The retail industry is good for a vast employment of over 300.000 persons, but has unfortunately been in decline over the last few years. Automation is the key to more productivity growth and to regain our position within the international economy, while also providing benefits in food safety. To this end, we need general purpose cobots which are able to quickly learn how to handle a variety of goods in complex dynamical environments dominated by human presence. Learning in robotics is mainly vision-driven because of the ease-of-access to cameras compared to other sensors, but this comes at the cost of elaborate image processing algorithms and long training times. It is our belief that the use of tactile input and its fusion with other sensory information is crucial to learn complex robot manipulation tasks in an efficient way. We propose to develop a versatile and modular sensing platform with stretchable circuit technology for the instrumentation of everyday objects, so that the data required for such a task can be conveniently extracted from the object under manipulation. The instrumentation will lead to faster learning of control policy without relying on complex and demanding learning pipelines.