Skip to content

TDMA Slotted Aloha, TDMA Manual Transmit, TDMA bug fixes and lots of new CI testing for TDMA

Caileigh Fitzgerald requested to merge dev/slotted_aloha into devel

Breaking Changes

  • TdmaStatus added field slot_duration_seconds.
  • removed the following dynamic reconfigure params from tdma_advanced (also effects tdma_scripted and tdma_slotted_aloha since they use tdma_advanceds' dynamic reconfigure method on top of their own). These params were mostly unused. The only tdma node where you can set slots with dynamic reconfigure is tdma_slotted_aloha (and only aloha_slots).
    • active_slots_
    • nav_slots_
    • comms_slots_
  • tdma_node will no longer run with a configuration that does not allow transmit because of padding and self imposed guard time. If intentional, this forces users to make use of tdma_advanced's software_mute feature (this is the tested and maintained code path for software mute with TDMA). If unintentional, this draws attention to the configuration mistake immediately.
  • tdma_scripted burns it's very first active slot after enabling a test plan (or when it is reloaded or changed with dynamic reconfigure). Burning the first slot means we can depend on the test plan order (defined in the yaml file).
  • changes to: TdmaAdvancedStatus.msg, TdmaScriptedStatus.msg, TdmaSlottedAlohaStatus.msg. Switching from nested message types for super class, to a flat message type. e.g., we_are_active, etc.. is now at the top level of these messages. Previously they were nested in the super classes type.
    • active_slots changed field type from int8 to int8[].
    • nav_slots is now a list type (int8[])
    • comms_slots is now a list type (int8[])
    • time_to_next_active in the context of tdma_slotted_aloha it means time to next own slot or aloha slot (which ever is closest)
  • GetNextPacketData.srv changed min_priority to minimum_priority to match other calls.
  • GetNextPacketData.srv and GetNextQueuedMessage.srv, minimum_priority is type: int8.
  • Setting all instances of minimum_priority and priority to type: int8.

New Features / Improvements

  • New Node: tdma_slotted_aloha uses queue priority to gate traffic sent in the aloha_slots.
  • tdma_slotted_aloha adds priority gating and a special kind of slot called aloha_slots
    • aloha_slot_priority condition: msg.priority >= aloha_slot_priority, msg can be transmitted during aloha_slot. Otherwise this msg can only be transmitted in our own slot.
    • aloha_slots are slots that have a minimum priority and are typically shared with other platforms.
    • aloha_slots allow traffic with high enough priority to be sent during this "group" slot. The traffic can also be sent in the dedicated slot. But only traffic with priority >= aloha_slot_priority will be transmitted during the aloha_slots
    • aloha_slots are allowed to:
      • be outside of active_slots completely (intended use)
      • overlap with active_slots (if active_slots are subset of aloha_slots this is only allowed when allow_aloha_slots_only=True)
      • be a list of slots
  • tdma_slotted_aloha rosparam allow_aloha_slots_only by default False. When True, it allows a configuration with only aloha_slots and no slots of our own i.e., active_slots is a subset of aloha_slots.
  • New DynamicReconfigure for tdma_scripted (overlaid on tdma_advanced DynamicReconfigure)
    • scripted_test_plan_enabled, default True if scripted_test_plan yaml is passed
    • scripted_test_plan_file, filename for test plan yaml
    • parse_test_plan_file_again, toggle to reload test plan
  • New DynamicReconfigure for tdma_slotted_aloha (overlaid on tdma_advanced DynamicReconfigure)
    • aloha_slot_priority
    • aloha_slots_, trailing '_' for this dynamic reconfigure param. This allows changing the aloha_slots after launch.
  • message_queue_node can now gate on a minimum priority. To get tdma_slotted_aloha working, message_queue_node needed the plumbing for minimum priority. It now has it but tdma_slotted_aloha is the only node that uses it currently (TODO: add priority slots to tdma_advanced? )

!117 (merged)

  • Added system in tdma super class (manual_transmit_queue) for sending modem nmea messages manually to be queued by tdma and sent in active slot (in place of next packet)
  • new rosparam ~maxlen_manual_queue default: 1
  • new rostopic tdma/manual_transmit_status type diagnostic_msgs/DiagnosticStatus. It shows the contents of all manual transmit queues on insert and on transmit
  • User publishes to nmea_to_modem topic inside of tdma/ namespace and tdma will hold up to ~maxlen_manual_queue message(s) and publish them (one at a time) instead of the next outgoing packet during our active_slot.
  • If using a value > 1 for maxlen_manual_queue, tdma will work through the manual_transmit_queue until it's empty. It will treat each nmea sentence like a packet with the same packet_length_seconds and guard_time_seconds. It will only transmit one message from this queue at a time so large values for maxlen_manual_queue are not recommended as tdma needs to work through the whole queue before getting back to packets.

