To navigate reliably in indoor environments, a mobile robot must know where it is. Thus, reliable position estimation is key problem robotics. We believe that probabilistic approaches are among the most promising candidates to providing comprehensive and real-time solution localization problem. However, current methods still face considerable hurdles. In particular problems encountered closely related type of representation used represent probability densities over robot's state space....

We present a novel data structure, the Bayes tree, that provides an algorithmic foundation enabling better understanding of existing graphical model inference algorithms and their connection to sparse matrix factorization methods. Similar clique tree encodes factored probability density, but unlike it is directed maps more naturally square root information simultaneous localization mapping (SLAM) problem. In this paper, we highlight three insights provided by our new structure. First, in...

In this paper, we present incremental smoothing and mapping (iSAM), which is a novel approach to the simultaneous localization problem that based on fast matrix factorization. iSAM provides an efficient exact solution by updating QR factorization of naturally sparse information matrix, thereby recalculating only those entries actually change. even for robot trajectories with many loops as it avoids unnecessary fill-in in factor periodic variable reordering. Also, enable data association real...

Solving the SLAM (simultaneous localization and mapping) problem is one way to enable a robot explore, map, navigate in previously unknown environment. Smoothing approaches have been investigated as viable alternative extended Kalman filter (EKF)-based solutions problem. In particular, looked at that factorize either associated information matrix or measurement Jacobian into square root form. Such techniques several significant advantages over EKF: they are faster yet exact; can be used...

We describe a particle filter that effectively deals with interacting targets--targets are influenced by the proximity and/or behavior of other targets. The includes Markov random field (MRF) motion prior helps maintain identity targets throughout an interaction, significantly reducing tracker failures. show this MRF can be easily implemented including additional interaction factor in importance weights filter. However, computational requirements resulting multitarget render it unusable for...

This paper describes an interactive tour-guide robot, which was successfully exhibited in a Smithsonian museum. During its two weeks of operation, the robot interacted with thousands people, traversing more than 44 km at speeds up to 163 cm/sec. Our approach specifically addresses issues such as safe navigation unmodified and dynamic environments, short-term human-robot interaction. It uses learning pervasively all levels software architecture.

Current approaches for visual-inertial odometry (VIO) are able to attain highly accurate state estimation via nonlinear optimization. However, real-time optimization quickly becomes infeasible as the trajectory grows over time, this problem is further emphasized by fact that inertial measurements come at high rate, hence leading fast growth of number variables in In paper, we address issue preintegrating between selected keyframes into single relative motion constraints. Our first...

This paper describes Minerva, an interactive tour-guide robot that was successfully deployed in a Smithsonian museum. Minerva’s software is pervasively probabilistic, relying on explicit representations of uncertainty perception and control. During 2 weeks operation, the interacted with thousands people, both museum through Web, traversing more than 44 km at speeds up to 163 cm/sec unmodified

The paper explores several statistical pattern recognition techniques to classify utterances according their emotional content. authors have recorded a corpus containing speech with over 1000 from different speakers. They present new method of extracting prosodic features speech, based on smoothing spline approximation the pitch contour. To make maximal use limited amount training data available, they introduce novel technique: majority voting subspace specialists. Using this technique,...

Recent results in monocular visual-inertial navigation (VIN) have shown that optimization-based approaches outperform filtering methods terms of accuracy due to their capability relinearize past states.However, the improvement comes at cost increased computational complexity.In this paper, we address issue by preintegrating inertial measurements between selected keyframes.The preintegration allows us accurately summarize hundreds into a single relative motion constraint.Our first...

Pose graph optimization is the non-convex problem underlying pose-based Simultaneous Localization and Mapping (SLAM). If robot orientations were known, pose would be a linear least-squares problem, whose solution can computed efficiently reliably. Since rotations are actual reason why SLAM difficult in this work we survey techniques for 3D rotation estimation. Rotation estimation has rich history three scientific communities: robotics, computer vision, control theory. We review relevant...

We present Panoptic Neural Fields (PNF), an object-aware neural scene representation that decomposes a into set of objects (things) and background (stuff). Each object is represented by oriented 3D bounding box multi-layer perceptron (MLP) takes position, direction, time outputs density radiance. The stuff similar MLP additionally semantic labels. MLPs are instance-specific thus can be smaller faster than previous approaches, while still leveraging category-specific priors incorporated via...

A method is presented to recover 3D scene structure and camera motion from multiple images without the need for correspondence information. The problem framed as finding maximum likelihood given only 2D measurements, integrating over all possible assignments of features measurements. This goal achieved by means an algorithm which iteratively refines a probability distribution set assignments. At each iteration new solved, using input 'virtual measurements' derived this distribution. needed...

To navigate reliably in indoor environments, a mobile robot must know where it is. This includes both the ability of globally localizing from scratch, as well tracking robot's position once its location is known. Vision has long been advertised providing solution to these problems, but we still lack efficient solutions unmodified environments. Many existing approaches require modification environment function properly, and those that work within environments seldomly address problem global...

We present iSAM2, a fully incremental, graph-based version of incremental smoothing and mapping (iSAM). iSAM2 is based on novel graphical model-based interpretation sparse matrix factorization methods, afforded by the recently introduced Bayes tree data structure. The original iSAM algorithm incrementally maintains square root information applying updates. analyze updates as simple editing operations conditional densities represented its cliques. Based that insight, we new method to change...

Wearable computers can certainly support audio-only presentation of information; a visual interface need not be present for effective user interaction. A system wearable audio navigation (SWAN) is being developed to serve as and orientation aid persons temporarily or permanently visually impaired. SWAN computer consisting output tactile input via handheld interface. aids in safe pedestrian includes the ability author new GIS data relevant their needs wayfinding, obstacle avoidance,...