Reflecting on the material covered in Chapters 2 through 7, what would you say is missing from the path planning discussed so far? How effective are the techniques discussed?
I think the biggest method missing is SLAM (simultaneous localization and mapping). So far, we have been able to navigate in a workspace that we know either nothing about or everything about. However, we do not yet have the tools to learn about our environment and make better decisions based on that new data.
I would also enjoy discussing path planning in a fully three-dimensional system, such as might be used for a spacecraft or satellite.
The techniques that we have discussed so far seem to be very effective. Each method has its strengths and weaknesses; however, taken as a whole, the path planners seem like they can handle any reasonably path planning problem.
One thing that has not been discussed is the efficiency of path planners. Ten years ago, this probably would have been a problem. However, current computers (even desktop computers) are powerful enough to quickly compute a path (or lack of path) for any reasonable workspace. Though, what defines a “reasonable workspace?” As path planning is applied to a more varied array of situations, such as biological devices, one can anticipate that the complexities of the paths will also increase. As long as computational power continues to increase, this won’t be a problem. However, many industry analysts believe that we are quickly reaching the maximum transistor density for CPUs. Are the currently path planning techniques efficient enough, or do we need to investigate more efficient ways to determine paths (perhaps parallel processing)?