Unmanned Surface Vehicle (USV) is an unmanned ship that is controlled through a remote control system (manual) or automatic control system (autopilot), move due to the thrust force from thruster machine and can turning due to the deflection angle of rudder. The USV path planning system becomes an important task so that the ship can make the global trajectory with the minimum travel distance according to the desired navigation while at the same able to avoid various obstacles from local dangerous situations that have the potential for collisions. To be able to do dynamic USV path planning, the Genetic Algorithm method with a sliding curve guidance system and PID MRAC controller is used. The use of this method gives smooth ship track performance with shortest distance in a 400x400 square meter map with static and dynamic obstacles. In a dynamic environment, the path replanning process that takes place in 0,98 seconds is able to find a new path that does not collide the obstacles. For the purposes of algorithm validation, the simulation is performed using MATLAB software with real ship parameters of 6 meters length USV.

No abstract

The Autopilot system for Unmanned Surface Vehicle (USV) can be applied by Sliding Mode Control which is a high frequency switching based control method and has discontinuous control action causing chattering on the system. Therefore, the Sliding Mode Control with natural control signal using a PID controller structure is applied to the USV. By using Sliding Mode Control, USV is expected to move accurately from the expected waypoint without any chattering on the system. The stability of the whole loop of system regulation is ensured using the Lyapunov stability function. The simulation results for autopilot system validate that the control parameters fit the time constant controller design specification and have zero steady-state errors. Further, autopilot system with waypoint navigation relatively generates small Mean Square Error (MSE) of waypoint.

This research is focusing on the development of metaheuristic algorithm to solve Dynamic Vehicle Routing Problem With Time Windows (DVRPTW) for the service provider of Inter-city Courier. The algorithm is divided into two stages which is static stage and dynamic stage. In the static stage, modified Ant Colony System is developed and in the dynamic stage, Insertion Heuristic is developed. In DVRPTW, vehicle’s routes are raised dynamically based on real time information, for example the reception of new order. To test the performances of the developed metaheuristic algorithm, the author compares the developed algorithm with the nearest neighbor algorithm and with the combination between the nearest neighbor and insertion heuristics algorithm. Experiment is done using Chen’s standard data test. The developed metaheuristic algorithm was applied on the network data of the roads in Surabaya, where the routes generated not only determine the order that the consumer must visit but also determine the routes that must be passed. After the experiment, the author conclude that the developed algorithm generates a better travel time total than the nearest neighbor and the combination between the nearest neighbor and insertion heuristics and can also generate route dynamically to respond to the new order well.

Vehicle Routing Problem is an issue in item delivery from depot to its customers using several vehicles which have limited capacity with a purpose to minimize transportation cost. The packing constraints exist because the vehicles which are usually used in item delivery have rectangular-box shaped container. Also, the items are commonly in shape of rectangular-box. Therefore, packing or loading method is needed so that containers could load all of the items without causing damage and could ease unloading process. The purpose of this final project is to develop a model and algorithm using metaheuristics method, especially genetics algorithm in order to minimize total delivery distance. A hybrid genetics algorithm and bottom-left fill algorithm also take place to solve the packing process. This algorithm delivered average solution 0.08% worse than ant colony optimization, but had 2.93% better solution than tabu search.