Search

The problem with the Model-based Mobile robot Language(MML) processor is that the code is unstructured, causing the system to be unstable; it is very difficult to read because of deficient source code documentation; and because of poorly defined function interfaces and extensive functional coupling, the system is hard to maintain. To fix the MML processor, we performed a manual static analysis of the existing source code to understand its structure. Next, based on the analysis, the software system was restructured and the functionality enhanced. Finally, explicit source code documentation was added in the form of comments. There are several results with the new system. First, global variables are reduced from 152 to zero. Secondly, function interfaces are clearly defined and function coupling is enhanced. Finally, the source code is extensively documented. Following from these results, the new system is more stable, easier to read and understand, and sampler to modify, http://archive.org/details/constructingreal00hugg, NA, U.S. Army (USA) author

The recognition of underwater objects and obstacles by sonar has been explored in many forms, particularly through the use of high-resolution imaging sonar systems. This work explores a method of providing real-time obstacle avoidance and navigational position updating for an Autonomous Underwater Vehicle (AUV) by applying regression analysis and geometric interpretation to sonar range data obtained from a low-cost, lowresolution, fixed-beam sonar. The algorithm utilized by this method first develops a leastsquares fit for sonar range data in a 2-D manner. The parameters developed by this method are then compared to an environmental model for position identification. If no match is achieved, then by applying the known geometry of the acoustic signal, an estimate for a 3- D surface is derived. This derived 3-D surface is then added to the environmental model to enable accurate path planning and post-mission analysis information. This method is currently implemented on an operational AUV operating in a well-defined orthogonal environment at NPS. The paper also discusses the simulation of the sonar systems using a ray tracing technique in a real-time dynamic graphical simulation implemented on a Silicon Graphics IRIS workstation., http://archive.org/details/designimplementa00floy, Commander, United States Navy, Approved for public release; distribution is unlimited.

There are two goals for autonomous vehicle navigation planning: shortest path and safe path. These goals are often in conflict; path safety is more important. Safety of the autonomous vehicle's navigation is determined by the clearances between the vehicle and obstacles. Because a Voronoi boundary is the set of points locally maximizing the clearance from obstacles, safety is maximized on it. Therefore Voronoi Diagrams are suitable for motion planning of autonomous vehicles. We use the derivative of curvature k of the vehicle motion (dk/ds) as the only control variable for the vehicle where s is the length along the vehicle trajectory. Previous motion planning of the autonomous mobile robot Yamabico-11 at Naval Postgraduate School used a path tracking method. Before the mission began the vehicle was given a track to follow; motion planning consisted of calculating the point on the track closest to the vehicle and calculating dk/ ds then steering the vehicle to get onto track. We propose a method of planning safe motions of the vehicle to calculate optimal dk/ds at each point directly from the information of the world without calculating the track to follow. This safe navigation algorithm is fundamentally different from the path tracking using a path specification. Additionally motion planning is simpler and faster than the path tracking method. The effectiveness of this steering function for vehicle motion control is demonstrated by algorithmic simulation and by use on the autonomous mobile robot Yamabico 11 at the Naval Postgraduate School, http://archive.org/details/optimizingsafemo1094528432, Lieutenant Junior Grade, Japan Maritime Self-Defense Force, Approved for public release; distribution is unlimited.

The Computer Science Department at the Naval Postgraduate School in Monterey, California performs research on the control and operation of autonomous mobile robots. One such robot, Yamabico-11, is an excellent test platform for the study of path planning and obstacle avoidance. The ability to operate in an area where unforeseen obstacles are present, and still attain the specified goal, is a highly desirable behavior in an autonomous mobile robot. This thesis takes a step in that direction by proposing and implementing an expert system for high level motion control of the robot. The expert system combines basic path planning routines and advanced obstacle avoidance techniques to direct the robot as it performs the mission, http://archive.org/details/anexpertsystemfo1094527046, Lieutenant Commander, United States Navy, Approved for public release; distribution is unlimited.

This research addresses the problems experienced by the autonomous mobile robot, Yamabico- 11, with its ultrasonic sonar system. It explains the basics of acoustic theory as related to Yamabico- 11 and explains the sources of limitations imposed on Yamabico- 11 by the physical nature of the problem. This paper documents the basic characteristics of the sonar hardware and examines causes of sonar range errors. Finally, this research leads to improvements of the current sonar system to provide better directional coverage through a new sonar configuration, http://archive.org/details/analysisndimprov1094527943, Lieutenant Commander, United States Navy, Approved for public release; distribution is unlimited.

