The health emergency caused by the Covid-19 pandemic has given new urgency to the long-standing objective of building machines which can help people carry out their physical work safely - even in environments which, once familiar, have suddenly become inaccessible or potentially hostile.

Fortunately, recent research advancements in the field of robotics have made it possible not only to have machines that approach or beat the computational intelligence of humans, but are also capable of ever more natural motion and exploit the "physical" intelligence embodied in their structure. Informed by neuroscientific models of human behavior in interaction with the physical world, new robots can safely touch humans and the environment to physically act on it. New sensing and display tools make it possible for other senses than just vision to share information on the world between a robot and a human. The union of such technologies, together with a deeper understanding of how to interface humans and machines, is enabling a new relationship between humans and robots, that is really more an integration than an interaction in the classical sense.

We will consider examples of partial integration, as in prosthetics and rehabilitation, augmentation with exoskeletons and supernumerary limbs, and shared-autonomy robotic avatars, with the robot executing the human's intended actions and the human perceiving the context of his/her actions and their consequences. Finally, we will discuss how human-robot integration could be leveraged to program the machine through "body language'' instructions which even robotics-naive users could intuitively give.

Session Chair: Jens Kober

There have been significant advances in the field of robot learning in the past decade. However, many challenges still remain when considering how robot learning can advance interactive agents such as robots that collaborate with humans. This includes autonomous vehicles that interact with human-driven vehicles or pedestrians, service robots collaborating with their users at homes over short or long periods of time, or assistive robots helping patients with disabilities. This introduces an opportunity for developing new robot learning algorithms that can help advance interactive autonomy.

In this talk, I will discuss a formalism for human-robot interaction built upon ideas from representation learning. Specifically, I will first discuss the notion of latent strategies— low dimensional representations sufficient for capturing non-stationary interactions. I will then talk about the challenges of learning such representations when interacting with humans, and how we can develop data-efficient techniques that enable actively learning computational models of human behavior from demonstrations, preferences, or physical corrections. Finally, I will introduce an intuitive controlling paradigm that enables seamless collaboration based on learned representations, and further discuss how that can be used for further influencing humans.

Session Chair: Claire Tomlin

To become intelligent and effective human helpers, robots must make complex decisions in highly dynamic and interactive human environments: driving through a crowded intersection with many pedestrians and vehicles, or picking up a coffee mug in a pile of different objects. Planning enables the robot to reason about the consequences of its actions far into the future, based on a prior model of the environment. However, it eventually succumbs to the combinatorial complexity of reasoning. Learning builds up past experiences to identify commonality among similar tasks and exploits the "remembered" solutions, but the challenge is then generalization to new, unseen tasks. In this talk, we will look at several ideas that try to fuse planning and learning by (i) inserting learning into planning, (ii) inserting planning into learning, and (iii) stacking them up in a hierarchy. In doing so, we achieve improved efficiency for planning, improved generalization for learning, and most importantly, scalable and generalizable robot decision making.

Session Chair: Fabio Ramos