Modeling Interaction via the Principle of Maximum Causal Entropy
ICML runner up for best student paper by Brian Ziebart, J. Andrew Bagnell, and Anind Dey.

@inproceedings{bziebart-maxcausalent,
author = {Brian D. Ziebart and J. Andrew Bagnell
and Anind K. Dey},
title = {Modeling Interaction via the Principle of Maximum Causal Entropy},
year = {2010},
booktitle = {International Conference on Machine Learning}
}

The principle of maximum entropy provides a powerful framework for statistical models of joint, conditional, and marginal distributions. However, there are many important distributions with elements of interaction and feedback where its applicability has not been established. This work presents the principle of maximum causal entropy — an approach based on causally conditioned probabilities that can appropriately model the availability and influence of sequentially revealed side information. Using this principle, we derive Maximum Causal Entropy Influence Diagrams, a new probabilistic graphical framework for modeling decision making in settings with latent information, sequential interaction, and feedback. We describe the theoretical advantages of this model and demonstrate its applicability for statistically framing inverse optimal control and decision prediction tasks.

Online modeling of experience base for potential environmental hazards

Learning Rough-Terrain Autonomous Navigation, by J. A. Bagnell, D. M. Bradley, D. Silver, B. Sofman, A. Stentz. Currently under review for Robotics and Automation Magazine.

Autonomous navigation by a mobile robot through natural, unstructured terrain is one of the premier challenges in field robotics. The DARPA UPI program was tasked with advancing the state of the art in robust autonomous performance through challenging and widely varying environments. In order to accomplish this goal, machine learning techniques were heavily utilized to provide robust and adaptive performance, while simultaneously reducing the required development and deployment time. This paper describes the autonomous system, Crusher, developed for the UPI program, and the learning approaches that aided in its successful performance.

A short video of the UPI system’s performance leveraging the learning techniques discussed can be seen by clicking on the image to the left.

Online modeling of experience base for potential environmental hazards

Anytime Online Novelty Detection for Vehicle Safeguarding, by Boris Sofman, J. A. Bagnell, and Anthony Stentz. Currently under review for ICRA 2010.

We present an online novelty detection algorithm that allows a mobile robot to identify stimuli that are outside its experience base, avoiding potentially hazardous situations. This algorithm addresses many of the limitations of existing novelty detection approaches, including sensitivity to high-dimensional and noisy feature spaces and the inability to efficiently update their models online. Additionally, this algorithm has anytime properties that make it highly suitable for mobile robot use. The included appendix provides a proof that the run-time of our algorithm for maintaining previously seen examples is constant-competitive with respect to any other algorithm.

A short video of this system’s performance can be seen by clicking on the image to the left. The novelty model is initialized to be empty and as the environment is perceived by the robot, the model is adjusted online so that future similar stimuli are no longer novel. Novel examples are shown with a red shade.