Stojce Deskovski

Intelligent system for freeway ramp metering control

The number of techniques implemented in subsystems of the Intelligent System of infrastructure in transportation system in terms of agents for signal control, ramp metering, detecting incidents is numerous. Challenges, however, are still there for the researchers to optimize traffic operations. The aim of this paper is to prove the ability of artificial intelligence technique known as reinforcement learning implemented in intelligent system for freeway control. Intelligent agents are implemented as controllers in order to provide optimal performance on the freeway corridor via ramp metering control on a corridor. The algorithm used in the research was Q learning algorithm. The results are promising proving that the technique is capable for optimal control of entrance freeway ramps and suitable for building the intelligent system of the freeway.

One Approach to the Integration of Low-Cost Inertial Sensors and Global Positioning System for Mobile Robots

Abstract Mobile robots are reality these days. Advances in the new technologies, including relatively cheap sensors, new communication and internet technologies together with the increases in the computational speed provide significant advances in the mobile robots autonomy. Precise navigation parameters (position, velocity and attitude) play crucial role for mobile robots to respond autonomously. Supplying wrong or not precise navigation parameters to the Guidance and Control System can turn to be catastrophic for mobile robots. In navigation the trend is to use integrated systems. The idea is to integrate more navigation systems and to achieve better navigation accuracy. In this paper a low-cost integrated navigation system was developed where Extended Kalman Filter (EKF) was used for integration of the Global Positioning System (GPS) information and measurements from the single yaw gyro. Extensive practical experiments were carried out to validate the presented approach. The algorithm was tested on a mobile robot platform Itar Pejo and a car. The results show improvements of the integrated navigation system then the GPS and gyro used alone.

Intelligent missile gudance system developed for training at Macedonian military academy

Abstract We have developed a simulation model of homing-guided anti-aircraft or surface-to air (SA) missiles – SAM, suitable for research, development and testing of different guidance and control laws, obtained from synthesis process. This system was developed for the teaching and training purposes of several subjects at Macedonian Military Academy. The topics in External ballistics give cadets knowledge in guided missiles movement, which are considered as dynamic systems and controlled objects. In homing-guidance systems for SA missiles – proportional navigation is most commonly used, and also improved solutions and optimal guidance laws based on modern control theory. All systems for guidance and control of SA missiles are complex and typically non-linear. In order to provide quality process and small guidance error – we have applied fuzzy logic to obtain some parameters for the guidance law. In this paper we present results from investigation of different guidance laws for hypothetic SA missile. The results from fuzzy guidance law combined with proportional navigation are also given. Simulation programs are realized in MATLAB/SIMULINK.

Fuzzy-Petri Net Reasoning System and Transfering of Knowledge to the Markov Chain

This work presents an overview of a reasoning expert system we have developed and implemented in C++. It is based on fuzzified Petri nets, with rule-based decision-making and appropriate knowledge base (KB). The reasoning algorithm is consisting ofcalculating the degrees of fulfillment (DOFs) for all rules of the KB and their assigning to the places of the Petri net. After this, it follows reasoning process with firing of active transitions and calculating of DOFs for output places (propositions of KB) and determining of fuzzy-distribution for output variables, as well as their defuzzified values. As final step, we are transferring these values to the states of a Markov chain in order to perform different command and control tasks. Markov chains are efficient tool for simulation and modeling of stochastic discrete event processes, especially those in military operations, like command and control activities.