We present an architecture for robot control which can be viewed as a very fine-grained layered architecture motivated by the principles underlying the subsumption architecture. The subsumption architecture provides a powerful means for defining intelligent robot control mechanisms through the layered composition of simple behaviors. However, we have found that there are basic limitations inherent in this architecture due to the inaccessibility of information internal to a behavior. While adhering to the basic concept of building a robot control system through successive layers of competence, task achieving behaviors in our system are fragmented into many smaller decision-making units. Each of these units simply has the task of transforming a set of input activations into an output activation, so that the role that any unit plays in the system is defined entirely by how it is connected to other units. This fine-grained nature of our architecture permits a more flexible arbitration of c...

Most autonomous outdoor navigation systems tested on actual robots have centered on local navigation tasks such as avoiding obstacles or following roads. Global navigation has been limited to simple wandering, path tracking, straight-line goal seeking behaviors, or executing a sequence of scripted local behaviors. These capabilities are insufficient for unstructured and unknown environments, where replanning may be needed to account for new information discovered in every sensor image. To address these problems, we have developed a complete system that integrates local and global navigation. The local system uses a scanning laser rangefinder to detect obstacles and recommend steering commands to ensure robot safety. These obstacles are passed to the global system which stores them in a map of the environment. With each addition to the map, the global system uses an incremental path planning algorithm to optimally replan the global path and recommend steering commands to reach the goal. An arbiter combines the steering recommendations to achieve the proper balance between safety and goal acquisition. This system was tested on a real robot and successfully drove it 1.4 kilometers to find a goal given no a priori map of the environment. Autonomous outdoor navigators have a number of uses, including planetary exploration, construction site work, mining, military reconnaissance, and hazardous waste remediation. These tasks require a mobile robot to move between

. Consider a robot that has to travel from a start location s to a target t in an environment with opaque obstacles that lie in its way. The robot always knows its current absolute position and that of the target. It does not, however, know the positions and extents of the obstacles in advance; rather, it finds out about obstacles as it encounters them. We compare the distance walked by the robot in going from s to t to the length of the shortest (obstacle-free) path between s and t in the scene. We describe and analyze robot strategies that minimize this ratio for different kinds of scenes. In particular, we consider the cases of rectangular obstacles aligned with the axes, rectangular obstacles in more general orientations, and wider classes of convex bodies both in two and three dimensions. For many of these situations, our algorithms are optimal up to constant factors. We study scenes with non-convex obstacles, which are related to the study of maze-traversal. We also show scenes ...

We present a set of architectural concepts which address the needs for integrating high-level planning activities with lower-level reactive or participatory behaviors. Based on lessons learned from our experience with a hierarchical architecture for autonomous crosscountry navigation, we have come to recognize various pitfalls that may arise from the misuse of abstraction. Consequently, we have adopted a new approach which emphasizes the minimization of information loss both within and between system layers. This change in perspective has allowed us to greatly enhance the overall capabilities and performance of our system. I. INTRODUCTION An autonomous mobile robot must be constantly involved in the processing of large amounts of sensory data in order to produce meaningful actions. The ability of a control architecture to support this immense processing task in a timely manner is significantly affected by the organization of information pathways within the architecture. Some architectu...

m-RI-m-95-33 submitted in partial fulfiumtnr of the reqimlmts for the degm of

Introduction 1 The evidence grid representation was formulated at the CMU Mobile Robot Laboratory in 1983 to turn wide angle range measurements from cheap mobile robot-mounted sonar sensors into detailed spatial maps. It accumulates diffuse evidence about the occupancy of a grid of small volumes of nearby space from individual sensor readings into increasingly confident and detailed maps of a robot’s surroundings.

Our experience over the years with different architectures and planning systems for mobile robots has led us to a distributed approach where an arbiter receives votes for and against commands from each subsystem and decides upon the course of action which best satisfies the current goals and constraints of the system. Centralized arbitration of votes from distributed, independent decision-making processes provides coherent, rational, goal-directed behavior while preserving real-time responsiveness to its immediate physical environment. The Distributed Architecture for Mobile Navigation (DAMN) has been successfully used to integrate various independently developed subsystems, providing systems that perform road following, cross-country navigation, or teleoperation while avoiding obstacles and meeting mission objectives. Examples of implemented systems are given. Further research will seek to more rigorously define the behavior of the system. Keywords: mobile robots, architecture, behav...

Autonomous navigation in vegetation is challenging because the vegetation often hides the load-bearing surface which is used for evaluating the safety of potential actions. It is difficult to design rules for finding the true ground height in vegetation from forward looking sensor data, so we use an online adaptive method to automatically learn this mapping through experience with the world. This approach has been implemented on an autonomous tractor and has been tested in a farm setting. We describe the system and provide examples of finding obstacles and improving roll predictions in the presence of vegetation. We also show that the system can adapt to new vegetation conditions.

This paper describes an approach for using several levels of data fusion in the domain of autonomous off-road navigation. We are focusing on outdoor obstacle detection, and we present techniques that leverage on data fusion and machine learning for increasing the reliability of obstacle detection systems. We are combining color and infrared (IR) imagery with range information from a laser range finder. We show that in addition to fusing data at the pixel level, performing high level classifier fusion is beneficial in our domain. Our general approach is to use machine learning techniques for automatically deriving effective models of the classes of interest (obstacle and nonobstacle for example). We train classifiers on different subsets of the features we extract from our sensor suite and show how different classifier fusion schemes can be applied for obtaining a multiple classifier system that is more robust than any of the classifiers presented as input. We present experimental results we obtained on data collected with both the eXperimental Unmanned Vehicle (XUV) and a CMU developed robotic vehicle.

Current methods for off-road navigation using vehicle and terrain models to predict future vehicle response are limited by the accuracy of the models they use and can suffer if the world is unknown or if conditions change and the models become inaccurate. In this paper, an adaptive approach is presented that closes the loop around the vehicle predictions. This approach is applied to an autonomous vehicle driving through unknown terrain with varied vegetation. Features are extracted from range points from forward looking sensors. These features are used by a locally weighted learning module to predict the load-bearing surface, which is often hidden by vegetation. The true surface is then found when the vehicle drives over that area, and this feedback is used to improve the model. Results using real data show improved predictions of the load-bearing surface and successful adaptation to changing conditions.