OCP Based Online Multisensor Data Fusion for Autonomous Ground Vehicle Nanang Syahroni Electronic Engineering Polytechnic Institute of Surabaya, Surabaya 60111, Indonesia email: nanang@eepis-its.edu Abstract In this paper, online multisensor data fusion algorithm using CORBA event channel is proposed, in order to deal with simplifying problem in sensor registration and fusion for vehicle’s state estimation. The networked based navigation concept for Autonomous Ground Vehicle (AGV) using several sensors is presented. A simulation of various application scenarios are considered by choosing several parameters of UKF, i.e. weighting constant for sigma points  and , and square root matrix . Normalized mean square error (MSE) of Monte Carlo simulations are computed and reported in the simulation results. Furthermore, the middleware infrastructure based on Open Control Platform (OCP) to support the interconnection between the whole filter structures also reported. Keywords: Autonomous Vehicle, OCP, CORBA, UKF 1. Introduction The software based controls for robotic and autonomous vehicle have been dominated in recent years. An OCP is an object-oriented software infrastructure implemented that allows seamless integration of cross- platform software and hardware components in any control system architecture. An OCP is a middleware that is based on the real-time common object request broker architecture (RT-CORBA). Middleware is connectivity software that consists of a set of services, allowing multiple processes running on one or more machines to interact across a network [1]. The multisensor data fusion using Kalman filter (KF) has been widely applied in integrated navigation system for many applications [2-3]. Estimation of navigation system in nonlinear system approach to use the extended Kalman filter (EKF) which simply linearizes all nonlinear models are reported in [4-5], so that the traditional KF can be used. An alternative approach, the unscented Kalman filter (UKF) where the random variable, Gaussian distributions is linearized while the nonlinear model equations are directly used in the calculations [6]. The centralized filter where all measured sensor data are communicated to the central site for processing [7], and distributed filter [8] where the local estimators from all sensor can yield the global optimal or sub optimal state estimator according to certain information fusion criterion. In this paper, an online decentralized multisensor data fusion of two stage federated UKF algorithms connected by RT-CORBA middleware network is proposed. We assumed that the problem solution of fault detection and isolation in the Autonomous Ground Vehicle (AGV) will made easily. 2. System Models When the vehicle negotiates a turn, this motion can be described as a rotation about Instantaneous Centre of Rotation (ICR) with the same angular speed, we assumed that there is no slip between the tires and the ground. In the figure1, a degree of freedom that appear are the steer angle  and drive speed  of the front wheel can be expressed as function of the control inputs which act on each wheel. B is the distance between the front and rear axles, and the width of front axle is H. The filters state space x consists of the position of the body (x 1 and x 2 ), its velocity (x 3 and x 4 ), and a parameter of its aerodynamic property (x 5 ). The vehicle state dynamics in a simple form are:   0 1 3 2 4 3 3 1 1 4 4 2 2 5 3 0 2 0 5 () () () () () () () () () () () () () () () () () () () ( )exp () () () () () ( )exp () m x k x k x k x k x k Akx k Dkx k k x k Akx k Dkx k k x k k Ak k R Rk Vk Dk r k k k x k D                         (1) where () A k is acceleration-related force, () Dk is breaking-related force, 1 2 3 , , and    is measurement noise,  is vehicle characteristic uncertainly, () Rk is distance from central 2 2 1 2 = () x k x  , and () Vk is absolute vehicle speed 2 2 3 4 = () x k x  . In summary, the state space is:   1 1 2 T t X X X R       (2)