In order to function autonomously in the real world a mobile robot must first be able to sense the boundaries of it's operating space. Once the enclosing features and/or obstacles have been sensed they must be interpreted and represented in some way meaningful to the robot's controlling algorithms. The objective of this work is the development of a system of ultrasonic sensors, or sonars, for the mobile robot YAMABICO-11 at the Naval Postgraduate School, and the implementation of a user friendly set of sonar language functions for the robot's control language MML. The sonar hardware includes twelve transducer pairs, their drivers and a bus mounted control card. The sonar control system operates autonomously under direction of the robot's central processor.Extraction of linear features is accomplished by the use of a least-square-fit algorithm of cartesian coordinate pairs to a parametric representation of the including line segment., http://archive.org/details/amobilerobotsona1094526635, Lieutenant, U.S. Navy, Approved for public release; distribution is unlimited.

The major problem addressed by this research is how to improve an existing real-time software system's readability, maintainability, stability and portability using reengineering techniques. A fundamental portion of the Model- based Mobile robot Language (MML) was the real-time system chosen as the basis for this study. The approach taken was to create a new system design. The new design was based on system specifications obtained by conducting static and dynamic analysis on the existing system. The results are that a new core system was implemented using a design that focused on creating independent software sub-systems while encapsulating data. Hardware dependencies were localized and assembly code minimized. The new system is easier to understand and modify and is portable to other hardware platforms. Autonomous vehicle, Robot, Software engineering, Real-time system, http://archive.org/details/improvingsoftwar1094528279, Lieutenant, United States Navy, Approved for public release; distribution is unlimited.

In order to continue to improve the usefulness of robots, it is becoming increasingly important to have them act as autonomous agents. A significant step toward this object is autonomous motion planning. This research was conducted as part of a broader effort to empower Yambico-11, a mobile robot under development at the Naval Postgraduate School, with ability to move autonomously. We believe that this problem is best attacked in layers. This thesis is our proposal for the initial layer. Given a robot's current location and its goal location, we use the homotopy relation to reduce the infinite set of path choices into a more manageable and smaller set of path classes. Specifically, we solve the problem of how to enable a robot to autonomously identify and label these classes of paths. We begin by decomposing the robot's operating environment into a collection of convex pieces called cells. The cells are transformed into a graph by adjacency. We show that every simple path on the graph corresponds to a unique simple homotopy class in the robot's world. We then search the graph to give each class a symbolic representation which also provides intermediate path planning clues. Subsequent layers can use these clues to form a more detailed plan. We implement the cell decomposition, graph transformation, and path class labeling as C programs, and preprocess them on a Unix workstation. This resulting data structures are then compiled and linked into the robot's kernel. All implementation has been integrated into the model- based mobile robot language (mml) used by Yamabico-11, http://archive.org/details/layeredpathplann1094528329, NA, NA, U.S. Army (USA) author

http://archive.org/details/threedimensional1094528610, NA, NA

The problem discussed in this dissertation is the development of an efficient method for visual navigation of autonomous vehicles. The approach is to significantly reduce the expensive computational time of landmark detection by straight-edge features. A novel, fast straight-edge-detection method for use in autonomous vehicle navigation and other image-understanding applications is presented. Straight edges in gray-scale images are detected using a new direction-control1ed edge tracking method, which gives precise estimate of the endpoints. To significantly reduce the number of exhaustive pixel computations, a random-hitting method using a pseudo-random number generator is proposed. Only if a generated pixel is significant do we start tracking the edge containing that pixel. To overcome the "noisy" gradient direction information, a robust least-squares linear fitting method is used to control the tracking process. The results of the algorithm show how it is efficient for landmark detection, which is important for motion control of autonomous vehicles. Thus the new method is implemented as a component of the image-understanding system in the autonomous mobile robot Yamabico-11 at the Naval Postgraduate School. An efficient world- modeling method based on the 2D model of the environment of the vehicle, including the heights of vertical edges in the environment, is presented. This modeling method is implemented with the new edge-detection method to improve the efficiency of the pose-determination algorithm (pose is a combination of the position and orientation of the camera), which is an essential task in the area of autonomous vehicle navigation, http://archive.org/details/anefficientmodel109458299, Major, Egyptian Air Force, Approved for public release; distribution is unlimited.

