Update RPP readme

nav2_regulated_pure_pursuit_controller/README.md

@@ -30,14 +30,12 @@ Note that a pure pursuit controller is that, it "purely" pursues the path withou
 
 ## Regulated Pure Pursuit Features
 
-We have created a new variation on the pure pursuit algorithm that we dubb the Regulated Pure Pursuit algorithm. We combine the features of the Adaptive Pure Pursuit algorithm with rules around linear velocity with a focus on consumer, industrial, and service robot's needs. We also implement several common-sense safety mechanisms like collision detection and ensuring that commands are kinematically feasible that are missing from all other variants of pure pursuit (for some remarkable reason).
+We have created a new variation on the pure pursuit algorithm that we dubb the Regulated Pure Pursuit algorithm. We combine the features of the Adaptive Pure Pursuit algorithm with rules around linear velocity with a focus on consumer, industrial, and service robot's needs. We also implement several common-sense safety mechanisms like collision detection.
 
 The Regulated Pure Pursuit controller implements active collision detection. We use a parameter to set the maximum allowable time before a potential collision on the current velocity command. Using the current linear and angular velocity, we project forward in time that duration and check for collisions. Intuitively, you may think that collision checking between the robot and the lookahead point seems logical. However, if you're maneuvering in tight spaces, it makes alot of sense to only search forward a given amount of time to give the system a little leeway to get itself out. In confined spaces especially, we want to make sure that we're collision checking a reasonable amount of space for the current action being taken (e.g. if moving at 0.1 m/s, it makes no sense to look 10 meters ahead to the carrot, or 100 seconds into the future). This helps look further at higher speeds / angular rotations and closer with fine, slow motions in constrained environments so it doesn't over report collisions from valid motions near obstacles. If you set the maximum allowable to a large number, it will collision check all the way, but not exceeding, the lookahead point. We visualize the collision checking arc on the `lookahead_arc` topic.
 
 The regulated pure pursuit algorithm also makes use of the common variations on the pure pursuit algorithm. We implement the adaptive pure pursuit's main contribution of having velocity-scaled lookahead point distances. This helps make the controller more stable over a larger range of potential linear velocities. There are parameters for setting the lookahead gain (or lookahead time) and thresholded values for minimum and maximum.
 
-We also implement kinematic speed limits on the linear velocities in operations and angular velocities during pure rotations. This makes sure that the output commands are smooth, safe, and kinematically feasible. This is especially important at the beginning and end of a path tracking task, where you are ramping up to speed and slowing down to the goal.
-
 The final minor improvement we make is slowing on approach to the goal. Knowing that the optimal lookahead distance is `X`, we can take the difference in `X` and the actual distance of the lookahead point found to find the lookahead point error. During operations, the variation in this error should be exceptionally small and won't be triggered. However, at the end of the path, there are no more points at a lookahead distance away from the robot, so it uses the last point on the path. So as the robot approaches a target, its error will grow and the robot's velocity will be reduced proportional to this error until a minimum threshold. This is also tracked by the kinematic speed limits to ensure drivability.
 
 The major improvements that this work implements is the regulations on the linear velocity based on some cost functions.  They were selected to remove long-standing bad behavior within the pure pursuit algorithm. Normal Pure Pursuit has an issue with overshoot and poor handling in particularly high curvature (or extremely rapidly changing curvature) environments. It is commonly known that this will cause the robot to overshoot from the path and potentially collide with the environment. These cost functions in the Regulated Pure Pursuit algorithm were also chosen based on common requirements and needs of mobile robots uses in service, commercial, and industrial use-cases; scaling by curvature creates intuitive behavior of slowing the robot when making sharp turns and slowing when its near a potential collision so that small variations don't clip obstacles. This is also really useful when working in partially observable environments (like turning in and out of aisles / hallways often) so that you slow before a sharp turn into an unknown dynamic environment to be more conservative in case something is in the way immediately requiring a stop.