Michael G. Kay

Multisensor integration and fusion in intelligent systems

The issues involved in integrating multiple sensors into the operation of a system are presented in the context of the type of information these sensors can uniquely provide. A survey is provided of the variety of approaches to the problem of multisensor integration and fusion that have appeared in the literature in recent years ranging from general paradigms, frameworks, and methods for integrating and fusing multisensory information to existing multisensor systems used in different areas of application. General multisensor fusion methods, sensor selection strategies, and world models are examined, along with approaches to the integration and fusion of information from combinations of different types of sensors. Short descriptions of the role of multisensor integration and fusion in the operation of a number of existing mobile robots are provided, together with proposed high-level multisensory representations suitable for mobile robot navigation and control. Existing multisensor systems for industrial and other applications are considered. >

Comparison of genetic algorithms, random restart and two-opt switching for solving large location-allocation problems

This paper examines the application of a genetic algorithm used in conjunction with a local improvement procedure for solving the location-allocation problem, a traditional multifacility location problem. This problem is difficult to solve using traditional optimization techniques because of its multimodal, nonconvex nature. The alternate location-allocation (ALA) method has been shown to be an effective local improvement procedure for the location-allocation problem. Using the ALA method, an empirical analysis was done to determine the number and size of the local minima of the location-allocation problem to demonstrate the reduction of the size of the search space that can be achieved through the use of the ALA method as an evaluator. A genetic algorithm that evaluates a series of ALA solutions was developed and compared to two traditional heuristic procedures for the problem: random restart and H4, a two-opt procedure. Like the genetic algorithm, both procedures evaluate a series of ALA solutions. A statistical analysis of the quality of the solutions provided by the three procedures for several problems of varying size demonstrated that the genetic algorithm provides the best solutions. An examination of the number of ALA evaluations performed by each procedure showed that the genetic algorithm also found solutions to the larger size problems much quicker than either the random restart or the H4 procedures.

Empirical investigation of the benefits of partial lamarckianism

Genetic algorithms (GAs) are very efficient at exploring the entire search space; however, they are relatively poor at finding the precise local optimal solution in the region in which the algorithm converges. Hybrid GAs are the combination of improvement procedures, which are good at finding local optima, and GAs. There are two basic strategies for using hybrid GAs. In the first, Lamarckian learning, the genetic representation is updated to match the solution found by the improvement procedure. In the second, Baldwinian learning, improvement procedures are used to change the fitness landscape, but the solution that is found is not encoded back into the genetic string. This paper examines the issue of using partial Lamarckianism (i.e., the updating of the genetic representation for only a percentage of the individuals), as compared to pure Lamarckian and pure Baldwinian learning in hybrid GAs. Multiple instances of five bounded nonlinear problems, the location-allocation problem, and the cell formation problem were used as test problems in an empirical investigation. Neither a pure Lamarckian nor a pure Baldwinian search strategy was found to consistently lead to quicker convergence of the GA to the best known solution for the series of test problems. Based on a minimax criterion (i.e., minimizing the worst case performance across all test problem instances), the 20% and 40% partial Lamarckianism search strategies yielded the best mixture of solution quality and computational efficiency.

A tutorial on multisensor integration and fusion

A tutorial introduction to the subject of multisensor integration and fusion is presented. The role of multisensor integration and fusion in the operation of intelligent systems is defined in terms of the unique type of information multiple sensors can provide. Multisensor integration is discussed in terms of basic integration functions and multisensor fusion in terms of the different levels at which fusion can take place. Numerical examples are given to illustrate a variety of different fusion methods. Speculations concerning possible research future directions and a guide to survey and review papers in the area of multisensor integration and fusion are presented.<<ETX>>

Mathes: An expert system for material handling equipment selection

Abstract This paper describes a rule-based expert system, MATHES (MATerial Handling Equipment Selection), that selects appropriate types of material handling equipment for intra-factory moves of unitized material. The equipment types are chosen by applying heuristic selection rules (acquired from a human expert) to responses from a user concerning relevant details about the characteristics of a material move. The heuristic rules relate characteristics of the move to appropriate types of material handling equipment. Associated with each selected equipment type is a factor that can be used to order the list of selected equipment as to each types degree of appropriateness.

