Some areas do not fit into the standard criteria for evaluation. These are just as important, but have not been needed, or heavily researched yet. This section is intended as a catch-all for the intangeable, or so far in-formal, factors in Path Planning.
Errors are inherent in every sort of system. Robotics is no exception, but in robotics errors can be very costly and dangerous. Thus there is a definite need to deal with error trapping, error reduction, error recovery, error simulation, and error prediction. All of these errors can arise in any part of the system, and in the domain of path planning they can have a drastic effect. An error in world modelling can result in a faulty path that may be less than optimal, or at worst cause a collision. In a feedback system, the errors could indicate fictitious collisions, or empty space when it is actually occupied. Thus all aspects of robotic path planning should eventually encorporate the ability to deal with unexpected events, and eliminate errors.
For some planning methods this is not applicable, but for other methods this becomes a critical factor. The real resolution of the environment is governed by accuracy and repeatability. The accuracy is the variance between the commanded and actual manipulator position. Repeatability is the variation in position that occurs when a position in space is re-found by the manipulator. Both of these factors are subject to possible errors which arise from sensor data. These errors can add a compounded error into the robot system, and thus they should be introduced into the path planning routines. Neither of these methods is considered by many of the authors of papers.
The resolution will have the most profound impact when the environment is represented in an array, oct-tree, Quad-tree, integers, or via sensors. Thus good research into resolution for a robotic system will allow auto scaling of Path Planning methods to be used, to match the environment, and provide realistic accuracies. One benefit of this approach would be running methods quickly at a poor resolution, for fast solutions, and at a high resolution for accurate solutions, in tight corners.
It was suggested by J.H.Graham [1984] that probabilistic methods for path planning help by allowing looser tolerances of parts involved, and allowing motions not normally considered.
To help overcome the resolution problem there are a couple of tricks. When modelling in three dimensions the surfaces may be multifaceted, and hard to calculate quickly. This aspect of calculation may be sped up by using simple approximating surfaces not near the start or stop point, where low resolution is acceptable. It is also possible to use gross motionapproximations when travelling through this space. Finally, if the path planner uses collision avoidance, errors in accuracy will be insignifigant, because the avoidance of objects may be greater than the resolution..