Background
Path planning is a crucial component of mobile robot navigation and has been extensively researched throughout the history of robotics. Advanced planning strategies, as represented in the Maverick path planner pioneered by SwRI, combine grid-based and sampling-based planners at various levels to generate a fast, high-performance path planner. This research extends SwRI’s path planning architecture (Maverick) by implementing a path planning strategy that improves planning speed, considers a larger plan area, and uses multiple sources of real-world a priori data.
Approach
Our research objective addressed limitations of the Maverick architecture by extending the planning window to greater than 1 km, leveraging a priori data to inform the system across large planning distances, and implementing the two different types of data to plan over. We implemented a multi-level hierarchical path planner: a higher-level planner with low resolution across a large planning area, and a lower-level planner using a higher resolution within a smaller planning area. We then linked the datasets using latitude and longitude and formulated a way to communicate between the two levels allowing for faster path planner operation.
Accomplishments
Implementing the hierarchical planner resulted in improved operation speeds, especially noticed in off-road environments. A path spanning several hundred meters at a resolution of 3 m/pixel was generated in 1.6 seconds, which would have taken over 5 seconds to generate without the hierarchical planner. We found that reducing the search area and incorporating the enhanced road network allowed the lower-level planner to generate a path in roughly 0.76 seconds.