One of the most difficult theoretical problems in robotics--motion planning for rigid body robots-- must be solved before a robot can perform real- world tasks such as mine searching and processing. This dissertation proposes a new motion planning algorithm for an autonomous robot, as well as the method and results of implementing this algorithm on a real vehicle. This dissertation addresses the problem of safely navigating an autonomous vehicle through free space of a two dimensional, world model with polygonal obstacles from a start configuration (position orientation) to a goal configuration using smooth motion under the structure of a layered motion planning approach. The approach proposes several new concepts, including v-edges and directed v-edges, and divides the motion planning problem of a rigid body vehicle into two subproblems: (1) finding a global path using Voronoi diagrams and for a given start and goal configurations planning an optimal global path; the planned path is a sequence of directed v-edges, (2) planning a local motion from the start configuration, using the aforementioned global path. The problem of how to design a safe and smooth path, is effectively solved by the steering function method and proximity. Another problem addressed here is how to make a smooth transition when the vehicle gets closer to an intersection of two distinct boundaries. This dissertation also presents a robust algorithm for the vehicle to continually eliminate its positional uncertainty while executing missions. This functionality is called self-localization which is an essential component of model-based navigation for indoor applications. This algorithm is based on the two-dimensional transformation group. Through this method, the robot can minimize its positional uncertainty, make safe and reliable motions, and perform useful tasks in a partially known world, http://archive.org/details/newmotionplannin109458737, Colonel, Egyptian Army, Approved for public release; distribution is unlimited.

Millions of acres of the U.S. formerly used defense sites (FUDS) are contaminated with unexploded ordnance (UXO) as a result of past military use. The process of returning the land to the civilian sector is sensitive, intensive, and costly (e.g., millions of dollars, and the loss of human life). Hence'clearing (i.e., site remediation, range clearance, and explosive ordnance disposal) UXO's from FUDS is a complex problem. Existing clearing methods are inaccurate, dangerous, and labor intensive. This thesis shows that through robotics technology (e.g., Shepherd rotary vehicle with three degrees of freedom) and the use of advanced computer technology it is possible to make clearing tasks safer, more cost effective, and more efficient. An over arching hardware and software architecture was developed for Shepherd (including a self- contained on board computer system). The software system was developed for timer control, motion control, user interface, and an operating kernel. The hardware and software organization, structure, and interaction provide the framework for real-time control. This research included the use of encoders, digital boards, and a counter board; required the handling of interrupts, electric motor manipulation by servomotor controllers, and communication using RS232 and VMEbus technology. The kinematics algorithms and a real-time operating kernel were implemented using the C language. Shepherd research has laid the foundation for the flexible, robust, and precise motion needed for UXO clearing, http://archive.org/details/shepherdrotaryve109458957, Major, United States Marine Corps, Lieutenant, United States Navy, Approved for public release; distribution is unlimited.

The Yamabico Research Group at the Naval Postgraduate School is actively pursuing improvements in design and implementation of applications for it's family of autonomous mobile robots. This paper describes a new high level language for controlling the Yamabico- 11, surnamed OOPS-MML (Object-Oriented Program Specification for a Mobile robot Motion control Language). Conceptual goals included a user friendly, high level interface coupled with an abstract, efficient and compartmentalized architecture to employ a path tracking and motion control application developed at NPS. The result is a robust and flexible robot control system that is intended to be implemented and employed onboard the Yamabico- 11., http://archive.org/details/anobjectorientpr1094524206, Lieutenant, United States Naval Reserve, Approved for public release; distribution is unlimited.

Through an array of ground based radar sights and optical cameras, the United States military tracks objects in near and far Earth orbit. The sensors provide epoch and ephemeris information that is used to update a database of known objects. While a majority of the sensor observations are matched to their corresponding satellites, a small percentage are beyond the capabilities of current software and can not be correlated. These uncorrelated targets, UCT's, must be manually fitted by orbital analysts in a labor intensive process. As an alternative to this human intervention, the use of artificial intelligence techniques to augment the present computer code was explored. Specifically, an expert system for processing UCT's at the Naval Space Surveillance Command was developed. Rules were generated through traditional knowledge engineering methods and by a novel application of machine learning. The initial results are very good with the operational portions of the system matching the performance of the experts with an accuracy of 99%. Although not yet complete, the code developed in this research definitely shows the potential of using artificial intelligence to process UCT's., http://archive.org/details/anexpertsystemfo1094523894, Lieutenant Commander, United States Navy, Approved for public release; distribution is unlimited.