!118 (merged)

  • New ci test node: test_tdma_slotted_aloha.py
  • test cases for tdma_slotted_aloha
    • test_high_pass()
      • publishes 3 msgs (on low priority topic) to a node with only aloha slots so it must timeout to pass
    • test_low_pass()
      • publishes 1 msg on high priority topic, confirms it's next rx.
    • test_collision()
      • makes sure we're not active and at least 1 slot away from shared aloha_slot
      • creates subscriber for /modem0/from_acomms/test_msg_aloha and /modem1/from_acomms/test_msg_aloha
      • publishes 1 msg on modem0 high priority topic (/modem0/test_msg_aloha), priority: 80
      • publishes 1 msg on modem1 high priority topic (/modem1/test_msg_aloha), priority: 80
      • when callback set for subscribers has set modem0_last and modem1_last, we check that the time delta and values indicate a collision (which we expect)
    • test_low_priority()
      • publishes 3 msgs on low priority topic
      • makes sure we recv all of them on rx side

!121 (merged)

  • tdma_scripted_node now sets rosparam server and attempts to delete rosparam server so re-launch does not use old cached value
  • tdma_slotted_aloha now has ~aloha_slots = [] by default for test cases
  • test_ros_acomms.py removed always_send_test_data case as test_tdma_extended.py now tests this case
  • test_tdma_extended.py test cases added:
    • manual_transmit_queue
    • manual_transmit_queue, with always_send_test_data
    • software_mute
    • always_send_test_data
    • always_send_test_data, with miniframe data
    • always_send_test_data, with dataframe data
  • test_tdma_scripted.py cases added:
    • test_cycling_rates_one_per_slot
      • passing criteria, cycle rates from test plan one packet per slot, rates in order from test plan
      • [1,3,5,1,3,5,1,3,5], [5,1,3,5,1,3,5,1,3], [3,5,1,3,5,1,3,1,5] are valid
    • test_cycling_rates_three_per_slot
      • passing criteria, cycle rates from test plan 3 packets per slot, so ALL 3 rates from test plan in test plan order per slot twice.
      • ONLY [1,3,5,1,3,5] is valid
    • test_cycling_rates_six_per_slot
      • passing criteria, cycle rates from test plan 6 packets per slot, so ALL 3 rates from test plan in test plan order twice per slot. Then confirm this pattern in the next active slot
      • ONLY [1,3,5,1,3,5,1,3,5,1,3,5] is valid

Bugfixes

!121 (merged), !120 (merged)

  • pipeline now uses tag: pipeline_$CI_PIPELINE_ID instead of latest
    • e.g., ros_acomms-tests:latest is now ros_acomms-tests:pipeline_$CI_PIPELINE_ID

!122 (merged)

  • tdma_scripted
    • #40 (closed) tdma_scripted will burn it's very first active slot IF there is an active test plan. This assures the rates/buffer sizes are cycled according to the test plan from the very first packet in the slot.
    • #40 (closed) tdma_scripted sets the parameters for the first transmit from the test plan rather than using values passed in launch for first TX
    • #40 (closed) when the test plan is toggled on from an off state (or it's re-read while running via dynamic reconf) the next active slot will be burned (in case the user toggled during an active slot. This keeps the code simple for handling this case and we won't activate a test plan mid slot)

477f2dd0

  • tdma_node, tdma_slotted_aloha_node added more informative error messages.
  • active_slots was never allowed to be an empty list but, now the error is more clear and the condition is checked more explicitly.
    • if active_slots was previously set, tdma will keep the prior active_slots rather than raise an Exception (general M.O. only raise exc right away but once we have launched, do everything in our power to keep system up)
  • tdma_slotted_aloha_nodeadded new rosparam allow_aloha_slots_only. By default this is False and should only be true in a special case (or for CI testing)
    • when allow_aloha_slots_only is True, the user is allowed to have a tdma configuration with only aloha slots and none of it's own.
    • when allow_aloha_slots_only is False, the user cannot configure tdma to only have aloha slots. It will raise an exception right away.

Notes

  • The manual_transmit_queue system is for sending nmea messages to the modem that make the modem transmit. There is only one topic logically compatible right now and it's the nmea_to_modem in the modem namespace tdma is currently servicing.
  • in the future there may be other topics or ways of using this manual transmit system. For now, it's just a simple interface for nmea commands that you want to send in band with tdma
  • Example for manual transmit:
    1. user publishes to $CCACM,1,1,1,CCCFG;pwramp.txlevel;0 to /modem0/tdma/nmea_to_modem
    2. $CCACM,1,1,1,CCCFG;pwramp.txlevel;0 is now queued to transmit next
    3. when tdma_status.we_are_active and ready to send a packet, we instead publish $CCACM,1,1,1,CCCFG;pwramp.txlevel;0 to /modem0/nmea_to_modem which in this case will make the modem transmit this CCACM.
Edited by Caileigh Fitzgerald

Merge request reports