Cora: An Anthropomorphic Robot Assistant for Human Environment

We describe the general concept, system architecture, hardware, and the behavioral abilities of Cora (Cooperative Robot Assistant, see Fig. 1), an autonomous non mobile robot assistant. Outgoing from our basic assumption that the behavior to perform determines the internal and external structure of the behaving system, we have designed Cora anthropomorphic to allow for humanlike behavioral strategies in solving complex tasks. Although Cora was built as a prototype of a service robot system to assist a human partner in industrial assembly tasks, we will show that Cora’s behavioral abilities are also conferrable in a household environment. After the description of the hardware platform and the basic concepts of our approach, we present some experimental results by means of an assembly task. 1 Why anthropomorphic autonomous service robots? Robots in the sense of human-like general purpose machines that are able to perform well in a human environment and can handle the various badly defined everyday tasks we humans spend a lot of time with are still a matter of the future. Besides the fascination to create our own technical analogon, the design of such robots poses a number of problems that are crucial for a lot of applications. The methods that need to be developed as a presupposition for successful robot design are important for many fields that at first glance have nothing to do with robotics. This includes automatic data acquisition, process control, user assistance for handling complex tasks like e.g. driving, and certainly entertainment. Robot design calls for a very high integration of mechanical, electrical, electronical and computational technology. Contrary to how most robots are constructed today, a look at the biological prototype reveals that not the extensive use of highly precise sensors is required, but that a few less perfect but Fig. 1. The serviceand assistance robot Cora. A seven DoF manipulator arm is mounted on a one DoF trunk which is fixed on a table. The robot possesses a two DoF stereo camera head with microphones. universally applicable input channels namely vision, audio and touch can be used to great effect. As useful as a collection of special purpose devices for distance measurement, illumination etc. may be in the short run, they imply additional system load and unnecessary complexity for special cases and thus are a dead end for the development of more generic systems. Another common misconception for intelligent systems is the call for complete user or operator control. Total and detailed control over every aspect of the system is something that a typical classical computer system offers. This is also the reason why those systems are hard to use, require a lot of knowledge and training and are nevertheless easy to make errors on, i.e. to generate command with a different outcome than that intended by the user. Our interaction with more intelligent ”systems” like e.g. a trained dog is very different from that: Although the communication channel speech and gestures is highly redundant, the syntax and semantics are mostly ill-defined and the animal is never guaranteed to even obey a given command, the interaction is quite intuitive, robust even under difficult conditions and generally pretty powerful. The dog is easy to deal with because it will generally use its own intelligence and only do what his master requests if the commands are not in contradiction to his own ”behavioral goals”. A dog that is commanded ”come here straight” and runs into a wall because it obeys the command would be called particularly stupid with good reason. The same holds true for a robot. Combining real autonomy of behavior with a user interface may be quite straightforward by defining the command input on the same level as the sensory input and the general behavioral goals; such as e.g. obstacle avoidance. Besides general purpose service robotics there is also the field of robotics for the disabled. Mobility assistances for the blind and wheelchairmounted robot arms are only two examples posing a technical challenge for highly robust robot control which allows the user to easily specify tasks. One possible and maybe ideal approach for robotic tools for the disabled would be a library of semiautonomous behaviors that perform tedious subtasks like e.g. locating an object for grasping or determining trajectories. Thus robotics for the disabled share the concept of behaviors or skills as coarsely parameterized atoms by which more complex tasks can be successfully performed. Anthropomorphic shape and the fluent, predictable movements of effectors controlled by dynamical systems rather than stiff trajectories also have a strong effect on users which makes working with such a robot anesthetically and emotionally more pleasing. 2 ”Anatomy” Body Shape, Sensors