Motion Planning is widely acknowledged as a fundamental problem of robotics. Due to the continuous efforts of the scientific community, various algorithmic families emerged that have different strengths and weaknesses. Finding a suitable motion planning program is often not trivial for real-world problems, as various domain-specific factors must be considered. An obvious example is a potential trade-off between path length, computation time, and resource constraints. We propose a technique to systematically explore the space of suitable programs, aiming to find Pareto optimal algorithm configurations. Our approach makes use of Combinatory Logic Synthesis to perform component-based software composition. Software components are injected with domain-knowledge, effectively restricting the solution space of synthesizable programs. We synthesize sample-based global planning programs that make use of the Open Motion Planning Library (OMPL) and evaluate the produced programs to yield numeric result vectors. These steps are encapsulated in a black-box function which is used with a multi-objective optimization tool (Hypermapper) to yield an automatic, learning-based search procedure for a given feature space. We validate our approach with a series of experiments that demonstrate the extensibility and transferability of our methodology regarding different robotic systems and planning instances.
Design Space Exploration for Sampling-Based Motion Planning Programs with Combinatory Logic Synthesis
Springer Proceedings in Advanced Robotics
International Workshop on the Algorithmic Foundations of Robotics ; 2022 ; , MD, USA June 22, 2022 - June 24, 2022
2022-12-15
16 pages
Article/Chapter (Book)
Electronic Resource
English
Synthesis of combinatory cam to generate a periodic spherical motion of small travel space
British Library Online Contents | 1998
|Sampling-based Motion Planning with Temporal Logic Missions and Spatial Preferences
BASE | 2020
|