|
Behavior-based systems are a very popular approach for robot control, and have very good robustness and real-time properties in dynamic, unstructured environments. However, to date they have been used for relatively simple tasks, which do not require a complex sequencing of the robot's behaviors. Also, behaviors are typically redesigned to encode the specifics (sequencing constraints) of any particular task. This work presents an architecture that extends the capabilities of behavior-based systems and addresses the above limitations. The key features of this architecture are: 1) the ability to encode and execute complex, sequential, hierarchically structured tasks within a behavior-based framework; 2) behavior reusability across different tasks; 3) support for automatic generation of a behavior-based system; 4) means for sequential and opportunistic task execution. [Details]
|
|
This project presents an interactive, natural method for robot teaching by demonstration, inspired from the approach people use when teaching each other. The key features of this robot teaching method are that: 1) it relies on a concurrent use of multiple means for interaction and learning, both on the part of the teacher and of the learner, similarily with the human-human teaching approach; 2) it enables the learning of high-level task representations of complex tasks, thus having major advantages in dynamic, changing environments. For effective teaching and learning approaches, during demonstrations, people most often make use of simple instructions and informative cues, that bias the learner's attention to the the relevant parts of the task. Also, after an initial demonstration, "students" usually may refine their acquired capabilities by practicing under the teacher's supervision. Depending on the quality of the learned task, the teacher may either demonstrate it again or provide specific feedback during the learner's practice trial for further refinement. Thus, instructive demonstrations, generalization over multiple demonstrations and practice trials are essential features for a successful human-robot teaching approach. [Details] [Poster]
|
|
The goal of this project is to develop an action-based framework for interaction between humans and robots that extends the communication strategies typically used in the mobile robotic systems domain. The challenge we are addressing is allowing the interaction to occur without the need of an explicitly shared vocabulary between the robot and the human. Performing actions produces outcomes that are easy for people to understand through their common sense. Actions carry intentional meanings, and therefore they could be used as a "vocabulary" for a robot to help express its intentions. We apply this strategy to enable a mobile robot to capture a human's attention and induce changes in a human's behavior in order to receive help when unexpected situations occur during the execution of its task. [Details]
|
|
Created by: Monica NICOLESCU (e-mail: monica@cs.unr.edu) Last update: 08/22/2003 |