High-mobility walking robots offer unique capabilities in complex off-road environments where wheeled vehicles are not able to travel. However, these environments can also pose significant autonomous navigation challenges. Key steps in planning a safe path for the robot autonomously include estimating the height of the support ground surface - which is often occluded by vegetation - and classifying the terrain and obstacles above the ground surface. This paper describes the development and experimental evaluation of a terrain classification and ground surface height estimation system to support autonomous navigation for a high-mobility walking robot. We provide experimental evaluation on an extensive, manually-labeled dataset collected from geographically diverse sites over a 28-month period.

Building robust and reliable autonomous navigation systems that generalize across environments and operating scenarios remains a core challenge in robotics. Machine learning has proven a significant aid in this task; in recent years learning from demonstration has become especially popular, leading to improved systems while requiring less expert tuning and interaction. However, these approaches still place a burden on the expert, specifically to choose the best demonstrations to provide. This work proposes two approaches for active learning from demonstration, in which the learning system requests specific demonstrations from the expert. The approaches identify examples for which expert demonstration is predicted to provide useful information on concepts which are either novel or uncertain to the current system. Experimental results demonstrate both improved generalization performance and reduced expert interaction when using these approaches.

GPS has become the de facto standard for obtaining a global position estimate during outdoor autonomous navigation. However, GPS can become degraded due to occlusion or interference, to the detriment of autonomous performance. In addition, GPS positions must be aligned with prior data, a tedious and continual process. This work presents a solution to these two problems based on learning generic observation models in the presence of GPS to use in its absence. The models are non-parametric and compared to traditional approaches require few assumptions about either the prior data available or a robot's onboard sensors. Along with allowing for localization to prior data under GPS-denied conditions, this learning approach can be coupled with an EM procedure to automatically register GPS and prior data positions. Experimental results are presented based on data from more than 15 km of autonomous navigation through challenging outdoor terrain.

Long range navigation by unmanned ground vehicles continues to challenge the robotics community. Efficient navigation requires not only intelligent on-board perception and planning systems, but also the effective use of prior knowledge of the vehicle's environment. This paper describes a system for supporting unmanned ground vehicle navigation through the use of heterogeneous overhead data. Semantic information is obtained through supervised classification, and vehicle mobility is predicted from available geometric data. This approach is demonstrated and validated through over 50 kilometers of autonomous traversal through complex natural environments

The need for reliable maps of subterranean spaces too hazardous for humans to occupy has motivated the use of robotic technology as mapping tools. As such, we present a systemic approach to autonomous topological exploration of a mine environment to facilitate the process of mapping. This approach focuses upon the interaction of three high-level processes: topological planning, intersection identification and local navigation. Topological planning tasks the robot to investigate stretches of mine corridor for the purpose of collecting data. Intersection identification converts sensory input into topological components used to construct an online topological map and provide the robot with a global sense of position. Local navigation transforms topological exploration objectives into robot actuation enabling traversal of mine corridors. These processes are described in detail with results presented from experiments conducted at a research coal mine near Pittsburgh, PA.

This paper presents a new map specifically designed for robots operating in large environments and possibly in higher dimensions. We call this map the hierarchical atlas because it is a multilevel and multiresolution representation. For this paper, the hierarchical atlas has two levels: at the highest level there is a topological map that organizes the free space into submaps at the lower level. The lower-level submaps are simply a collection of features. The hierarchical atlas allows us to perform calculations and run estimation techniques, such as Kalman filtering, in local areas without having to correlate and associate data for the entire map. This provides a means to explore and map large environments in the presence of uncertainty with a process named hierarchical simultaneous localization and mapping. As well as organizing information of the free space, the map also induces well-defined sensor-based control laws and a provably complete policy to explore unknown regions. The resulting map is also useful for other tasks such as navigation, obstacle avoidance, and global localization. Experimental results are presented showing successful map building and subsequent use of the map in large-scale spaces.

In this paper we present a new technique for improving the azimuth resolution of ultrasonic range sensors frequently used with mobile robots. This improvement is achieved without a significant increase in the latency, or processing delay, of the system. Our approach decreases the azimuth uncertainty of a sensor reading by eliminating portions of the reading that are contradicted by subsequent readings. Our idea bears resemblance to space carving as used by the vision community, where a ray of light is used to define the boundaries of an obstacle. A sonar model similar to that commonly utilized by occupancy grids is used. Our method, termed arc carving, can be used to produce maps that are both accurate and with low enough latency for robust mobile robot navigation. Experimental results verify this approach over spaces as large as 5000 square meters.

We present a robust method for detecting and recognizing topological features in underground mines. Our method involves performing Delaunay triangulations on range scans to extract points of interest, such as intersecting corridors. By combining these interest points into a topological map, we have a valuable tool for navigation and localization in large scale, highly cyclic environments. We present results from a research coal mine near Pittsburgh, PA.

This paper presents a novel method of combining topological and feature-based mapping strategies to create a hierarchical approach to simultaneous localization and mapping (SLAM). More than simply running both processes in parallel, we use the topological mapping procedure to organize local feature-based methods. The result is an autonomous exploration and mapping strategy that scales well to large environments and higher dimensions while confronting the issue of obstacle avoidance. We have obtained successful results of our approach in an area spanning 5000 square meters.