POMDP-based long-term user intention prediction for wheelchair navigation

Abstract

This paper presents an intelligent decision-making agent to assist wheelchair users in their daily navigation activities. Several navigational techniques have been successfully developed in the past to assist with specific behaviours such as "door passing" or "corridor following". These shared control strategies normally require the user to manually select the level of assistance required during use. Recent research has seen a move towards more intelligent systems that focus on forecasting users' intentions based on current and past actions. However, these predictions have been typically limited to locations immediately surrounding the wheelchair. The key contribution of the work presented here is the ability to predict the users' intended destination at a larger scale, that of a typical office arena. The systems relies on minimal user input -obtained from a standard wheelchair joystick -in conjunction with a learned Partially Observable Markov Decision Process (POMDP), to estimate and subsequently drive the user to his destination. The projection is constantly being updated, allowing for true userplatform integration. This shifts users' focus from fine motorskilled control to coarse control broadly intended to convey intention. Successful simulation and experimental results on a real wheelchair robot demonstrate the validity of the approach.

Learning to Identify Users and Predict Their Destination in a Robotic Guidance Application

“…Based on this acquired knowledge, the system could infer the most probable goal from a set of possibly noisy observations. Taha et al [10] achieved similar results by fusing the robot position and the possible goals as the new state definition in a common POMDP. Vasquez [11] described a growing hidden Markov model algorithm dedicated to continuous learning, clustering, and making inference about car motions in a parking lot or people motions in a hallway.…”

“…Topological maps were successfully used in the works of Taha [10]. This environment representation is compact and matches the human's natural description of a path better than metric representations [7] (e.g.…”

Performance Evaluation Methods for Assistive Robotic Technology

“…Performance measures include time to completion (i.e., time to pass through the doorway), number of interactions, and number of collisions. Recent performance measures include accuracy, legibility, and gracefulness of the motion [15,91].…”

A probabilistic approach to learn activities of daily living of a mobility aid device user

“…Researchers have therefore resorted to explore different stochastic/probabilistic modelling techniques capable of representing the possible uncertainties involved in ADLs. Probabilistic models such as Bayesian Network (BN) [13], Dynamic Bayesian Network (DBN) [14],Partially Observable Markov Decision Process(POMDPs) [15], [16], [17], Hidden Markov Models (HMM) [18] and Hierarchical Hidden Markov Models (HHMM) [19] have been proposed in the literature.…”