CRL Lab Overview
The Cognitive Robotics Laboratory (CRL) conducts research on humanoid robots, service robots, multi-robot teams, and human-robot symbiosis.
The main humanoid robot platform within the CRL is ISAC. ISAC stands for Intelligent SoftArm Control, a reference to its unusual pneumatically-powered arms, which are designed to allow safe interaction between the robot and people. ISAC can understand voice commands and respond with synthesized speech. Additionally, ISACís software is built on an agent-based paradigm containing short and long-term memory structures roughly modeled after the human brain.
Research work on ISAC is proceeding in a variety of directions. Though these directions may appear quite different, they are in fact tightly coupled with each other around the goal of developing an autonomous, intelligent robot capable of perceiving the world around it as well as learning to interact with that world.
ISAC is termed a cognitive robot. ISAC utilizes a variety of memory structures, learning algorithms, perceptual units, and decision making policies organized in a structured agent environment. Current agents in development include the human agent, self agent, central executive agent, and emotion agent. Current memory structures include a long term memory, a short term memory, an episodic memory (for remembering particular episodes or experiences), and a working memory system.
Exchanging verbal greetings with a person and shaking hands with that person is an example of some of the work that has been done on ISAC in the past. As the research on ISAC progresses, ISAC is taught how to learn other tasks and behaviors as well. Utilizing the cognitive architecture, ISAC will be able to prioritize tasks and discern generalizations about tasks.
In addition to the cognitive architecture research being performed, ISAC is also learning behaviors through simple finger pointing tasks (i.e. a person teaches ISAC the location of objects by pointing at them) and through the use of sensory-motor coordination (SMC).
SMC is very interesting in that it couples sensory information with motor action within the context of a particular task. We believe that intelligence emerges in a situated and embodied machine as a result of interaction and development. One of our fundamental hypotheses in this research is that sensory-motor coordination is the basis of intelligence. This is supported by the fact that SMC descriptors cluster, or self-organize, into categories.