This research deals with the problem of extracting features from an image using wavelets and then using these features to recognize objects present in the image. This technique is applied to recognition of Unexploded Ordnance (UXO) objects. However, the concepts described here can be extended to recognition of other objects such as ships, missiles and aircraft. This work is performed as part of an ongoing effort to develop an autonomous vehicle capable of detecting UXOs, http://archive.org/details/objectrecognitio109458755, Lieutenant, Brazilian Navy, Approved for public release; distribution is unlimited.

A method of controlling a rotary vehicle to navigate a heading using a combination of translational and rotational motions by a plurality of driving-steering wheels controlling the motion in three degrees of freedom for a manned or unmanned vehicle having at least two drive-steering wheels, wherein a drive-steering wheel is a wheel with its heading orientation and driving velocity positively controlled, wherein a global motion is a vehicle trajectory with vehicle orientation from the initial position (with orientation) to a final destination (with orientation), which comprises comparing a global motion selected to the vehicle's body position and orientation to compute a motion instruction in three degrees of freedom. the acceleration, path of curvature, and rotation rate, collectively known as the motion command then converting the motion command into a translational speed, a translational direction, and a rotational rate and converting the translational speed, the translational direction and the rotational rate into the direction and driving speed for each independent drive-steering wheel.

The problem discussed in this dissertation is the development of an efficient method for visual navigation of autonomous vehicles. The approach is to significantly reduce the expensive computational time of landmark detection by straight-edge features. A novel, fast straight-edge-detection method for use in autonomous vehicle navigation and other image-understanding applications is presented. Straight edges in gray-scale images are detected using a new direction-control1ed edge tracking method, which gives precise estimate of the endpoints. To significantly reduce the number of exhaustive pixel computations, a random-hitting method using a pseudo-random number generator is proposed. Only if a generated pixel is significant do we start tracking the edge containing that pixel. To overcome the "noisy" gradient direction information, a robust least-squares linear fitting method is used to control the tracking process. The results of the algorithm show how it is efficient for landmark detection, which is important for motion control of autonomous vehicles. Thus the new method is implemented as a component of the image-understanding system in the autonomous mobile robot Yamabico-11 at the Naval Postgraduate School. An efficient world- modeling method based on the 2D model of the environment of the vehicle, including the heights of vertical edges in the environment, is presented. This modeling method is implemented with the new edge-detection method to improve the efficiency of the pose-determination algorithm (pose is a combination of the position and orientation of the camera), which is an essential task in the area of autonomous vehicle navigation

Millions of acres of the U.S. formerly used defense sites (FUDS) are contaminated with unexploded ordnance (UXO) as a result of past military use. The process of returning the land to the civilian sector is sensitive, intensive, and costly (e.g., millions of dollars, and the loss of human life). Hence'clearing (i.e., site remediation, range clearance, and explosive ordnance disposal) UXO's from FUDS is a complex problem. Existing clearing methods are inaccurate, dangerous, and labor intensive. This thesis shows that through robotics technology (e.g., Shepherd rotary vehicle with three degrees of freedom) and the use of advanced computer technology it is possible to make clearing tasks safer, more cost effective, and more efficient. An over arching hardware and software architecture was developed for Shepherd (including a self- contained on board computer system). The software system was developed for timer control, motion control, user interface, and an operating kernel. The hardware and software organization, structure, and interaction provide the framework for real-time control. This research included the use of encoders, digital boards, and a counter board; required the handling of interrupts, electric motor manipulation by servomotor controllers, and communication using RS232 and VMEbus technology. The kinematics algorithms and a real-time operating kernel were implemented using the C language. Shepherd research has laid the foundation for the flexible, robust, and precise motion needed for UXO clearing

