The purpose of this thesis is to plan and execute trajectories that minimize "the risk" for a serial manipulator operating in a human-robot collaboration context.
The first step is, therefore, to formalize the risk concept, identify which risks we may incur, and create a metric to quantify them.
We then continue by presenting some optimization strategies in the offline trajectory planning phase that lead directly or indirectly to the minimization of the identified risks.
Finally, we simulated the execution of the found trajectories during their actual online execution, developing an interface for online evaluation of the risks indexes and a reactive strategy for online modification of the reference trajectories found during the offline optimization.