Existing approaches to reward inference from behavior typically assume that humans provide demonstrations according to specific models of behavior. However, humans often indicate their goals through a wide range of behaviors, from actions that are suboptimal due to poor planning or execution to behaviors which are intended to communicate goals rather than achieve them. We propose that supervised learning offers a unified framework to infer reward functions from any class of behavior, and show that such an approach is asymptotically Bayes-optimal under mild assumptions. Experiments on simulated robotic manipulation tasks show that our method can efficiently infer rewards from a wide variety of arbitrarily suboptimal demonstrations.

We investigate the Active Clustering Problem (ACP). A learner interacts with an $N$-armed stochastic bandit with $d$-dimensional subGaussian feedback. There exists a hidden partition of the arms into $K$ groups, such that arms within the same group, share the same mean vector. The learner's task is to uncover this hidden partition with the smallest budget - i.e., the least number of observation - and with a probability of error smaller than a prescribed constant $\delta$. In this paper, (i) we derive a non-asymptotic lower bound for the budget, and (ii) we introduce the computationally efficient ACB algorithm, whose budget matches the lower bound in most regimes. We improve on the performance of a uniform sampling strategy. Importantly, contrary to the batch setting, we establish that there is no computation-information gap in the active setting.

Unconditional scene inference and generation are challenging to learn jointly with a single compositional model. Despite encouraging progress on models that extract object-centric representations (''slots'') from images, unconditional generation of scenes from slots has received less attention. This is primarily because learning the multi-object relations necessary to imagine coherent scenes is difficult. We hypothesize that most existing slot-based models have a limited ability to learn object correlations. We propose two improvements that strengthen object correlation learning. The first is to condition the slots on a global, scene-level variable that captures higher-order correlations between slots. Second, we address the fundamental lack of a canonical order for objects in images by proposing to learn a consistent order to use for the autoregressive generation of scene objects. Specifically, we train an autoregressive slot prior to sequentially generate scene objects following a learned order. Ordered slot inference entails first estimating a randomly ordered set of slots using existing approaches for extracting slots from images, then aligning those slots to ordered slots generated autoregressively with the slot prior. Our experiments across three multi-object environments demonstrate clear gains in unconditional scene generation quality. Detailed ablation studies are also provided that validate the two proposed improvements.

In unstructured outdoor environments, robotics requires accurate and efficient odometry with low computational time. Existing low-bias LiDAR odometry methods are often computationally expensive. To address this problem, we present a lightweight LiDAR odometry method that converts unorganized point cloud data into a spherical range image (SRI) and filters out surface, edge, and ground features in the image plane. This substantially reduces computation time and the required features for odometry estimation in LOAM-based algorithms. Our odometry estimation method does not rely on global maps or loop closure algorithms, which further reduces computational costs. Experimental results generate a translation and rotation error of 0.86\% and 0.0036{\deg}/m on the KITTI dataset with an average runtime of 78ms. In addition, we tested the method with our data, obtaining an average closed-loop error of 0.8m and a runtime of 27ms over eight loops covering 3.5Km.

Radio frequency (RF)-based techniques are widely adopted for indoor localization despite the challenges in extracting sufficient information from measurements. Soft range information (SRI) offers a promising alternative for highly accurate localization that gives all probable range values rather than a single estimate of distance. We propose a deep learning approach to generate accurate SRI from RF measurements. In particular, the proposed approach is implemented by a network with two neural modules and conducts the generation directly from raw data. Extensive experiments on a case study with two public datasets are conducted to quantify the efficiency in different indoor localization tasks. The results show that the proposed approach can generate highly accurate SRI, and significantly outperforms conventional techniques in both non-line-of-sight (NLOS) detection and ranging error mitigation.

SRI International, an over 75-year strong pioneering research institute, has a rich history supporting government and industry. Our innovations have created new industries, billions of dollars in market value and lasting benefits to society. SRI is organized around broad disciplines and capabilities, with research and development divisions and labs to groups who excel at identifying new opportunities, developing products and creating custom solutions. Our organization is driven by impact – delivering unique solutions for the world's important challenges and transforming ideas into reality for clients and partners. From image capture to situational understanding, SRI's Center for Vision Technologies (CVT) offers end-to-end vision solutions that translate into real-world applications.

Global access to art, culture, and entertainment products – music, movies, books, and more – has undergone fundamental changes over the past 20 years in light of groundbreaking developments in artificial intelligence. For example, users of streaming services like Netflix and Spotify have data collected and analyzed by algorithms to determine their streaming habits – resulting in recommendations that cater to their tastes. But this is only one of the many ways in which AI tools are transforming the arts and culture industries. AI is also being used in the production of music and other art, with algorithms generating photos or writing songs on their own. Warner Music even "signed" an algorithm to a record deal in 2019.

A driver of a car can be in many different emotional states while driving: calm, anxious, angry, excited, relaxed or drowsy are just a few. SRI International researchers are teaching vehicles how to read and respond to a driver's emotional state as part of a project with Toyota Motor Corporation. This project seeks to improve driving safety and personalize the interaction between a person and their car. These efforts center on integrating emotional artificial intelligence (AI) into cars. SRI scientists are making strides in this with the development of the Driver Monitoring System (DMS).

A "chemputer" is a robotic method of producing drug molecules that uses downloadable blueprints to synthesize organic chemicals via programming. Originated in the University of Glasgow lab of chemist Lee Cronin, the method has produced several blueprints available on the GitHub software repository, including blueprints for Remdesivir, the FDA-approved drug for antiviral treatment of COVID-19. Cronin, who designed the "bird's nest" of tubing, pumps, and flasks that make up the chemputer, spent years thinking of a way researchers could distribute and produce molecules as easily as they email and print PDFs, according to a recent account from CNBC. "If we have a standard way of discovering molecules, making molecules, and then manufacturing them, suddenly nothing goes out of print," Cronin stated. Beyond creating the chemputer, Cronin's team recently took a second major step towards digitizing chemistry with an accessible way to program the machine.

The Stable Roommates problem (SR) is characterized by the preferences of agents over other agents as roommates: each agent ranks all others in strict order of preference. A solution to SR is then a partition of the agents into pairs so that each pair shares a room, and there is no pair of agents that would block this matching (i.e., who prefers the other to their roommate in the matching). There are interesting variations of SR that are motivated by applications (e.g., the preference lists may be incomplete (SRI) and involve ties (SRTI)), and that try to find a more fair solution (e.g., Egalitarian SR). Unlike the Stable Marriage problem, every SR instance is not guaranteed to have a solution. For that reason, there are also variations of SR that try to find a good-enough solution (e.g., Almost SR). Most of these variations are NP-hard. We introduce a formal framework, called SRTI-ASP, utilizing the logic programming paradigm Answer Set Programming, that is provable and general enough to solve many of such variations of SR. Our empirical analysis shows that SRTI-ASP is also promising for applications. This paper is under consideration for acceptance in TPLP.