Marcos Salganicoff

Tolerating concept and sampling shift in lazy learning using prediction error context switching

In their unmodified form, lazy-learning algorithms may have difficulty learning and tracking time-varying input/output function maps such as those that occur in concept shift. Extensions of these algorithms, such as Time-Windowed forgetting (TWF), can permit learning of time-varying mappings by deleting older exemplars, but have decreased classification accuracy when the input-space sampling distribution of the learning set is time-varying. Additionally, TWF suffers from lower asymptotic classification accuracy than equivalent non-forgetting algorithms when the input sampling distributions are stationary. Other shift-sensitive algorithms, such as Locally-Weighted forgetting (LWF) avoid the negative effects of time-varying sampling distributions, but still have lower asymptotic classification in non-varying cases. We introduce Prediction Error Context Switching (PECS) which allows lazy-learning algorithms to have good classification accuracy in conditions having a time-varying function mapping and input sampling distributions, while still maintaining their asymptotic classification accuracy in static tasks. PECS works by selecting and re-activating previously stored instances based on their most recent consistency record. The classification accuracy and active learning set sizes for the above algorithms are compared in a set of learning tasks that illustrate the differing time-varying conditions described above. The results show that the PECS algorithm has the best overall classification accuracy over these differing time-varying conditions, while still having asymptotic classification accuracy competitive with unmodified lazy-learners intended for static environments.

Active learning for vision-based robot grasping

Reliable vision-based grasping has proved elusive outside of controlled environments. One approach towards building more flexible and domain-independent robot grasping systems is to employ learning to adapt the robots perceptual and motor system to the task. However, one pitfall in robot perceptual and motor learning is that the cost of gathering the learning set may be unacceptably high. Active learning algorithms address this shortcoming by intelligently selecting actions so as to decrease the number of examples necessary to achieve good performance and also avoid separate training and execution phases, leading to higher autonomy. We describe the IE-ID3 algorithm, which extends the Interval Estimation (IE) active learning approach from discrete to real-valued learning domains by combining IE with a classification tree learning algorithm (ID-3). We present a robot system which rapidly learns to select the grasp approach directions using IE-ID3 given simplified superquadric shape approximations of objects. Initial results on a small set of objects show that a robot with a laser scanner system can rapidly learn to pick up new objects, and simulation studies show the superiority of the active learning approach for a simulated grasping task using larger sets of objects. Extensions of the approach and future areas of research incorporating more sophisticated perceptual and action representation are discussed

Gesture-speech based HMI for a rehabilitation robot

One of the most challenging problems in rehabilitation robotics is the design of an efficient human-machine interface (HMI) allowing the user with a disability considerable freedom and flexibility. A multimodal user direction approach combining command and control methods is a very promising way to achieve this goal. This multimodal design is motivated by the idea of minimizing the users burden of operating a robot manipulator while utilizing the users intelligence and available mobilities. With this design, the user with a physical disability simply uses gesture (pointing with a laser pointer) to indicate a location or a desired object and uses speech to activate the system. Recognition of the spoken input is also used to supplant the need for general purpose object recognition between different objects and to perform the critical function of disambiguation. The robot system is designed to operate in an unstructured environment containing objects that are reasonably predictable. A novel reactive planning mechanism, of which the user is an active integral component, in conjunction with a stereo-vision system and an object-oriented knowledge base, provides the robot system with the 3D information of the surrounding world as well as the motion strategies.

A virtual head-stick rehabilitation robot system

For individuals with upper-extremity motor disabilities, the head-stick is a simple and intuitive means of performing manipulations because it provides direct proprioceptive information to the user. Through practice and use of inherent proprioceptive cues, users may become quite adept at using the head-stick for a number of different tasks. The traditional head-stick is limited, however, to the users achievable range of head motion and force generation, which may be insufficient for many tasks. The authors describe an interface to a robot system which emulates the proprioceptive qualities of a traditional head-stick while also allowing for augmented end-effector ranges of force and motion. The design and implementation of the system in terms of coordinate transforms, bilateral telemanipulator architecture, safety systems, and system identification of the master is described, in addition to preliminary evaluation results.