Autonomous grasping is a key requisite for the autonomy of robots.
However, grasping of unknown objects in domestic environments is difficult due to the presence of unpredictability and clutter.
In this paper, a novel algorithm capable of finding an unobstructed grasping pose on unknown regular object shapes in cluttered environments is proposed. At first, an axis aligned bounding box is fitted around the target object point cluster, proposing a limited set of possible grasps.
A Support Vector Machine is trained to take the decision of the preferred grasp via supervised Machine Learning. Such decision is based on the distribution of obstacles around the target object, inferred by sampling the OctoMap of the scene.
Four different versions of the algorithm have been trained, tested and validated with simulation and real data, with known and unknown target object shapes, in more than 120 different scenarios. The best performing algorithm version returned an unobstructed grasping position in more than 90\% of the cases.
Thanks to the the use of the OctoMap representation, significant results are achieved with a limited set of training examples, with a 10 to 10.000 fold reduction of training examples amount with respect to comparable state-of-the-art algorithms. The algorithm also proved itself capable to generalise to novel object shapes and from simulation to real data.

A major societal challenge in occidental countries nowadays is the continuously increasing ageing population. Due to this ageing society, solutions are needed to alleviate the limited care personnel and the ever increasing healthcare costs. To this end, service robots are proposed to meet this demand by assisting the elderly with their daily activities and providing them with a more independent life.

When people are present, new behavioral concepts are necessary to assure acceptance of robots as daily assistants. Robot motion must take safety, as well as human comfort into account. This is the goal of Human-Aware Navigation (HAN). Most current HAN approaches focus on maintaining appropriate interaction spatial distances between the robot and the human, aiming at respecting their proximity space. However, these approaches are effective only when the human lies in the close vicinity of the robot. Predictive methods, on the other hand, aim at foreseeing how humans will move in the environment and plan a navigational behavior based on this anticipated motion. Though, these methods usually fail to produce socially acceptable robot motion as they do not account for proximity constraints.

This thesis is concerned with the implementation of a Human-Aware navigation framework that incorporates the proximity space around a human as well as human intention, as layers of a layered 2D costmap architecture, used for navigation. The proximity space is modeled as a Gaussian cost function around the person. Human intention is defined by the probability that a person will move to a specific destination within the environment and is used as a
means of human motion prediction model. In order to infer the path that the human will follow to reach the destination, a simplistic path planner is implemented. Then the intention cost model is assigned along this expected path.

The proposed framework is benchmarked against two current state-of-the-art navigation methods. The contestants are: the navigation method used in the ubiquitous Robot Operating System (ROS) and a social navigation method that accounts only for the proximity space around humans. The goal of the thesis is to show that the incorporation of human intention in the robot path planning process is able to produce friendlier motion and increase human comfort. Simulated experiments have been conducted and metrics have been defined
to evaluate the methods in terms of human comfort and navigation performance. Results demonstrate that the proposed navigation framework outperformed the other two methods, proving to be able to produce friendlier robot motion while exhibiting similar navigation performance.