Collisions, pillars and their octomap, tether task, ground go_to, config folders, student README - #2

On hardware the mocap markers sit on top of the scout and herover, so
their poses report that height, not 0. The control stack adds a body-frame
dock's z to its host's z, so docks on a UGV were lower in the sim than on
hardware.

Ground robots now take pose.z, published on /<name>/pose and in the
world -> <name> TF. It defaults by model to the top of the deck in the
mesh: scout 0.227, herover 0.095, robots without a model 0. The RViz mesh
is shifted down by z so it stays on the floor.

Not run: the pytest suite (docker run interrupted). Scenario parsing was
checked on the host.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

A winch drags an object across the floor; the tether runs through a
pulley on a ground robot and a ring on a Crazyflie. A scenario sets it
up with an optional tether: section (config/tether_demo.yaml). Without
one, nothing extra runs.

- hero_sim/tether.py: ROS-free quasi-static model. Slack leaves the
  object, taut slides it toward the point under the UAV, it is never
  lifted, and it jams against pillars; the winch slips in both cases.
- hero_sim/tether_sim.py: node publishing /winch/* and /tether/* plus
  RViz markers, started by sim.launch.py only for a tether: section.
- config: validated tether: section, documented in README and
  example.yaml; RViz gets a Tether display.
- Tests for the model, config and node; CI smoke tests check tether_sim
  starts with a tether: section and not without one.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…r object

Pillars move from tether.obstacles to a top-level obstacles: list and live
in the same ObstacleMap as the octomap grid (kept as exact circles, also
with map: null). Ground robots, Crazyflies (at any height) and the tether
object all stop at the same obstacles, under the one sim.collision switch.

ground_robot_sim publishes the pillars latched on /hero_sim/obstacles
(one CYLINDER each), drawn by a new Obstacles display in sim.rviz.
tether.obstacles now fails with a pointer to the new key.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Every ground robot serves /<name>/go_to with crazyflie_interfaces/srv/GoTo,
the Crazyflies' type. It reaches goal x, y and yaw in `duration`, stretched
to stay within ground_max_speed and ground_max_yaw_rate (duration 0: as
fast as allowed). omni drives straight while turning; diff turns, drives,
then turns (kinematics.GroundGoTo). relative offsets x, y in the world
frame; goal.z and group_mask are ignored. Any cmd_vel cancels it, and it
stops at the map and pillars like every other move.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Under sim.collision, each ground robot's move is checked against the other
ground robots as circles of their radius, in the same pass as the map and
the pillars (ObstacleMap.move's new `others`). Robots move one after
another each tick against the others' current positions, so they never
overlap; radius 0 opts out. load() now returns an empty map without a map
or pillars, so this works in bare scenarios too.

A blocked go_to now waits where it stopped, its clock paused, and carries
on once the way clears, instead of jumping to where its schedule got to.

config/two_ugvs.yaml: two herovers and the tether demo's pillar, no tether.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

… size

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Configs can sit in config/ or a folder in it: tether/ for the tether task,
mrs/ for the control stack's Crazyflie scenarios, the rest on top
(example, ground-robot-only setups). A bare filename is searched in every
folder (names are unique; two matches are an error), folder/name.yaml is
relative to config/ unless it's an existing path. Same rules in
config.find_config (launch files, nodes) and startup/_setup.sh.

setup.py installs the folders and skips the broken links a
--symlink-install build keeps to moved configs, so existing workspaces
rebuild without a clean. Tests cover both resolvers and run every config
in every folder.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Setup, start/stop, scenarios and folders, moving the robots, obstacles,
the tether task and troubleshooting, without the internals. Every command
in it was run against the sim. Mesh-making notes move to models/README.md.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Adds the sim_env_a / sim_env_a12 experiment scenarios, the
test_perching_full(2) and test_unperch scenarios to config/mrs/, moves
dock_hardware2 there, and makes test_perching_full2 the default. The
example scenario gets its pillar, sim.rviz the current view, and the
README's tether section the task diagram (media/tether_demo.png).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

… or obstacles

With obstacles:, octomap.launch.py writes the pillars as an octomap .bt
(<tmp>/hero_sim/<config>_obstacles.bt, floor to each pillar's height, no
ground plane) and serves it with octomap_server like a sim.map, so RViz
shows the same occupied cells, projected map and topics. obstacles.write_bt
is read_bt's inverse.

Pillars take an optional height (default 2 m) for RViz and the octomap;
collisions still ignore it. A scenario sets sim.map or obstacles:, not
both: obstacles drop the default map, and setting both is an error.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

README: one Maps section for the two kinds of world, a .bt map or
obstacles:, what each looks like in RViz and where the pillar map file
goes, instead of the scattered mentions.

sim.rviz: the occupied-cells displays subscribe Transient Local, as
octomap_server publishes the map once, latched. RViz started after the
octomap now still draws it, so session.yml needs no sleep before it.

example.yaml: the pillar commented out again, so the example uses the
default map.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…only

The example scenario uses its pillar (no map) again. sim.rviz hides the
pillar cylinders by default, which showed through the octomap cells; the
Obstacles display can still be ticked on.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>