Multilayered fuzzy behavior fusion for real-time reactive control of systems with multiple sensors

Fuzzy linguistic rules provide an intuitive and powerful means for defining control behavior. Most applications that use fuzzy control feature a single layer of fuzzy inference, mapping a function from one or two inputs to equally few outputs. Highly complex systems, with large numbers of inputs, may also benefit from the use of qualitative linguistic rules if the control task is properly partitioned. This paper presents a modular fuzzy control architecture and inference engine that can be used to control complex systems. The control function is broken down into multiple local agents, each of which samples a subset of a large sensor input space. Additional fuzzy agents are employed to fuse the recommendations of the local agents. Real-time implementation without special hardware is possible by using singleton output values during fuzzy rule evaluation. A development tool is used to translate a fuzzy programming language offline for fast execution at run time. Using this system, a multilayered fuzzy behavior fusion based reactive control system has been implemented on an autonomous mobile robot, MARGE, with great success. MARGE won first place in Event III of the 1993 Robot Competition sponsored by the American Association for Artificial Intelligence.

Autonomous mobile robot global motion planning and geometric beacon collection using traversability vectors

Approaches in global motion planning (GMP) and geometric beacon collection (for self-localization) using traversability vectors have been developed and implemented in both computer simulation and actual experiments on mobile robots. Both approaches are based on the same simple, modular, and multifunctional traversability vector (t-vector). Through implementation it has been found that t-vectors reduce the computational requirements to detect path obstructions, Euclidean optimal via-points, and geometric beacons, as well as to identify which features are visible to sensors. Environments can be static or dynamic and polygons are permitted to overlap (i.e., intersect or be nested). While the t-vector model does require that polygons be convex, it is a much simpler matter to decompose concave polygons into convex polygon sets than it is to require that polygons not overlap, which is required for many other GMP models. T-vectors also reduce the data size and complexity of standard V-graphs and variations thereof. This paper presents the t-vector model so that the reader can apply it to mobile robot GMP and self-localization.

Multi-layered fuzzy behavior fusion for real-time control of systems with many sensors

A modular architecture for real-time fuzzy mapping of sensors to control signals is presented. The function is broken down into multiple agents, each of which samples a subset of a large sensor input space. Additional fuzzy agents are employed to fuse the recommendations of the local agents. Real-time implementation without special hardware is possible by using singleton output values during fuzzy rule evaluation. A linguistic syntax for fuzzy systems development is presented, allowing complex nonlinear control functions to be defined using qualitative expressions rather than mathematical terms. Our development tool, PCFUZ, translates this syntax off-line into a data structure for fast execution at run time. Using this system, a fuzzy behavior-based reactive control system has been implemented on an autonomous mobile robot, MARGE, with great success.<<ETX>>

Multilayered fuzzy behavior fusion for reactive control of an autonomous mobile robot

Fuzzy linguistic rules provide an intuitive and powerful means for defining control behavior. Most applications that use fuzzy control feature a single layer of fuzzy inference, mapping a function from one or two inputs to equally few outputs. Highly complex systems, however, may benefit from qualitative rules as well if the control task is properly partitioned. This paper presents a modular fuzzy control architecture and inference engine. A control function is broken down into multiple agents, each of which samples a subset of a large sensor input space. Additional fuzzy agents are employed to fuse the recommendations of the local agents. Real-time implementation without special hardware is possible by using singleton output values during fuzzy rule evaluation. Using this system, a fuzzy behavior-based reactive control system has been implemented on an autonomous mobile robot MARGE, with great success.

Multisensor Integration And Fusion: Issues And Approaches

Issues concerning the effective integration of multiple sensors into the operation of intelligent systems are presented, and a description of some of the general paradigms and methodologies that address this problem is given. Multisensor integration, and the related notion of multisensor fusion, are defined and distinguished. The potential advantages and problems resulting from the integration of information from multiple sensors are discussed.