦視覺、人工智能和機器學習的高級學生、研究人員和專業人士閱讀。