This research deals with the problem of extracting features from an image using wavelets and then using these features to recognize objects present in the image. This technique is applied to recognition of Unexploded Ordnance (UXO) objects. However, the concepts described here can be extended to recognition of other objects such as ships, missiles and aircraft. This work is performed as part of an ongoing effort to develop an autonomous vehicle capable of detecting UXOs

This article deals with the problem of planning and controlling a radially symmetric six-legged walker on an uneven terrain when a smooth time-varying body motion is required. The main difficulties lie on the planning of gaits and foot trajectories. As for the gaits, this article discusses the forward wave gait of a variable duty factor and a variable wave direction. With the commanded body motion, the maximum possibly duty factor is computed using the speed limit of the leg swing motion. Guaranteeing this maximum duty factor contributes to obtain higher stability.Weprove the ‘‘continuity’’ of this forward wave gait planning algorithm adds the versatility to gaits planned. The foot trajectory planning algorithm dynamically generates a smooth foot trajectory as a function of the instantaneous body motions by modifying standard leg motion templates. The robot can negotiate an uneven terrain by modifying a vertical leg motion by a signal of tactile sensors on the foot. The experiments prove that the robot can successfully track smooth curves with body rotations on an uneven terrain, and thus prove the robustness of the algorithms.

This paper deals with the problem of planning and controlling a radially symmetric six legged walker on an uneven terrain when a smooth time varying body motion is required. The main difficulties lie on the planning of gaits and foot trajectories. As for the gaits, this paper discusses the forward wave gait of a variable duty factor and a variable wave direction. With the commanded body motion, the maximum possible duty factor is computed using the speed limit of the leg swing motion. Guaranteeing this maximum duty factor contributes to obtain higher stability. We proved the "continuity" of this forward wave gait planning algorithm adds the versatility to gaits planned. The foot trajectory planning algorithm dynamically generates a smooth foot trajectory as a function of the instantaneous body motions by modifying standard leg motion templates. The robot can negotiate an uneven terrain by modifying a vertical leg motion by a signal of tactile sensors on the foot. The experiments proved that the robot can successfully track smooth curves with body rotations on an uneven terrain, and thus proved the robustness of the algorithms.

This report contains 26 summaries of research projects in the Department of Computer Science which were carried out under funding of the Naval Postgraduate School Research Program. A list of recent publications is also included which consists of conference presentations and publications, books, contributions to books, published journal papers, and technical reports.

The path planning algorithm in Yamabico is based on a variation of Dijkstra's algorithm which has time complexity of O(n2). This algorithm works well in a dynamic environment, but a faster algorithm, called the All-Pairs Minimum Cost Paths algorithm, works even faster, O(1), in the case of a static environment. The computational complexity of the All-Pairs algorithm is O(n3), but if we know all pairs in advance, that is, the environment is static, we can preprocess them in advance, and use table lookup instead of Dijkstra's algorithm. Thus, we implemented a table lookup version for the static case, and kept Dijkstra's algorithm for the dynamic case. This results in both speed and flexibility. This thesis also investigated the Linear Fitting Algorithm for Sonar testing. Range and angle data, from sonar, was fit to a straight line, giving resolution of 1 to 2.5 cm when the robot is within 100 to 150 cm of the line.

One of the most difficult theoretical problems in robotics--motion planning for rigid body robots-- must be solved before a robot can perform real- world tasks such as mine searching and processing. This dissertation proposes a new motion planning algorithm for an autonomous robot, as well as the method and results of implementing this algorithm on a real vehicle. This dissertation addresses the problem of safely navigating an autonomous vehicle through free space of a two dimensional, world model with polygonal obstacles from a start configuration (position orientation) to a goal configuration using smooth motion under the structure of a layered motion planning approach. The approach proposes several new concepts, including v-edges and directed v-edges, and divides the motion planning problem of a rigid body vehicle into two subproblems: (1) finding a global path using Voronoi diagrams and for a given start and goal configurations planning an optimal global path; the planned path is a sequence of directed v-edges, (2) planning a local motion from the start configuration, using the aforementioned global path. The problem of how to design a safe and smooth path, is effectively solved by the steering function method and proximity. Another problem addressed here is how to make a smooth transition when the vehicle gets closer to an intersection of two distinct boundaries. This dissertation also presents a robust algorithm for the vehicle to continually eliminate its positional uncertainty while executing missions. This functionality is called self-localization which is an essential component of model-based navigation for indoor applications. This algorithm is based on the two-dimensional transformation group. Through this method, the robot can minimize its positional uncertainty, make safe and reliable motions, and perform useful tasks in a partially known world

