Home * Call for Papers * Committees * Registration * Hotel * Paper Submission * List of Papers * General Information * Contact * Final Program * Travel * Balatonf├╝red

Plenary Lecture

Oussama Khatib:

The Human Frontier: Robotics New Quest and Challenge


Oussama Khatib
Department of Computer Science
Stanford University, Stanford, USA
ok@cs.stanford.edu

Summary
Robotics is rapidly expanding into human environments and vigorously engaged in its new emerging challenges. Interacting, exploring, and working with humans, the new generation of robots will increasingly touch people and their lives. The successful introduction of robots in these environments will rely on the development of competent and practical systems that are dependable, safe, and easy to use. To effectively work and cooperate with a person, robots must display abilities and skills that are compatible with those of humans. The discussion focuses on the ongoing effort for the design and development of human-friendly robotic systems that can safely and effectively interact and work with humans.
A major component in these developments is a new framework for the modeling and control of complex human-like robotic systems. In this framework, the various problems associated with (i) the motion coordination of the large number of degrees of such robots; (ii) the effective control of their contacts and interactions with the environment; (iii) the maintenance of their internal and external constraints; (iv) and the strategies for dealing with their underactuation and balance are all treated in a unified fashion within a general whole-body control structure. This is a task-oriented control structure that addresses the whole body dynamics for specifications involving multiple distributed tasks and postures in consistency with the requirements of multiple distributed contacts and constraints.
The second component in this effort is concerned with the synthesis of natural human movements to produce human-like robot behaviors. The objective is to unveil the underlying characteristics of human motion through an elaboration of its physiological basis. The aim is to formulate general strategies for whole-body robot behaviors. This exploration has employed models of human musculoskeletal dynamics and made use of extensive experimental studies of human subjects with motion capture techniques. Our study of human motion has revealed the dominant role physiology plays in shaping human motion. The characteristics of human motion revealed in this study have allowed the development of generic motion creteria that efficiently and effectively encode human motion behaviors.
The third component in our effort is concerned with the critical issue of safety in robot design. Our work in human-friendly robot design has led to the development of a new actuation methodology which has been shown to be well-suited for the emerging generation of robots conceived to operate in human environments. This methodology of distributed macro mini actuation, $DM^2$, addresses both the safety and performance characteristics of a robot. The approach has led to the design and construction of several prototypes, the last of which is a two-arm on a common torso robotic testbed. This new system represents a unique platform to explore the competing issues of safety and performance in the design of robot mechanisms. The new two-arm torso testbed is being used to validate the promise of safety and performance and to establish meaningful measures for safety and performance. Of particular interest is the analysis of impact forces in a three dimensional collision between a robot and its surroundings. Two safety standard measures are used to quantify the improvement in safety in terms of reduction of impact force, while the robot performance characteristics are evaluated agains traditional design.
Other fundamental issues in human-centered robotics will be also examined in this presentation. These include the elastic planning methodology for real-time modifications of existing plans, and various other effective methodologies and efficient algorithms that address the computational challenges associated with human-like robotic structures.