<aside> 💡 Clearance-biasing: a method of targeting workspace exploration based on the free-space between obstacles in the workspace. It is particularly applied to Dynamic Region-biased Rapidly-exploring Random Trees (DR-RRT), a skeleton guided RRT.
</aside>
Existing state-of-the-art planners uses the workspace topology for exploration and to speed up the planning process, but improvements can be made to such methods to exploit these topological characteristics and guide exploration based on other features of the environment and capabilities of the robot.
To approach this problem, we implemented the clearance-bias method which uses a skeleton-guided RRT to guide workspace exploration based on the clearance value annotated to the skeleton. Our experiments showed an overall improvement in planning time and less number of collision detection calls
Goal: To show how Clearance-bias works using DR-RRT(a skeleton-guided RRT).
We demonstrate our algorithm on holonomic robots in two 3D environments, MazeTunnel (Figure 1.1 ), and Obstacles (Figure 2.1). We used a Reeb Graph for skeleton guidance (MazeTunnel: Figure 1.2 and 1.3, Obstacles: Figure 2.2 and 2.3 ) and the results are average over 10 random seeds.
Experiment success is measured by the exploration of the RRT towards regions with the specified clearance value. For comparison, we would use Basic RRT and DR-RRT with and without (minimum and maximum) clearance bias and we check the average of the following:
O B S T A C L E S
Figure 1.1: 3D Environment, Obstacles, with Query (start and goal)
Figure 1.2: Reeb Graph of 3D Obstacles with Query (start and goal)
Figure 1.3: Reeb Graph of 3D Obstacles
M A Z E T U N N E L
Figure 2.1: 3D MazeTunnel environment with with Query (start and goal)
Figure 2.2: Reeb Graph of 3D MazeTunnel Environment with Query (start and goal)
Figure 2.3: Reeb Graph of 3D MazeTunnel Environment with Query (start and goal)