Relevant Degree Programs
Graduate Student Supervision
Doctoral Student Supervision (Jan 2008 - Mar 2019)
Cognitive impairments prevent older adults from using powered wheelchairs because of safety concerns, thus reducing mobility and resulting in increased dependence on caregivers. An intelligent powered wheelchair system (NOAH) is proposed to help restore mobility, while ensuring safety. Machine vision and learning techniques are described to help prevent collisions with obstacles, and provide reminders and navigation assistance through adaptive audio prompts. The intelligent wheelchair is initially tested in various controlled environments and simulated scenarios. Finally, the system is tested with older adults with mild-to-moderate cognitive impairment through a single-subject research design. Results demonstrate the high diversity of the target population, and highlight the need for customizable assistive technologies that account for the varying capabilities and requirements of the intended users. We show that the collision avoidance module is able to improve safety for all users by lowering the number of frontal collisions. In addition, the wayfinding module assists users in navigating along shorter routes to the destination. Prompting accuracy is found to be quite high during the study. While compliance with correct prompts is high across all users, we notice a distinct difference in the rates of compliance with incorrect prompts. Results show that users who are unsure about the optimal route rely more highly on system prompts for assistance, and thus are able to improve their wayfinding performance by following correct prompts. Improvements in wheelchair position estimation accuracy and joystick usability will help improve user performance and satisfaction. Further user studies will help refine user needs and hopefully allow us to increase mobility and independence of several elderly residents.
Master's Student Supervision (2010-2017)
Mobility is one of the most significant factors that determines older adults’ perceived level of health and well being. Cognitively impaired older adults are deprived of using powered wheelchairs because of the operational safety risks. These users can benefit from intelligent assistance during cognitively or visually challenging tasks such as back-in parking. An intelligent powered wheelchair that assists a cognitively impaired elderly user to perform a back-in parking task is proposed. A single subject participatory action design method is used with a cognitively impaired older adult to identify design guidelines for the proposed system. Based on analysis of transcripts from semi-structured interviews with the participant, a semi-autonomous back-in parking system is designed to drive the powered wheelchair into a pre-specified back-in parking space when the user commands it to. A prototype of a non-intrusive steering guidance feature for a joystick handle is also designed to render shear force in a way that can be associated with steering behavior of a car. The performance of the proposed system is evaluated in a pilot study. Experiments with the autonomous trigger and autonomous assisted modes are conducted during a back-in parking task with real-life obstacles such as tables and chairs in a long-term care facility. A single-subject research design is used to acquire and analyze quantitative data as a pilot study. Results demonstrate an increase in the user’s perception of ease of use, effectiveness and feeling of safety with the proposed system. While the user experienced at least one minor contact in 37.5% of the trials when driving unaided, the proposed system eliminated all minor contacts. No statistically significant difference in completion time and route length is observed with the proposed system. In the future, improved back-in parking systems can use this work as a benchmark for single subject participatory action design. Future iterations could also replicate the usability study on a larger population.
One of the major challenges in the field of Artificial Intelligence is dealing with uncertainty. Finding the optimal solution in the presence of uncertainty is computationally quite costly. This makes it impossible to solve large problems. In this thesis, we propose a new heuristic, named the pairwise heuristic, which efficiently finds a near-optimal solution for such problems. The pairwise heuristic is based on optimal solutions for the pairs of states. For each pair, it solves the problem assuming that the uncertainty exists only between the two states of the pair. A greedy online strategy uses these solutions to solve the main problem. We tested the pairwise heuristic on two problems where uncertainty plays a major role, i.e., localization and planning under uncertainty. Our achievements in connection with both problems are novel in their respective fields. In the field of localization, we have developed an efficient method to localize a robot in any kind of environment in a fullyautonomous way. In the field of planning under uncertainty, our method finds a near-optimal solution in a time shorter than the time required by any other current method in the field.
Model-based reinforcement learning methods make efficient use of samples bybuilding a model of the environment and planning with it. Compared to model-freemethods, they usually take fewer samples to converge to the optimal policy. Despite that efficiency, model-based methods may not learn the optimal policy due tostructural modeling assumptions. In this thesis, we show that by combining model-based methods with hierarchically optimal recursive Q-learning (HORDQ) undera hierarchical reinforcement learning framework, the proposed approach learns theoptimal policy even when the assumptions of the model are not all satisfied. Theeffectiveness of our approach is demonstrated with the Bus domain and InfiniteMario – a Java implementation of Nintendo’s Super Mario Brothers.
In this dissertation, we study the problem of knowledge reuse by a reinforcement learning agent. We are interested in how an agent can exploit policies that were learned in the past to learn a new task more efficiently in the present. Our approach is to elicit spatial hints from an expert suggesting the world states in which each existing policy should be more relevant to the new task. By using these hints with domain exploration, the agent is able to detect those portions of existing policies that are beneficial to the new task, therefore learning a new policy more efficiently. We call our approach Spatial Hints Policy Reuse (SHPR). Experiments demonstrate the effectiveness and robustness of our method. Our results encourage further study investigating how much more efficacy can be gained from the elicitation of very simple advice from humans.