Most existing real-time control systems use ad hoc static priority scheduling methods, in spite of the fact that the rate monotonic scheduling algorithm was proved to be the optimal static priority scheduling algorithm over 20 years ago. In this paper, we discuss a task library we are using for the real-time control of autonomous robots. The task library comprises a preemptive rate-monotonic scheduler which provides guaranteed optimal scheduling when certain conditions of processor utilization are met. The task library has been implemented as a collection of lightweight threads, which operate entirely in user-space for maximum efficiency. We show the performance advantages resulting from the reduced overhead of this approach, compared with commercial operating systems. The task system is robust, extensible, and portable, and has been successfully used to control the autonomous mobile robot Yamabico-11 developed at the Naval Postgraduate School

The major problem addressed by this research is how to plan a safe motion for autonomous vehicles in a two dimensional, rectilinear world. With given start and goal configurations, the planner performs motion planning which will lead a vehicle to achieve its task safely. During the planning, in addition to the safety consideration, motion's smoothness is also taken into account. The approach taken was to divide whole motion planning task into two layers. The top layer finds a global path by decomposing the free space into convex regions, then searching for an optimal global path class. The bottom layer performs local motion planning which further subdivides the planning problem into mid-portion and end-portion motion planning. The local motion planning is carried out region by region along the global path class. As results, simple motion instructions are generated for each region. For execution of planned motion, a software system, Model-based Mobile robot Language (MML- 11), was developed. This easy- to-use robot language provides users a convenient tool to program their applications through its function library. The results of the research were implemented in MML-1l and tested on an experimental robot Yamabico-11 successfully. (AN)

The major problem addressed by this research is how to develop a motion control algorithm for local motion planning of an autonomous mobile robot. The approach taken was to clearly define the robot's motion descriptions and to design a high-level, machine independent robot control language called MML (Model-based Mobile robot Language). The results are that the robot can implement smooth rotation, symmetric, and initial motion tracking on an given environment. Based on the motion control algorithm developed in this thesis, the robot with rigid body can be applicable in both local and global motion planning. The experimental results were successful. The algorithms were implemented first on a simulator, then on the autonomous mobile robot Yamabico- Ji. Given an initial and final configuration, the vehicle demonstrated the capability to safely achieve its goal.

The major problem addressed by this research is how to integrate an image understanding subsystem into an autonomous mobile robot so that the robot will be self-contained and independent of any Unix workstation for extracting image information. The resulting image understanding subsystem should be a part of the total intelligent autonomous robot and should provide functionality that will allow the robot to determine its position and that of obstacles in a partially known environment. The image understanding subsystem should be fully integrated with existing motion control and sonar software. The approach taken was to develop software to capture images that uses a VME bus to interface with a standard image manager. The software was written to provide a clean interface between the image grabber hardware and the robots existing Model-based Mobile Robot Language. By maintaining functionality similar to that provided in previous image understanding software, the changes needed to incorporate previous research software is kept to a minimum. The result is a autonomous robot that can capture images using a standard image manager, display those images and then convert them to a format needed by existing image understanding software to extract information for position determination or avoid obstacles. The image understanding software uses only the computing power it has available on-board the robot. This gives the robot the capability to extract image information about its environment and remain autonomous and self- contained.

This paper describes the objectives, organizations and research results of an international joint research project on an autonomous underwater walking robot conducted at the Naval Postgraduate School in the US and the Port and Harbour Research Institute in Japan. The main purpose of this project has been to investigate several scientific research questions in the US side and to enhance the capability of the existing six-legged underwater walking robot, AQCAUOBOT. Tlw period of this project is three years beginning in February 1992. Almost two dozen researchers have been involved in this project. The research results briefed in this paper include smooth motion planning for rigid body robots, versatile extended gait planning, dynamic foot motion planning: experiments with the real AQUAROBOT linkage dynamics, virtual reality simulation model, and task and mission level control architecture. In order to conduct these subprojeds, there have been frequent and close communications between two countries.

