This paper intends to describe the current status of our group in trying to make a humanoid robot autonomously build an internal representation of an object, and later on to find it in an unknown environment. This problem is named "treasure hunting". In both cases, the main difficulty is to be able to find the next best position of the vision sensor in order to realize the behavior while taking care of the robots limitation. We briefly describe the models and the processes we are currently investigating in building this overall behavior. Along the description we stress the current key problems faced while trying to solve this problem.