When simulating a spiking neuron, numerical integration of synaptic input is often utilized to compute the neuron's depolarization. This report shows that the Numerical Integration Model(NIM) for spiking neuron simulations have a cumulative error that diverges unless the expectancy value for the local truncation error is zero. An alternative neuron simulation scheme, KM, was developed and is presented in this text. Experimental and theoretical results shows that the $kappa M$ error varies within a bounded domain.Experiments have been conducted on sample--and--hold implementations of the two models. A KM_{100} simulation, a KM simulation with 100 iterations per forcing function period, was compared with NIM simulations with finer temporal resolutions. It is shown that before 15 periods of a sinusoidal depolarizing input current has been simulated, the KM_{100} simulation produced a smaller error than a NIM_{10.000} simulation.Since the NIM simulation has a number of time steps that is two orders of magnitude larger than the KM simulation, this represents a significant efficiency improvement.

Horizontal wells with multistage hydraulic fracturing are today the most important drilling technology for shale gas extraction. Considered unprofitable before, the production has now become economically profitable due to advances in technology. Shales main characteristics is its low permeability, making the gas challenging and expensive to extract. Hydraulic fracturing stimulates the wells by creating additional conductivity, making the gas flows from storage pores to the well. This flow only possible in a short time scale, and states the need for multistage fracturing. Shale gas flow therefore exhibits a high initial peak, followed by a rapid decline in production rates. The use of shut-ins of shale gas wells allows for pressure build-up and may prevent liquid loading, as a means of boosting production. Shut-ins are used as on/off control variables in short-term model-based optimization of multiple shale gas wells with the objective of tracking a reference rate, while at the same time avoiding liquid loading. Previous work have focused on open-loop optimization. Here, an open-loop formulation is compared to a closed-loop formulation, in the form of mixed integer model predictive control. Both formulations are implemented in IBM ILOG CPLEX, with and without disturbances. Optimal production settings are solved in the presence of global constraints on production rates and minimal shut-in time. This allows for shut-ins with variable periods. The implementation is sensitive to initial conditions, horizons and weighting factors. The closed-loop formulation shows the best ability to reduce the effects of disturbances.

The NTNU Test Satellite (NUTS) project, is part of the Norwegian Student Satellite programANSAT. The goal of the project is to design and launch a double CubeSat by the endof 2014. During earlier satellite projects at NTNU, solid work on design of the attitudedetermination and control for a small satellite has been done.One of the considered estimation methods for the attitude determination is the extendedquaternion estimation method (EQUEST). Further development and testing of themethod is described in this thesis. In addition to the new EQUEST method, a nonlinearobserver has been implemented and tested. The simulation results for the two methods arecompared in order to find the attitude estimation method best suited for the NUTS satellite.The new EQUEST method has several advantages over the more common Kalmanfiltering for use in small CubeSats. It is less computationally costly, and has a fast startupand settling time. Magnetorquers, which affect the local magnetic field, are used asactuators for the satellite. This makes a fast algorithm preferable, since the attitude estimationand the attitude control should be performed separately. The nonlinear observer isslower than the EQUEST method, but it can guarantee global exponential stability and itis less vulnerable to disturbances. It is therefore introduced as an alternative solution forthe attitude determination problem.The original EQUEST method builds upon the QUEST method which has been extendedto include non-vectorized terms for gyroscope measurements and attitude predictionin the method’s cost function. In these terms, subtractions between the estimated andmeasured quaternions are used. The result is not entirely mathematically correct, eventhough previous testing of the method has been successful. The subtractions in the includedterms will not result in new attitude error quaternions.In this thesis, the method is further developed by replacing the subtraction terms withquaternion products. The new method is tested and compared to the original EQUESTmethod and an extended Kalman filter. It is also compared to the implemented nonlinearobserver. If the computational power of the NUTS satellite is sufficiently large, a combinationof the developed EQUEST method and the nonlinear observer could be considered forthe estimation. A combination of these two attitude estimation methods is implementedand the simulation results are analyzed.

