Christopher G. Atkeson

Locally Weighted Learning

This paper surveys locally weighted learning, a form of lazy learning and memory-based learning, and focuses on locally weighted linear regression. The survey discusses distance functions, smoothing parameters, weighting functions, local model structures, regularization of the estimates and bias, assessing predictions, handling noisy data and outliers, improving the quality of predictions by tuning fit parameters, interference between old and new data, implementing locally weighted learning efficiently, and applications of locally weighted learning. A companion paper surveys how locally weighted learning can be used in robot learning and control.

Cyberguide: a mobile context-aware tour guide

Future computing environments will free the user from the constraints of the desktop. Applications for a mobile environment should take advantage of contextual information, such as position, to offer greater services to the user. In this paper, we present the Cyberguide project, in which we are building prototypes of a mobile context‐aware tour guide. Knowledge of the users current location, as well as a history of past locations, are used to provide more of the kind of services that we come to expect from a real tour guide. We describe the architecture and features of a variety of Cyberguide prototypes developed for indoor and outdoor use on a number of different hand‐held platforms. We also discuss the general research issues that have emerged in our context‐aware applications development in a mobile environment.

The Aware Home: A Living Laboratory for Ubiquitous Computing Research

We are building a home, called the Aware Home, to create a living laboratory for research in ubiquitous computing for everyday activities. This paper introduces the Aware Home project and outlines some of our technology-and human-centered research objectives in creating the Aware Home.

Prioritized Sweeping: Reinforcement Learning with Less Data and Less Time

We present a new algorithm, prioritized sweeping, for efficient prediction and control of stochastic Markov systems. Incremental learning methods such as temporal differencing and Q-learning have real-time performance. Classical methods are slower, but more accurate, because they make full use of the observations. Prioritized sweeping aims for the best of both worlds. It uses all previous experiences both to prioritize important dynamic programming sweeps and to guide the exploration of state-space. We compare prioritized sweeping with other reinforcement learning schemes for a number of different stochastic optimal control problems. It successfully solves large state-space real-time problems with which other methods have difficulty.

Constructive incremental learning from only local information

We introduce a constructive, incremental learning system for regression problems that models data by means of spatially localized linear models. In contrast to other approaches, the size and shape of the receptive field of each locally linear model, as well as the parameters of the locally linear model itself, are learned independently, that is, without the need for competition or any other kind of communication. Independent learning is accomplished by incrementally minimizing a weighted local cross-validation error. As a result, we obtain a learning system that can allocate resources as needed while dealing with the bias-variance dilemma in a principled way. The spatial localization of the linear models increases robustness toward negative interference. Our learning system can be interpreted as a nonparametric adaptive bandwidth smoother, as a mixture of experts where the experts are trained in isolation, and as a learning system that profits from combining independent expert knowledge on the same problem. This article illustrates the potential learning capabilities of purely local learning and offers an interesting and powerful approach to learning with receptive fields.

Locally Weighted Learning for Control

Lazy learning methods provide useful representations and training algorithms for learning about complex phenomena during autonomous adaptive control of complex systems. This paper surveys ways in which locally weighted learning, a type of lazy learning, has been applied by us to control tasks. We explain various forms that control tasks can take, and how this affects the choice of learning paradigm. The discussion section explores the interesting impact that explicitly remembering all previous experiences has on the problem of learning to control.

Rapid prototyping of mobile context-aware applications: the Cyberguide case study

We present the Cyberguide project, in which we are building prototypes of a mobile context-aware tour guide that provide information t,o a tourist based on knowledge of position and orientation. We describe features of existing Cyberguide prototypes and discuss research issues that have emerged in our context-aware applications development in a mobile environment.

Estimation of inertial parameters of manipulator loads and links

The inertial parameters of manipulator rigid-body loads and links have been automatically estimated as a result of gen eral movement. The Newton-Euler equations have been recast to relate linearly the measured joint forces or torques via acceleration-dependent coefficients to the inertial parame ters, which have then been estimated by least squares. Load estimation was implemented on a PUMA 600 robot equipped with an R TI FS-B wrist force-torque sensor and on the MIT Serial Link Direct Drive Arm equipped with a Barry Wright Company Astek wrist force-torque sensor. Good estimates were obtained for load mass and center of mass, and the forces and torques due to movement of the load could be pre dicted accurately. The load moments of inertia were more difficult to estimate. Link estimation was implemented on the MIT Serial Link Direct Drive Arm. A good match was ob tained between joint torques predicted from the estimated parameters and the joint torques estimated from motor cur rents. The match actually proved superior to predicted torques based on link inertial parameters derived by CAD modeling. Restrictions on the identifiability of link inertial parameters due to restricted sensing and movement near the base have been addressed. Implications of estimation accu racy for manipulator dynamics and control have been consid ered.

Predicting human interruptibility with sensors

A person seeking another persons attention is normally able to quickly assess how interruptible the other person currently is. Such assessments allow behavior that we consider natural, socially appropriate, or simply polite. This is in sharp contrast to current computer and communication systems, which are largely unaware of the social situations surrounding their usage and the impact that their actions have on these situations. If systems could model human interruptibility, they could use this information to negotiate interruptions at appropriate times, thus improving human computer interaction.This article presents a series of studies that quantitatively demonstrate that simple sensors can support the construction of models that estimate human interruptibility as well as people do. These models can be constructed without using complex sensors, such as vision-based techniques, and therefore their use in everyday office environments is both practical and affordable. Although currently based on a demographically limited sample, our results indicate a substantial opportunity for future research to validate these results over larger groups of office workers. Our results also motivate the development of systems that use these models to negotiate interruptions at socially appropriate times.

Simultaneous tracking and activity recognition (STAR) using many anonymous, binary sensors

In this paper we introduce the simultaneous tracking and activity recognition (STAR) problem, which exploits the synergy between location and activity to provide the information necessary for automatic health monitoring. Automatic health monitoring can potentially help the elderly population live safely and independently in their own homes by providing key information to caregivers. Our goal is to perform accurate tracking and activity recognition for multiple people in a home environment. We use a “bottom-up” approach that primarily uses information gathered by many minimally invasive sensors commonly found in home security systems. We describe a Rao-Blackwellised particle filter for room-level tracking, rudimentary activity recognition (i.e., whether or not an occupant is moving), and data association. We evaluate our approach with experiments in a simulated environment and in a real instrumented home.