No / This paper presents an investigation into the challenges in implementing a hard real-time optimal non-stationary system using general regression neural network (GRNN). This includes investigation into the dynamics of the problem domain, discretisation of the problem domain to reduce the computational complexity, parameters selection of the optimization algorithm, convergence guarantee for real-time solution and off-line optimization for real-time solution. In order to demonstrate these challenges, this investigation considers a real-time optimal missile guidance algorithm using GRNN to achieve an accurate interception of the maneuvering targets in three-dimension. Evolutionary Genetic Algorithms (GAs) are used to generate optimal guidance training data set for a large missile defense space to train the GRNN. The Navigation, Constant of the Proportional Navigation Guidance and the target position at launching are considered for optimization using GAs. This is achieved by minimizing the. miss distance and missile flight time. Finally, the merits of the proposed schemes for real-time accurate interception are presented and discussed through a set of experiments. (C) 2012 Elsevier Ltd. All rights reserved.

Global Navigation Satellite System/Inertial Measurement Unit (GNSS/IMU) Integrated Navigation Systems (INS) integrate the positive features of GNSS and IMU for optimal navigation guidance in high accuracy outdoor navigation systems, for example using Extended Kalman Filter (EKF) techniques. Time synchronization of IMU data with precise GNSS based time is necessary to accurately synchronize the two systems. This must be done in real-time for time sensitive navigation applications such as autonomous vehicles. The research is done in two parts. The first part is the simulation of inaccurate time-stamping in a single axis of nonlinear input data in a gyroscope and an accelerometer, to obtain the timing error value that is tolerable by a high accuracy GNSS/INS system. The second part is the creation of a real-time algorithm using an STM32 embedded system enabled with FreeRTOS real-time kernel for a GNSS receiver and antenna, along with an IMU sensor. A comparative analysis of the time synchronized system and an unsynchronized system is done based on the errors produced using gyroscope and accelerometer readings along a single axis from the IMU sensor, by conducting static and rotational tests on a revolving chair.The simulation concludes that a high accuracy GNSS/INS system can tolerate a timing error of up to 1 millisecond. The real-time solution provides IMU data paired with updated GNSS based time-stamps every 5 milliseconds. The timing jitter is reduced to a range of ±1 millisecond. Analysis of final angular rotation error and final position error from gyroscope and accelerometer readings respectively, indicate that the real-time algorithm produces a reduction in errors when the system is static, but there is no statistical evidence showing the reduction of errors from the results of the rotational tests. / GNSS / IMU integrerade navigationssystem kombinerar de positiva egenskaperna hos GNSS och IMU för optimal prestanda i noggranna navigationssystem. Detta görs med hjälp av sensorfusion, till exempel EKF. Tidssynkronisering av IMU-data med exakt GNSS-baserad tid är nödvändigt för att noggrant synkronisera de två systemen. Detta måste göras i realtid för tidskänsliga navigationsapplikationer såsom autonoma fordon. Forskningen görs i två delar. Den första delen är simulering av icke-linjär rörelse i en axel med felaktig tidsstämpling hos ett gyroskop och en accelerometer. Detta görs för att erhålla det högsta tidsfel som är acceptabelt hos ett GNSS / INS-system med hög noggrannhet. Den andra delen är skapandet av en realtidsalgoritm med ett inbyggt STM32-system med FreeRTOS som realtidskärna för en GNSSmottagare och antenn, tillsammans med en IMU-sensor. En jämförande analys av det tidssynkroniserade systemet mot ett osynkroniserat system görs baserat på de positionsfel längs en axel som produceras av gyroskopoch accelerometermätningar. Detta görs genom att utföra statiska och roterande tester med hjälp av en roterande stol.Simuleringen visar att ett noggrant GNSS / INS-system tolererar ett tidsfel på upp till 1 millisekund. Realtidslösningen ger IMU-data med tidsstämplar synkroniserade med GNSS-tid var femte millisekund. Tidsjittret reduceras till ett intervall mellan ± 1 millisekund. Analysen av det slutliga vinkelrotationsfelet och positionsfelet från gyroskopoch accelerometermätningar indikerar att realtidsalgoritmen ger ett lägre fel när systemet är statiskt. Det finns dock inga statistiska bevis för förbättringen från resultaten av rotationstesterna.

Collision avoidance is an essential component of autonomous shipping. As ships begin to advance towards autonomy, developing an advisory system is one of the first steps. An advisory system with a strong collision avoidance component can help the crew act more quickly and accurately in dangerous situations. One way to avoid colission is to make the vessel stop as fast as possible. In this work, two scenarios are studied, firstly, stopping along a predefined path, and secondly, stopping within a safe area defined by surrounding obstacles. The first scenario was further worked with to formulate a real-time solution. Movements of a vessel, described in three degrees of freedom with continuous dynamics, were simulated using mathematical models of the forces acting on the ship. Nonlinear optimal control problems were formulated for each scenario and solved numerically using discretization and a direct multiple shooting method. The results for the first problem showed that the vessel could stop without much deviation from the path. Paths with different curvatures were tested, and it was shown that a slightly longer distance was traveled when the curvature of the path was greater. The results for the second problem showed that the vessel stays within the safe area and chooses a relatively straight path as the optimal way of stoping. This results in a shorter distance traveled compared to the solution of the first problem. Two different real-time approaches were formulated, firstly a receding-horizon approach and secondly a lookup-based approach. Both approaches were solved with real-time feasibility, where the receding-horizon approach gave a better solution while lookup-based approach had a shorter computational time.