Rough terrain autonomous navigation continues to pose a challenge to the robotics community. Robust navigation by a mobile robot depends not only on the individual performance of perception and planning systems, but on how well these systems are coupled. When traversing complex unstructured terrain, this coupling (in the form of a cost function) has a large impact on robot behavior and performance, necessitating a robust design. This paper explores the application of Learning from Demonstration to this task for the Crusher autonomous navigation platform. Using expert examples of desired navigation behavior, mappings from both online and offline perceptual data to planning costs are learned. Challenges in adapting existing techniques to complex online planning systems and imperfect demonstration are addressed, along with additional practical considerations. The benefits to autonomous performance of this approach are examined, as well as the decrease in necessary designer effort. Experimental results are presented from autonomous traverses through complex natural environments. 1

With the aim of developing a vehicle that can fully autonomously operate in highly vegetated terrain, the University of Kaiserslautern’s Robotics Research Lab is conducting research in the field of off-road robotics. Their platform RAVON is a 4WD vehicle equipped with a large number of different sensor systems, which are used as basis for the hazard detection. RAVON’s navigation system is three-layered, consisting of a deliberative navigator on top of a behaviour-based pilot. The navigator’s job is to create and update topological maps of the robot’s environment, in which paths are calculated. The pilot tries to move the robot along such paths while keeping it away from obstacles. Collision avoidance is realised by a large number of behaviours that access the sensor data in a unified and straightforward way. An intermediate layer shall mediate between navigator and pilot and solve problems they cannot address properly. The purpose of this paper is to provide information about the platform RAVON and to present the concepts underlying the work in detail.

Abstract — This paper addresses the problem of vegetation detection from laser measurements. The ability to detect vegetation is important for robots operating outdoors, since it enables a robot to navigate more efficiently and safely in such environments. In this paper, we propose a novel approach for detecting low, grass-like vegetation using laser remission values. In our algorithm, the laser remission is modeled as a function of distance, incidence angle, and material. We classify surface terrain based on 3D scans of the surroundings of the robot. The model is learned in a self-supervised way using vibrationbased terrain classification. In all real world experiments we carried out, our approach yields a classification accuracy of over 99%. We furthermore illustrate how the learned classifier can improve the autonomous navigation capabilities of mobile robots. I.

Novelty detection is often treated as a one-class classification problem: how to segment a data set of examples from everything else that would be considered novel or abnormal. Almost all existing novelty detection techniques, however, suffer from diminished performance when the number of less relevant, redundant or noisy features increases, as often the case with high-dimensional feature spaces. Additionally, many of these algorithms are not suited for online use, a trait that is highly desirable for many robotic applications. We present a novelty detection algorithm that is able to address this sensitivity to high feature dimensionality by utilizing prior class information within the training set. Additionally, our anytime algorithm is well suited for online use when a constantly adjusting environmental model is beneficial. We apply this algorithm to online detection of novel perception system input on an outdoor mobile robot and argue how such abilities could be key in increasing the real-world applications and impact of mobile robotics 1. 1 Most figures in this paper are best viewed in color.

As mobile robotics continues to advance, we are beginning to see intelligent robots with complex perception and planning systems onboard. These systems are often engineered for a specific task, or trained using machine learning to be adaptable and robust. Unfortunately, when faced with the complexities of the real world, almost every current robotic system will eventually encounter a situation which it has not been not trained or designed to handle. Unless a system operates in a highly structured or controlled environment, it will be impossible to determine all of the types of obstacles a robot may encounter. Instead of trying to make robots perfect perception and planning machines, we seek to enable robotic systems to detect situations in which the robot is unfamiliar. When a robot regularly visits an area more than once, we propose the use of a change detection algorithm to identify significant changes in that area over time and inform the robot about possible hazards. If a robot has traversed an area before, we can generally assume it to be safe, but if an unexpected change happens in the robots environment, this may represent a danger to the robot or the safety of humans

Abstract — On-line, efficient perception based on sensing is essential for an autonomous robot to operate in an unknown and unpredictable environment. An efficient on-line algorithm is introduced to determine whether a robot at a future time t and a configuration q will be guaranteed collision-free, directly from real-world sensor data of the robot’s environment at the current time τ, using stereo vision sensor. Such a problem can be formulated [1] as checking the intersection between the so-called dynamic envelope, which relates to the robot at a configuration-time (CT) point (q, t) and the current sensing time τ, and the atomic obstacles, which are obtained directly from low-level sensory data at τ. The algorithm achieves real-time efficiency, as confirmed by the experimental results, by classifying the atomic obstacles possibly intersecting the dynamic envelope and by grouping relevant atomic obstacles on the fly. It is suitable to be used on-line by sensing-based motion planners. I.

Many mobile robot applications require robots to act safely and intelligently in complex unfamiliar environments with little structure and limited or unavailable human supervision. As a robot is forced to operate in an environment that it was not engineered or trained for, various aspects of its performance will inevitably degrade. Roboticists equip robots with powerful sensors and data sources to deal with uncertainty, only to discover that the robots are able to make only minimal use of this data and still find themselves in trouble. Similarly, roboticists develop and train their robots in representative areas, only to discover that they encounter new situations that are not in their experience base. Small problems resulting in mildly sub-optimal performance are often tolerable, but major failures resulting in vehicle loss or compromised human safety are not. This thesis presents a series of online algorithms to enable a mobile robot to better deal with uncertainty in unfamiliar domains in order to improve its navigational abilities, better utilize available data and resources and reduce risk to the vehicle. We validate these algorithms through extensive testing onboard large mobile robot systems and argue how such approaches can increase the reliability and robustness of mobile robots, bringing them closer to the capabilities