Cognitive Programming Interface : from Task Level Programming to Coherent Task Level Programming
Raphaël Gerin, Julie Dumora, Olivier David, Baptiste Gradoussoff · 2024
One of the main barriers to the uptake of robotics in industry is the cost of integration. For small and medium-sized enterprises with small production runs and frequent changes to their robotic cells, the cost of integration and programming is too high an investment. To tackle this problem, a number of techniques exist in the literature to perform Task-Level-Programming.However most provide a set of tasks without defining, or in a very limited way, their context of use and their effects on the robot’s environment. As a result, users need to master a wide range of skills and the context in which they can be used.In this paper, we propose a different approach considering object’s interfaces level instead of object level, to allow the user to perform Coherent Task Level Programming where available skills/tasks are context dependent. We describe a Cognitive Programming Interface based on a multimodal planner relying on semantic reasoning that allows two programming modalities: programming by selecting a feasible skill according to the context, or programming by demonstration. These two methods are complementary and can be used in the same programming sequence if required. They rely on an ontological skill definition that replaces the classical Planning Domain Definition Language approach. Consequently, we propose a new method to recognize demonstrated skills from demonstration.