IoT and Smart Homes get more and more popular these days, thereby the desire to automize more areas of daily life, grows stronger. The aim of this thesis is to create an autonomous smart home garden assistant that supports the user in the daily garden care and during longer time periods, like holidays. The garden assistant assumes the following tasks: Watering, (soil) heating, artificial lighting and hazardous precipitation protection by a shutter. Additionally, a new AI-based irrigation algorithm is developed to reduce the water consumption. This intelligent irrigation system uses an Artificial Neuronal Network-based evaporation prediction and the information of weather forecasts to generate optimal irrigation plans. Classically, equations with many specific parameters have to be used for an accurate evaporation prediction. Hence, the goal of the usage of ANNs is to simplify the evaporation prediction for the individual conditions in each garden, as the system is self-adapting. A fully functional prototype of the garden assistant is developed and evaluated during a long-term experiment (end April to mid of July 2021), to prove the advantages of the system. In this experiment, the intelligent irrigation system saves approx. 25% water and has a gain in growth, compared to the simpler irrigation method. The prototype showed the ability to use and prefer rain, instead of watering. The standard error of the evaporation prediction of this method was approx. 2 percentage points of soil moisture level per day, this is 0.64 mm/m^2 * day^(-1). A gain in growth, according to the other assumed tasks, could not be measured in this experiment and has to be evaluated in a greater scaled experiment. The prototype is based on the previous 'Fachstudie' 'Garden Planner for a Greenhouse - Autonomous Garden Assistant for Smart Homes' by Michael Hersam, Philipp Wonner and Jannis Westermann.