Gas lift is an artificial lift technique which is intended to be used in oil and gas production systems to enhance the oil recovery rate. The technique is used in production systems which suffers from insufficient production rates because of inadequate reservoir pressure. The principle is to inject extra gas, from an external source, into the fluid mixture flowing out of the reservoir. This reduces the weight of the fluid column, which in turn enables the reservoir pressure to lift the mixture of fluids up to the surface. Thus, the restriction in production rate, resulting from insufficient reservoir pressure, is resolved. The casing heading instability cycle is a phenomenon which may occur in such gas lift systems at certain pressure and flow conditions. In such a case, a blocking constraint leads to an accumulation of gas that is building up a high pressure in the container storing the extra gas. Eventually the pressure is able to overcome the blocking constraint which further leads to a production flow blow out at the surface. This process repeats itself over a period of several hours. The resulting flow regime may be damaging to the receiving surface equipment. It also leads to a decrease in the average production rate, which is highly undesirable.In this thesis a new low order mathematical model to describe the casing heading instability is presented. The model is further compared and fitted to a gas lift case implemented in the advanced OLGA flow simulator, which is regarded as the real flow process in this thesis. A controllability analysis is performed on the newly fitted model, and the results of the analysis are used to investigate which control structures that seems reasonable for stabilizing the casing heading instability. The most promising control strategies are further tested in simulations, both on the low order model and on the advanced flow simulator. The results from the analysis, which is further confirmed in simulations, show that out of the measurement candidates that were considered in this thesis, the bottom hole pressure, i.e., the pressure measured in the well or near the depth of the producing formation, is the single best measurement to use for both disturbance rejection and stabilization of the system. However, this measurement is known to be unreliable, if at all available, because of its location in a harsh environment. Another control structure which showed promising results both in analysis and simulations is therefore concluded as the better choice. This is a cascaded control structure, which only relies on realistically available top side pressure measurements. It proved successful in stabilizing the system.

To aid in faster and better decision making it is interesting to couple advanced simulators with optimization tools.Most simulators however does not offer gradients, therefore derivative-free methods must be used. In this thesis optimization of and oil and gas field with free routing is considered. By embedding the structural information in the optimization problem and approximating the simulators by polynomials a MINLP problem is formed which can be solved by gradient based solvers. This approach requires that the polynomial models are updated frequently to fit the simulators. Each update requires several simulations and creates a trade-off between robustness and computation time. Different updating strategies for the models are considered in this thesis. By solving a separate optimization problem to update the models the MINLP problem can be formulated as a convex problem which is solved in a branch and bound framework and with an interior-point. Two approaches to updating the models in respect to the branch and bound method are explored, and it is found to be more robust to update the models for each node of the branch and bound tree, ensuring a local fit before branching.

Operation and maintenance O&M is an expensive but necessary part of almost any industrial endeavor. This becomes even more expensive when the object in need of O&M is located in remote places. This will be the case if you have a wind farm at sea. To get the cost of this down research into using a robot or possibly a robotic arm to do remote O&M has been initiated. This Thesis deals with some of the challenges when it comes to remote presence or telepresence and the use of a robotic arm.Telepresence systems have some unique challenges when it comes to control. The time delay and limits when it comes to field of view has led many to advocate a solution called a Predictive display.Predictive displays lets the user see an immediate response to control input. It does this by simulating the system that is being controlled and then in some way synthesize an image that would show a simulated response.This Thesis is concerned with the development of such a display. The solution proposed here leads to the use of a real time simulator of a robotic arm. In addition to this I have worked with a possible solution for synthesizing the view of the robot by using a game engine to render the robot and its environment.The resulting application as far as the predictive display is concerned leads me to believe that the tools I have used can be used for the purposes described in this thesis.The predictive display method seams like the best alternative when it comes to providing the operator with information about the robot's surroundings. And thus increasing the operator's situational awareness.The report shows how the application can be created and provides you with information on how the solution can be extended.Further, the application created should make it possible to carry out further experiments and help in future developments within this project.

With ever cheaper and more versatile robots, the use of robotic systems in creases rapidly. Although robots are becoming more intelligent, the cognitive capabilities of humans can still not be matched. By combining the intelligence of a human operator with the strength, endurance and size of a robot, in addition to separating the robot and operator to avoid danger to the operator, the applications are innumerable. The use of an operator to remotely control a robot is often referred to as teleoperation.In a teleoperation system it is important to present the state of the robot and the remote environment with high accuracy and in a comprehensible way. With a large number of sensor data, a solution is to enhance the feeling of telepresence or transparency of the system. That is, making the human operator feel like he or she is interacting directly with the manipulated environment. This could be achieved by using a haptic joystick, which is able to generate force feedback to the operator, to present information about the slave robot. Examples of such informations are the distance to an obstacle or deviation from a desired movement. Such a system is often called a bilateral teleoperation system, where the stability is especially sensitive to transmission delay. This time-delay is often introduced by the communication network between the human operator and remote robot.This thesis presents a control architecture for interpreting a change in the joystick position to a desired end-effector velocity for a mobile manipulator. In addition to calculating the velocity, the controller is designed to comply with the joint limits, optimize the manipulability and handle time-varying transmission delay. A force, that depends on the deviation between the desired and measured end-effector position, is sent back to the human operator, as well as a visual feedback. To increase the precision of the end-effector movement the position of the movable robot base is fixed when the manipulability is above a given threshold, and moves only to increase the workspace of the robot. The designed system is implemented using Robot Operating System (ROS) and tested on a virtual mobile manipulator. The virtual robot is based on a model of a Schunk LWA3 7-DOF manipulator, mounted on a Seekur Jr. wheeled mobile base.Several experiments prove that the system with the proposed control architecture is stable when under influence of constant, as well as variable time-delay. Any standard deviation between the measured and desired end-effector position is eliminated, and the trajectory of the end-effector is almost identical the desired, though delayed when affected by communication delay. Neither the force feedback nor end-effector position show indications of dramatic change at the transition between fixed and moving robot base. Simulations with human operators show that they are able to move the end-effector of a virtual mobile manipulator from an initial position to a predefined goal, with the use of a Phantom Omni, haptic joystick.

