This thesis concerns the application of different multi-objective optimization (MOO) methods and strategies for finding the optimal envelope for a given building plot and lighting performance indicators. More specifically, the PV potential and daylighting potential of the building are maximized using different optimization solvers. Auxilliary objectives are introduced to constrain the model to a certain compactness and size. The method utilises an existing data framework called TopoGenesis and solves the problem using the PyGmo library.

A ray tracing is used to find all possible collisions between the objective test
points and the building mass and environment. The problem is first presented as a standard integer programming problem, but solving this problem is not feasible if complexity needs to be kept at a reasonable level. An alternative method of continuous optimization is therefore proposed that uses (meta)heuristics to find an optimal solution for maximizing the objective functions. The occupation status of the massing is used as inputs for the decision variables.

After the application of this method on small scale toy problems, a few of the
design options are selected and evaluated by their performance indicators, as well as the measure with which the option makes sense from a more traditional design perspective. The comparison of the performance and results of both methods give insight into the recommended workflow, settings, and pitfalls for finding an optimal solution to a multi-criteria design problem with visibility objectives. From the initial results, the Non-Sorted Genetic Algorithm seems to be the best option for solving these types of problems, and the PV potential objective is validated. The daylighting potential objective performs less satisfactory and suggestions are made on alternative approaches for this metric.