| Commit message (Collapse) | Author | Age |
| |
|
|
|
|
|
|
|
|
|
|
|
| |
At the start of simulation, the vehicle is actually hovering slightly
above the platform; this prevents the simulator from erroneously
inferring initial conditions under which the car intersects the
platform. This does, however, mean that the car has to fall a couple
inches onto the platform when the simulation begins, so if you record
the absolute speed of the car, you’ll get a weird spike at the start of
the simulation. This commit fixes that issue by confining speed
measurements to the xy plane.
|
|
|
|
|
|
|
|
| |
The sensor was having difficulty reading far-away objects because its
readings would collide with the ground.
Reported-By: Constantin Berzan <cberzan@gmail.com>
Reported-Reference: <CALYHk4UkCX4og+xULtrokM86+tH_=necKG1imH0a5HH_nP0Nuw@mail.gmail.com>
|
|
|
|
|
|
|
| |
In commit 50969a2, I changed the API I used to get distance data from
the vision sensors. Somewhat disappointingly, the new API returns data
in the reverse order from the old API; this commit corrects the order in
Lua before passing it off to C++.
|
|
|
|
|
|
|
|
|
| |
The documentation string (and README documentation) for
'simExtAutomobileRequestNoise' specified that distance noise was passed
before intensity noise. In actuality, the reverse is true.
This is an API-compatible change; only the documentation and example
have been updated.
|
|
|
|
|
|
|
|
|
|
|
| |
V-REP models vision sensors’ fields of vision as frusta – i.e., V-REP
assumes the camera’s field of view terminates in a plane whose points
are generally not equidistant from the aperture. Unfortunately, most
humans think of a camera’s field of view terminating in a paraboloid.
Thus, to the human brain, the raw depth buffer V-REP returns appears
distorted. This commit fixes that by getting depth data from a filter,
rather than the raw depth buffer; this causes V-REP to perform
correction on the data.
|
|
|