This paper describes the MineFleet distributed vehicle performance data mining system designed for commercial fleets. MineFleet analyzes high throughput data streams onboard the vehicle, generates the analytics, sends those to the remote server over the wide-area wireless networks and offers them to the fleet managers using stand-alone and web-based user-interface. The paper describes the overall architecture of the system, business needs, and shares experience from successful large-scale commercial deployments. MineFleet is probably one of the first commercially successful distributed data stream mining systems. This patented technology has been adopted, productized, and commercially offered by many large companies in the mobile resource management and GPS fleet tracking industry. This paper offers an overview of the system and offers a detailed analysis of what made it work.

This paper presents an overview of an experimental mobile and distributed data stream mining system that allows real time vehicle-health monitoring and driver characterization. It offers the motivation behind this application, explains the system architecture, discusses many challenges that the project faced, and shares some of the adopted solutions. The main contribution of the paper is our experience in building one of the very early distributed data stream mining systems for wireless applications that performs most of the data analysis related tasks using light-weight on-board computing devices. This paper points out that the distributed data mining technology can play a key role in solving real-life problems in a mobile application environment where computing, storage, power, and communication resources are limited. The paper also illustrates how privacy-preserving distributed data mining can play an important role in this type of applications.

A neural map algorithm has been employed to control a five-joint pneumatic robot arm and gripper through feedback from two video cameras. The pneumatically driven robot arm (SoftArm) employed in this investigation shares essential mechanical characteristics with skeletal muscle systems. To control the position of the arm, 200 neurons formed a network representing the three-dimensional workspace embedded in a four-dimensional system of coordinates from the two cameras, and learned a three-dimensional set of pressures corresponding to the end effector positions, as well as a set of 3/spl times/4 Jacobian matrices for interpolating between these positions. The gripper orientation was achieved through adaptation of a 1/spl times/4 Jacobian matrix for a fourth joint. Because of the properties of the rubber-tube actuators of the SoftArm, the position as a function of supplied pressure is nonlinear, nonseparable, and exhibits hysteresis. Nevertheless, through the neural network learning algorithm the position could be controlled to an accuracy of about one pixel (/spl sim/3 mm) after 200 learning steps and the orientation could be controlled to two pixels after 800 learning steps. This was achieved through employment of a linear correction algorithm using the Jacobian matrices mentioned above. Applications of repeated corrections in each positioning and grasping step leads to a very robust control algorithm since the Jacobians learned by the network have to satisfy the weak requirement that the Jacobian yields a reduction of the distance between gripper and target. The neural network employed in the control of the SoftArm bears close analogies to a network which successfully models visual brain maps. It is concluded, therefore, from this fact and from the close analogy between the SoftArm and natural muscle systems that the successful solution of the control problem has implications for biological visuo-motor control.>