This thesis is a contribution to the Unmanned Aerial Vehicle (UAV) project at the Department of Engineering Cybernetics, which is a project where contributions from master students and Phd students will result in an autonomous aerial vehicle. The unmanned vehicle laboratory has its own UAV, the Odin Recce D6 delta-wing aircraft which is to be considered in the overall project. When the UAV is in the air on a mission, one important thing is to ensure that the UAV detects obstacles, such as mountains, buildings and other aircrafts. No-fly areas should be avoided by the path planner. This thesis considers a guidance system that will set up a path from the initial position to the final destination, and make sure that the generated trajectory is safe.One problem with the design of the optimal path has been that the designed path gives textit{corner cutting} when obstacles from the environment is included in the path-planner. To avoid this problem, which happens because discrete time is considered, two different solutions to avoid this problem have been discussed closer. Implementation of constraints and different cost functions for path planning with collision avoidance using the Mixed Integer Linear Programming (MILP) is one of the purposes of this thesis. The MILP algorithm is developed for the case of planar motion where the UAV has to fly around the obstacle, and can't fly over or under it. The design of the path path planner using MILP is done in two different ways. One where obstacles are known at the beginning of the optimization, and one where obstacles are added as information to the path planner when they are in the range of the UAVs radar. It is shown that the implementation with obstacle radar detection is more realistic, and that it also improves the computation time. As the author knows this method has not been published in articles up to this date. Two different approaches for search of a defined area with an arbitrary number of UAVs with camera systems have been developed and implemented through this thesis. As far as the author of this thesis knows these approaches for search have not been published up to this date. Efficient search and low computational complexity has been important design factors during the development of these approaches.The final systems are simulated in MATLAB for some test-scenarios. Also, reflection and discussion on further improvement on the path planning system are included in the report. This includes further improvement of the guidance system using receding horizon strategies.A literature study on path planning with receding horizon has been done.

Visualization is a technique for creating a graphical representation of a reality, with possible augmentations. It can be used in a wide variety of applications in the area of robotics. An example usage is for teleoperation, where a human operator can control a robot from a different location by having the robotic process visualized in real-time. Furthermore, visualization software can also be used for doing detailed analysis of a process, for instance to debug a system that is under development.This thesis proposes and implements a system that visualizes robot cells in 3D. The system presented here uses RViz as the actual visualization program, and Robot Operating System (ROS) to connect the different parts of the system together. In all, the system is organized in a way that makes it easy to combine several different robots and other devices in one view. The main part of the system presented in this thesis is a Graphical User Interface (GUI), which is used to control the actual visualization, as well as the rest of the system. This GUI provides functionality to record and play back measurement data from the different devices and sensors, making it possible for a human operator to pause, resume and step through the data one value at a time. To be able to use this general system on a real robotic process, it has to be customized by specifying the available devices, and information about these, in a configuration file.The developed system was tested on the Ekornes sewing cell at SINTEF Raufoss Manufacturing. The sewing cell consists of a sewing machine and two industrial robots, that are used for sewing together leather covers. Based on this testing it was seen that the program is a powerful tool for analyzing robotic processes, because of the possibility to record and replay data, as well as providing the operator with a very informative visualization view.

This thesis will present a design of a guidance and control system to use on aircrafts, primarily on UAVs. One control method for heading control and two for pitch and altitude control will be investigated. The control methods are Proportional-Integral-Derivative (PID) and sliding mode control. PID will be tested on both heading and pitch and altitude control, while sliding mode will only be applied to pitch and altitude. There will be presented a path-following method, Line of Sight, for heading guidance and a kinematic controller for altitude reference. The presented methods are implemented in Matlab Simulink while the aircraft model used comes from the flight simulator X-Plane. X-Plane is also used to visualize the performance of the autopilot design.PID and sliding mode control are tested in four different scenarios to investigate which controller who has the best performance. After the simulations, it was observed that the PID had better performances than sliding mode control.