Published Aug 22, 2018

Teaching Bots Learn by Watching Human Behavior - Ep. 67

Explore the transformative potential of robotics as hosts Marynel Vázquez and Animesh Garg delve into how robots learn from human behavior, tackle challenges in mobile robot autonomy, and envision a future where advanced sensor technology and neural task programming make robots an integral part of everyday life.
Episode Highlights
The AI Podcast logo

Popular Clips

Episode Highlights

  • Memory Constraints

    Mobile robots face significant challenges due to memory and processing power constraints. explains that while their robots perform well in brief occlusions, they struggle with long-term tracking due to limited onboard processing capabilities. She highlights the importance of optimizing computational resources to ensure real-time processing on mobile platforms 1.

    One of the things that I think is special about this project is that it is, at least to my knowledge, one of the first robots that has had a full onboard GPU to try to do as much as we can in terms of deep learning and computer vision mobile on the go.

    ---

    To address these issues, they are enhancing their robots with additional GPUs to improve performance and efficiency 1.

       

    Social Navigation

    Navigating social spaces is another complex challenge for robots. discusses their efforts to make robots polite and socially aware, such as adding a tie to symbolize politeness. She notes that people often interact with robots out of curiosity, which complicates the robots' navigation tasks 2.

    One of the first reactions that you get is that people get all excited about standing in front of the robot just to kind of tease and see how it would react to them.

    ---

    Their goal is to develop robots that can navigate crowded spaces without causing disruption, a task that requires sophisticated social engineering and human-robot interaction techniques 3.

       

    Tracking Humans

    Tracking human movement in crowded spaces is a significant technical hurdle. describes their approach, which involves using deep learning to predict human trajectories by considering appearance, velocity, and interaction with others. This method aims to improve tracking accuracy beyond mere pixel association 4.

    The idea was to think about not only the appearance of people, which is what typically the pixels directly tell you, but also think about the way that people are moving and also think about the way that they are interacting with other people in their surroundings.

    ---

    Additionally, understanding subtle human gestures can significantly enhance robot interaction, though this requires advancements in both hardware and software capabilities 5.

Related Episodes