REDCP - Robotic Exploration and Data Collection Planning
The course consists of lectures and tasks. The successful fulfillment of the course is to finish all the tasks, which are principally tasks
Successful course completion is according to the developed autonomous mobile exploration task and data collection planning with the following scoring targeting ten points.
Implement t1f-exploration with some of the following features.
(1 point) Determination of the frontiers and goal candidates with the closest goal candidate as the next navigational goal.
(1 point) Ensure the planned paths towards the frontier (goal candidate) are within the robot size distance from the obstacles using
obstacle growing.
(1 point) Consider the relatively slow rotation motion of the robot in selecting the next navigational waypoint to avoid oscillation behavior.
(2 points) Select the navigational goal based on some utility measure, such as the expected covered area or mutual information gain.
(2 points) The OccupancyGrid map container dynamically grows with the explored area.
(2 points) Using the non-myopic selection of the next navigational waypoint, such as TSP distance cost, see
lec04.
Data-collection path planning
Implement t2-tspn with some of the following features.
(1 point) Exploiting the non-zero neighborhood of the disk-shaped instances of the TSPN (aka Close Enough TSP) in the unsupervised learning-based method.
(3 points) Improve the performance of the planner using the GSOA algorithm.
(3 points) Solve the TSPN instances using a sampling-based approach and its transformation to the GTSP.
(3 points) Solve the TSPN instances using the decoupled approach with the sequencing part solved as an instance of the Euclidean TSP followed by the determination of the locations of visits to the neighborhoods by location sampling and forward-graph search.