The problem with the Model-based Mobile robot Language(MML) processor is that the code is unstructured, causing the system to be unstable; it is very difficult to read because of deficient source code documentation; and because of poorly defined function interfaces and extensive functional coupling, the system is hard to maintain. To fix the MML processor, we performed a manual static analysis of the existing source code to understand its structure. Next, based on the analysis, the software system was restructured and the functionality enhanced. Finally, explicit source code documentation was added in the form of comments. There are several results with the new system. First, global variables are reduced from 152 to zero. Secondly, function interfaces are clearly defined and function coupling is enhanced. Finally, the source code is extensively documented. Following from these results, the new system is more stable, easier to read and understand, and sampler to modify

There are two goals for autonomous vehicle navigation planning: shortest path and safe path. These goals are often in conflict; path safety is more important. Safety of the autonomous vehicle's navigation is determined by the clearances between the vehicle and obstacles. Because a Voronoi boundary is the set of points locally maximizing the clearance from obstacles, safety is maximized on it. Therefore Voronoi Diagrams are suitable for motion planning of autonomous vehicles. We use the derivative of curvature k of the vehicle motion (dk/ds) as the only control variable for the vehicle where s is the length along the vehicle trajectory. Previous motion planning of the autonomous mobile robot Yamabico-11 at Naval Postgraduate School used a path tracking method. Before the mission began the vehicle was given a track to follow; motion planning consisted of calculating the point on the track closest to the vehicle and calculating dk/ ds then steering the vehicle to get onto track. We propose a method of planning safe motions of the vehicle to calculate optimal dk/ds at each point directly from the information of the world without calculating the track to follow. This safe navigation algorithm is fundamentally different from the path tracking using a path specification. Additionally motion planning is simpler and faster than the path tracking method. The effectiveness of this steering function for vehicle motion control is demonstrated by algorithmic simulation and by use on the autonomous mobile robot Yamabico 11 at the Naval Postgraduate School

The autonomous mobile robot, Yamabico-11. recognizes distance from obstacles by a transmit and receive sonar pair. However. the current sonar amplification has not been enough to obtain reliable range information. This thesis describes methods to improve the sonar analog circuits on the autonomous mobile robot so that they obtain more robust range information. One improvement was a change in the driving voltage of the transmit transducer from 5 volts to 12 volts which doubles the strength of sonar signal received by the pickup sensor. After changing the voltage source, it wa.., fouwi that there was spillover leakage directly from the transmitter to the receiver transducer. The amplifier sensitivity was decreased for the first one millisecond to reduce the spil!over.

The major problem addressed by this research is how to develop a motion control algorithm for stable and precise control of the motion of an autonomous mobile robot The approach taken was to clearly define the robot's motion descriptions and to design a high-level, machine independent robot control language called MML (Mode-based Mobile robot Language). The results are that the robot can implement line to line, line to circle, circle to circle path tracking or the combinations of these. Based on the motion control algorithm which was developed in this thesis, the robot is able to use external sensors to execute complicated missions such as obstacle avoidance (sonar is used in this thesis work).

This research addresses the problems experienced by the autonomous mobile robot, Yamabico- 11, with its ultrasonic sonar system. It explains the basics of acoustic theory as related to Yamabico- 11 and explains the sources of limitations imposed on Yamabico- 11 by the physical nature of the problem. This paper documents the basic characteristics of the sonar hardware and examines causes of sonar range errors. Finally, this research leads to improvements of the current sonar system to provide better directional coverage through a new sonar configuration

In order to continue to improve the usefulness of robots, it is becoming increasingly important to have them act as autonomous agents. A significant step toward this object is autonomous motion planning. This research was conducted as part of a broader effort to empower Yambico-11, a mobile robot under development at the Naval Postgraduate School, with ability to move autonomously. We believe that this problem is best attacked in layers. This thesis is our proposal for the initial layer. Given a robot's current location and its goal location, we use the homotopy relation to reduce the infinite set of path choices into a more manageable and smaller set of path classes. Specifically, we solve the problem of how to enable a robot to autonomously identify and label these classes of paths. We begin by decomposing the robot's operating environment into a collection of convex pieces called cells. The cells are transformed into a graph by adjacency. We show that every simple path on the graph corresponds to a unique simple homotopy class in the robot's world. We then search the graph to give each class a symbolic representation which also provides intermediate path planning clues. Subsequent layers can use these clues to form a more detailed plan. We implement the cell decomposition, graph transformation, and path class labeling as C programs, and preprocess them on a Unix workstation. This resulting data structures are then compiled and linked into the robot's kernel. All implementation has been integrated into the model- based mobile robot language (mml) used by Yamabico-11

