Progress Review 10

December 9, 2022 rachelzh

Off

MRSLAM

Added check before correcting odometry with laser scan
Added mutex’s to make algorithm multi-threading safe
Adjusted raytrace behavior to avoid generating free-space due to sensor limitations
Still investigating erroneous scans being added to map

Added repulsive forces when robots are within a repulsion zone
- using an inverse cube of distance
When robots are within a repulsion zone, robots are assigned a repulsive velocity.
Testing it with the robots introduced a lot of bugs because this is another velocity component added on the robot along with the velocity obstacle.
Tuning the parameters for velocity obstacle

Implemented search behavior in robot after it reaches a goal based on the type of goal specified.
Robot rotates in place and tries to detect a victim.
Not using it right now because we are discussing other ideas.

Improvements to RRT Frontier Exploration
- Filtering frontiers based on information gain
- Tuning hyperparameters: RRT branch length, sampling frequency, info gain threshold
- Making frontiers more stable: saving previous frontier centroids and adding to current frontier list to be reclustered + filtered
Integrating frontier tasks with task allocator
- Greedy task allocation based on distance
- Greedy allocation sends robots to not globally optimal tasks
- Work on market-based task allocation
- Use A* path costs instead of euclidean