GitLab

The class PotentialFieldPedestrianCompactSoftshell implements the soft shell repulsive pedestrian potential described in sivers-2016b (p. 46, eq. 4.1). There are differences between the manuscript and the implementation. It is not clear if the implementation should be used or adapted to the manuscript. Also, it was observed that the manuscript is not 100% consistent.

Points of irritation in the manuscript:

• \delta_{int} and \delta_{per} are both termed "radius" in the manuscript even though they should not contain the pedestrians radius. Therefore, they are not a radius around the pedestrians position.

• On page 34 of the manuscript, Halls paper "The Hidden Dimension" (1966) is summarized saying that there are 4 zones. The one closest around a pedestrian is the close phase of the intimate distance which is 15 cm (6 inches). Since a pedestrian has a radius of about 20 cm, this distance has to be added to the pedestrians radius, it does not contain it.
• in eq. 4.2 the bounds seem to mistaken for cases 2 and 3
  • case 2: (2 r_p <= d_j(x) <= \delta_{int} + 2 r_p
  • case 3: (\delta_{int} + 2 r_p <= d_j(x) < \delta_{per} + 2 r_p

It is not clear if in eq 4.1 for the definition of \psi_2 and \psi_1 r_p should be replaced by 2 r_p

Differences to the code

• In the code, in functions \psi_1 and \psi_2 as well as cases 2 and 3 of eq. 4.2 radii (which is 2r_p) is used where r_p is used in the manuscript
• eq 4.1 \psi_3^j(x): the factor within the exponential function is 2 in the manuscript but 4 in the code.
• [minor] The parameter personalSpacePower is not mentioned in the manuscript. (It's default value is 1, so there is no difference at this point using the defaults)

At the moment, as defaults, the values from sivers-2016b are used. It is not clear whether the parameters from the manuscript work well with the differences in implementation.

Possibly related with xref #90 (closed)