SECOND is building on the results obtained in FIRST (Fundamentals of Intelligent Reliable Robot Systems), action 3274. The context for SECOND stems from the observation that current intelligent autonomous systems are unreliable, and that the basis of their unreliability is their failure to deal properly with uncertainty. This is due to a lack of integration between the individual elements of sensing, planning, modelling and control. FIRST explored the inter-relationships and mutual constraints between the individual elements, and made progress in pair-wise integration of some of these elements; the objective of SECOND is to study and demonstrate the integration of all elements.
Studies have been made of the integration of sensing, planning, modelling and control of intelligent, autonomous robot systems.
During the first year 2 joint demonstrators have been developed. Each demonstrator integrates methodologies and software from different partners.
Demonstrator 1, located at Leuven, concentrates on vision based recognition and localization of a cylindrical workpiece, followed by automatic derivation of stable grasping positions, and force controlled assembly (insertion). The recognition is based on affine invariant contour descriptions. Demonstrator 2, located at Grenoble, concentrates on the integration of vision, automatic planning and execution of a grasping task in a partly known environment. The demonstration combines these 3 activities to detect unknown obstacles, update an a priori computer aided design (CAD) model, and then plan a collision free trajectory for a manipulator in order to go, grasp a known object, and put it on top of another known object. Besides integration and demonstration, new methodologies continue to be developed in the areas of sensing (vision/force, and their combination), planning, and sensor based task execution.
APPROACH AND METHODS
The following areas have been selected for investigation: i) integration of sensing and modelling: the objective is to achieve a two way interaction between sensing and modelling: use sensors to construct or update the model, and use available models to guide the sensing; ii) model-based planning: the objective is to develop task planners which are robust against uncertainties and incomplete models; iii) sensor based planning and execution: the objective is to integrate the results on sensing and planning from (i) and (ii) with execution of navigation, manipulation, and gross and fine motion tasks, in order to achieve intelligent execution: monitoring of the task execution will feature online error detection and autonomous recovery. The consortium intends to develop four demonstrators: two will show the integration of new results obtained in areas (i) and (ii), and a further two showing the results of the overall integration (iii).
SECOND will contribute to more intelligent and reliable planning and execution by autonomous robot systems. A major objective is to develop demonstrators which exhibit such enhanced behaviour. Broadly speaking, exploitation of the results of SECOND follows three routes: Commercial - Research results that are taken up by, for example, industrial collaborators; Explicit - Research results lead directly to further work, such as future research and development goals; Indirect - Research results are infused into the scientific community, either via visiting students and researchers or through publications and meetings.