Detect obstacles from side mono-cameras

[m3d]

Reference docker image ver120side, run f556a76f-ea33-4975-8ccf-5a1a498a95dc (CS1) and 34bcaef9-5bf4-4c24-b0e5-9788b87a99bc (UP3).

[jisa]

f556a76f-ea33-4975-8ccf-5a1a498a95dc has had some networking issues (log/rostopic_395_1628849844460.log):

[rospy.internal][WARNING] 2021-08-13 03:22:59,350: Unknown error initiating TCP/IP socket to 172.16.98.37:37657 (http://172.16.98.37:42489/): Traceback (most recent call last):
  File "/opt/ros/melodic/lib/python2.7/dist-packages/rospy/impl/tcpros_base.py", line 560, in connect
    self.socket.connect((dest_addr, dest_port))
  File "/usr/lib/python2.7/socket.py", line 228, in meth
    return getattr(self._sock,name)(*args)
error: [Errno 111] Connection refused

[rospy.internal][INFO] 2021-08-13 03:22:59,351: topic[/statistics] removing connection to http://172.16.98.37:42489/
[rospy.internal][WARNING] 2021-08-13 03:23:01,140: Unknown error initiating TCP/IP socket to 172.16.98.37:45577 (http://172.16.98.37:41603/): Traceback (most recent call last):
  File "/opt/ros/melodic/lib/python2.7/dist-packages/rospy/impl/tcpros_base.py", line 561, in connect
    self.write_header()
  File "/opt/ros/melodic/lib/python2.7/dist-packages/rospy/impl/tcpros_base.py", line 642, in write_header
    self.stat_bytes += write_ros_handshake_header(sock, protocol.get_header_fields())
  File "/opt/ros/melodic/lib/python2.7/dist-packages/rosgraph/network.py", line 417, in write_ros_handshake_header
    sock.sendall(s)
  File "/usr/lib/python2.7/socket.py", line 228, in meth
    return getattr(self._sock,name)(*args)
error: [Errno 104] Connection reset by peer

The strange part is that this is after the robot started moving, which means we got enough messages to think everything initialized OK.

[m3d]

I am not sure if this was our problem or CloudSim problem :(. I re-run both and the both scored, although I do not know the details (yet).

[jisa]

I am just looking at the logs and trying to convince myself it is an improvement. I haven't fully succeeded with that yet.

[jisa]

In CS1, the robot looks confused. Hard to tell if more or less than without this change. In the Urban world, Freyja did not get past the doorstep into the first room, so not a very informative run.

I have just started a FQ run with one Freyja going left (urban -> tunel -> cave) and one right (cave -> tunel -> cave). Let's see if we get more info from that.

[jisa]

In FQ, it mostly works and the confused robot's behavior comes mostly from other parts of code. The local map looks OK.

However (!), it reports false obstacles all around the robot when in fog. This is a blocker and needs extra care before merging this PR.

[m3d]

However (!), it reports false obstacles all around the robot when in fog. This is a blocker and needs extra care before merging this PR.

@jisa - thanks for discovery and warning!

[m3d]

in NOISH SR tunnel (config B) the right side was closed probably by "ghost" ... but I am not sure if that was from mono-cameras