Development of a 2D Optimal Path Simulation for Ship-to-Shore Cranes : Safe Trajectories within Interchangeable Obstalce Environments

Abstract: The most advanced ports as of writing of this report are at least somewhat autonomous. Whether discussing the transporters between crane and stack (temporary storage) or cranes, the ports are shifting into a completely autonomous system. This ultimate goal presents a challenge in regards to unloading and loading cargo ships in the harbour. How do you achieve unloading of a ship without human intervention while still guaranteeing secure trajectories for the containers? ABB Ports in collaboration with the Division of Vehicular Systems at Linköping University have developed a simulation that utilises a simple control model to investigate the behaviour, limitations and capabilities of such an autonomous crane. Specifically, this simulation utilises a model of the dynamics of a Ship-to-Shore crane (STS), which has the task of unloading a ship. In order to set the crane model in context of realistic scenarios, some additions to the simulation are needed. One of these additions is obstacles. Before this thesis work, the model enjoyed an empty simulation environment to freely optimise how quickly the containers could be transported off of the ship. The addition of obstacles in the form of other containers on the cargo ship as well as the physical presence of the crane’s legs presents new challenges for the optimiser used to solve the optimal control problems formulated through the model in the simulation. The implementation of obstacles is one of the objectives for this thesis. This addition was implemented by modeling the obstacle dimensions and ship limitations by looking at the largest container ships in the world. Due to the simulation not containing obstacles previous to this thesis work, the initial guess provided to the solver initialised the solving in an area of convergence that is unfair to the solver, This rendered the simulation useless, as any obstacle presented to the solver would generate an infeasible solution. Another functionality needed for the obstacle implementation to be meaningful is a solution for guaranteeing safe trajectories for the containers from ship to shore. The solution utilised to reach this goal was to combine a convex hull and safety conditions where the convex hull covers the obstacles, including some padding to prevent collisions between the container carried (load) and obstacles. The safety conditions however calculates the potential positions of the load when an emergency stop occurs, and therefore can prevent the load from swinging into obstacles if there is an emergency stop. These implementations however changes the usefulness and performance of the simulation because of how they shrink the area of convergence for the solver and making some problems non-solvable. When considering both a convex hull and safety conditions, the usability of the simulations is harmed, but can still be utilised to learn about autonomous performance of the simulation. The optimal solutions include some interesting characteristics that can learn crane operators about how the control systems can be utilised. Such a simulation would benefit from continuous development in order to investigate further technologies and features that could improve both performance and usability. Areas such as homotopy, modelling ropes, comparison between simple and nuanced model would be truly interesting for future areas of investigation.