Integrated Computer Vision and Decision Theoretic Modeling for Automated Assisted Living The goal of the COACH(Cognitive Orthosis for Assisting aCtivities in Home) project in the IATSL (Intelligent Assistive Technology and Systems Lab) is to develop automated computer vision systems to assist the human caregiver's role. Such systems would be able to non-invasively monitor the patient, stepping in to provide help in the form of verbal or visual prompts when necessary. Our research focusses on learning models of ADL behavior. The principal benefit of the model we describe is that it does not require patient behaviors to be labeled in video sequences. The learning method discovers the classes of behaviors present in the training data, and what their relationship is to the task state. The burden on human experts for the training of the system is thus reduced, for they only need to provide intermittent annotations of some small number of variables.

Wheelchair Collision Avoidance Using 3D Sensors This project investigates the use of 3D sensors for collision avoidance system for powered wheelchairs used by people with cognitive disabilities. Such systems increase mobility and feelings of independence, thereby enabling reversal of some symptoms of depression and cognitive impairment and improvement of quality of life. The collision avoidance system uses obstacle detections from 3D sensors to take actions that optimise over the desires of the user (to get somewhere) and the need for safety.

Decision Theoretic Models of Human Non-Verbal Behavior We are researching methods for learning decision theoretic models of facial expressions and gestures from video data. The meaning of faces and gestures for an observer is contained in their relationship to context, actions and outcomes. We study ways in which an agent can capitalize on these relationships by distinguishing facial displays and gestures according to their affordances, or how they help the agent to maximize utility. Partially observable Markov decision processes (POMDPs) are the basis of our modeling scheme, with outputs over entire video sequences of a person's face or hands.

Planning under uncertainty SPUDD is a stochastic planner which fuses Markov decision processes (MDPs) and algebraic decision diagrams (ADDs) to solve large planning problems. Representing MDPs as ADDs allows conditional independence in the Markovian dynamics to be exploited, giving large efficiency gains. Visit the SPUDD home page to find out more, read publications, submit examples to the SPUDD server, or download the SPUDD software. You can also jump directly to online SPUDD, where you can upload your own problem domain to the SPUDD server, and get results online.

Robotics The Robot Partners group at UBC is involved in creating embodied autonomous robotic agents that interact with people in real time. Our robots use a behavior based architecture. Our team created Jose the robotic waiter, winner of the 2001 AAAI Hors D'oeuvres competition.