This work considers path-planning processes for manipulation tasks such as assembly, maintenance or disassembly in a virtual reality (VR) context. The approach consists in providing a collaborative system associating a user immersed in VR and an automatic path planning process. It is based on semantic, topological and geometric representations of the environment and the planning process is split in two phases: coarse and fine planning. The automatic planner suggests a path to the user and guides him trough a haptic device. The user can escape from the proposed solution if he wants to explore a possible better way. In this case, the interactive system detects the users intention and computes in real-time a new path starting from the users guess. Experiments illustrate the different aspects of the approach: multi-representation of the environment, path planning process, users intent prediction and control sharing.
layer path planning control for the simulation of manipulation tasks : involving semantics and topology. (2019) Robotics and Computer-Integrated Manufacturing, 57. 17-28.
Simulating complex industrial manipulation tasks (e.g., assembly, disassembly and maintenance tasks) under strong geometric constraints in a virtual environment, requires the joint usage of task and path planning, not only to compute a sequence of primitive actions (i.e., a task plan) at task planning level to identify the order to manipulate different objects (e.g., assembly order), but also to generate and validate motions for each of these primitive actions in a virtual environment by computing valid collision-free paths for these actions at path planning level. Although task and path planning have been respectively welly discussed by artificial intelligence and robotic domain, the link between them still remains an open issue, in particular because path planning for a primitive action often uses purely geometric data. This purely geometric path planning suffers from the classical failures (i.e., high-possibility of failure, high processing time and low path relevance) of automated path planning techniques when dealing with complex geometric models. Thus, it can possibly lead to high computational time of the joint task and path planning process and can probably produce a poor implementation of a task plan. Instead of geometric data, involving higher abstraction level information related to a task to be performed in the path planning of a primitive action could lead to a better relevance of simulations. In this work, we propose an ontology-based approach to generate a specific path planning query for a primitive action, using a well-structured taskoriented knowledge model. This specific path planning query aims at obtaining an increased control on the path planning process of the targeted primitive action.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.