The major problem addressed by this research is how to allow an autonomous vehicle to dynamically recognize changes in its environment, to map its environment, and alter its path to avoid obstacles while still reaching its goal point. The approach taken was to modify existing sonar functions in previous work, to better utilize sonars, and to perform many experiments to determine what data to expect from sonars while the vehicle is in motion. By applying the linear square fitting algorithm, the robot has the ability to map the objects within sensor range of an autonomous vehicle. The results are that, given an initial and goal point, the robot can proceed on a directed path, utilize its sonar sensor(s) used to detect obstacles, and when an obstacle is detected have the capability to dynamically compute a parallel path and smoothly alter its motion to the parallel path. The robot now has the capability to track the obstacle, and, once clear of the obstacle smoothly alter its motion to a path that will reach its goal point. The ability for the robot to combine smooth motion with obstacle avoidance has now been successfully programmed. Autonomous vehicle, Robot, Obstacle avoidance, Sonar sensing.

The major problem addressed by this research is how to improve an existing real-time software system's readability, maintainability, stability and portability using reengineering techniques. A fundamental portion of the Model- based Mobile robot Language (MML) was the real-time system chosen as the basis for this study. The approach taken was to create a new system design. The new design was based on system specifications obtained by conducting static and dynamic analysis on the existing system. The results are that a new core system was implemented using a design that focused on creating independent software sub-systems while encapsulating data. Hardware dependencies were localized and assembly code minimized. The new system is easier to understand and modify and is portable to other hardware platforms. Autonomous vehicle, Robot, Software engineering, Real-time system

We determine near-optimal paths with respect to work against friction and gravity on the surface of a vertical-axis ideal cone, assuming a moving agent of negligible size, friction proportional to the normal force, and power-maximum and stability limitations on the agent. This can provide good paths across hilly terrain for mobile robots. Our previous work required difficult-to-obtain polyhedral terrain models; cone surface patches permit easier and better models. We prove that our near-optimal paths on a vertical-axis cone surface are not much more complex than on polyhedra: There are qualitatively 22 kinds of path behavior (as opposed to four on a polyhedral face), and the area of the surface optimally reachable from a fixed start point by a given qualitative behavior has mathematically simple boundaries (only line segments, circle arcs, and arcs of logarithmic spirals in azimuth projection). We examine the possible cases based on agent characteristics and cone steepness, and provide "behavior maps" for quickly finding near-optimal paths between any two points on the cone surface. Our models incorporate discontinuous effects with respect to traversal heading. Comparisons with a program using uniform-grid path planning show our methods run considerably faster in less space, and our paths are much simpler to describe and easier to follow in the real world.

This thesis develops the mathematical relationships necessary to implement alternating tripod gaits on the hexapod underwater walking machine, AquaRobot. Analysis of documentation and application of Denavit-Hartenberg kinematic modeling techniques determine the fundamental vehicle parameters. Smooth leg motion models following elliptical and cycloidal trajectories are devised. Gait planning algorithms, using the elliptical smooth leg motion model, are developed for both discrete and continuous body motion. Statically stable, alternating tripod gait simulations are implemented in the C + + programming language. A stick figure graphics display allows examination and testing of the gait algorithms prior to incorporation in follow-on 3D graphics simulations or in real-time operation.

The problem addressed in this research was to provide a capability for sensing previously unknown rectilinear, polyhedral-shaped objects in the operating environment of the autonomous mobile robot Yamabico-11. The approach to the system design was based on the application of edge extraction and least squares line fitting algorithms of PET92 to real-time camera images with subsequent filtering based on the environmental model of STE92. The output of this processing was employed in the recognition of obstacles and the determination of object range and dimensions. These measurements were then used in path tracking commands, supported by Yamabico's Model-based Mobile Robot Language (MML), for performing smooth, safe obstacle avoidance maneuvers. This work resulted in a system able to localize objects in images taken from the robot, provide location and size data, and cause proper path adjustments. Accuracies on the order of one to ten centimeters in range and one-half to two centimeters in dimensions were achieved.