Browsing by Author "Kaski, Tuomas"

Now showing 1 - 2 of 2

Gamification in a web service for learning the use of construction modeling software
(2013) Salminen, Maija
School of Science | Master's thesis
In this thesis, gamification in Tekla Campus self-learning online service was researched by following a user centric gamification process. It was concluded that the users prefer gamified features related to social status and community forming, tracking and planning one's progress, learning step-by-step and constant feedback. These features were suggested to be implemented in the service to engage and motivate the users. The research begins with explaining the history of gamification and studying the relation of games and learning. Next, online learning services and other building information modelling software companies' online services than Tekla Campus are mapped out. The research was performed by following a gamification process involving the users. Motivating features were ideated with the users and implemented in a prototype. The prototype was tested with users to validate the design based on the features, and the results were analysed to guide further development of the service.
Learning and performing skill demanding tasks with WorkPartner robot
(2004) Kaski, Tuomas
Helsinki University of Technology | Master's thesis
This thesis researches the learning and performing of skill demanding tasks with an advanced service robot called WorkPartner. The skill demanding task studied in this work composes of finding a box from the surroundings, moving to the box, lifting the box and taking the box to a specific place. In this thesis a learning method was used, where a genetic algorithm is used in certain sub tasks of the robot to seek better solutions for the subtasks. With this algorithm the robot's performance of a task gets better with each attempt at performing the task. This means that the robot can learn through trial and error. The genetic algorithm improves the performance of subtasks by changing the parameters of the microtask which are parts of the subtask. In addition to this the algorithm can add and remove microtasks and change the order in which the microtasks are executed within the subtask. This increases the performance of the skill demanding task considerably. When a robot can learn to perform a task, making difficult and complicated skill demanding tasks gets easier. The designer of a task doesn't have to waste time trying to come up with the perfect solution to the task. It is enough if WorkPartner is told what components (microtasks) the task has, it is given starting values (random or specific) and its task performance is ranked with a fitness value. The robot does the rest. This research produced a computer program that is used to teach the "find and lift box" -task to WorkPartner. By further developing the program, a programming platform can be made for producing other skill demanding tasks. This one platform could control all learned tasks of WorkPartner.