most functionality to splashdwon, find neutral and start mission. short timeouts still in code for testing, will adjust to go directly to sit_idle after splashdown
Dependencies: mbed MODSERIAL FATFileSystem
changelog.txt
- Committer:
- tnhnrl
- Date:
- 2017-11-29
- Revision:
- 24:c7d9b5bf3829
- Child:
- 39:58375ca6b6ff
File content as of revision 24:c7d9b5bf3829:
Starting from Trent's Linear Actuator code from 2017-10-19, these modifications by Dan add an outer loop controller for depth and pitch to command the inner linear actuator loops. Modified 2017-10-20 revA by Dan - added outer loop controller, but it is hanging the mbed. (turns out it was the imu update) Modified 2017-10-22 revA by Dan - outer loop now works with a call to outerloop.update() in main loop(), not with an attached ticker Modified 2017-10-22 revB by Dan - enabled both depth and pitch outer loop controllers - added ability to keyboard reset Modified 2017-10-22 revC by Dan - major update to the IMU library processing to make a state machine that doesn't hang - added lat/lon/alt and GNSS fix information to the IMU library - brought out the pin names into the constructors of IMU, omega, SpiADC Modified 2017-10-22 revD by Dan - everything seems to be working, shy of re-checking on the hardware - Depth sensor call done inside the OuterLoop, but should somehow be set as a callback instead - IMU sensor call done inside the OuterLoop, but should somehow be set as a callback instead Modified 2017-10-23 revA by Dan/Matt - linear actuator hardware works great, limit switches, sensing, etc. - outer loops run, but move the BCE in the wrong direction. - new IMU code doesn't read from the sensor correctly, but doesn't hang up either. - depth sensor worked fine, just needs zero bias adjustment. Modified 2017-10-24 by Troy - added offset to outerloop Modified 2017-10-26 by Dan - brought over a state machine and new keyboard controls ... currently just dumped into main. Modified 2017-10-26 by Matt - new IMU code imported and working well with the hardware. Modified 2017-10-26 revB by Dan - This version has been in the pool. - Get occasional ADC bad string pot batt & piston. On initial sensor reads, see negative positions. But after running FLOAT_BROADCAST, string pot positions are normal. Not sure why. - Repeatedly got stuck in RISE with a +60s timeout. Battery hit end bell and stalled out. - keyboard and state machine are in main, probably shouldn't be, but easier to debug. - Really happy with the logic and flow of the state machine. Timeouts work. - Need to add a means to drive the linear actuators manually ... toggle out stop() in SIT_IDLE. - Need to add keyboard commands to modify the zeroOffset positions. Modified 2017-10-30 by Dan, Trent, Matt, Troy - changed .stop() to .pause() and many .start() to .unpause() ... fixed the negative ADC and lets the PVF's keep running even though the motor isn't moving. - changed exit conditions to use filtered depth from the outer loop. less noisy. Modified 2017-10-31 by Dan, Matt - added oversampling and a tare function to the depth sensor. - updated main and keyboard to include tare in setup() and a keyboard tare option. - looks like the piston isn't zeroing correctly, unclear why, should be unrelated to depth sensor. Modified 2017-10-31 (again) by Dan, Matt, and Troy - POOL TESTED! - Tuned the PID gains with magic. Works to hold mostly level during BCE runs. Dive is awesome. - Neutral won't do what we want ... without large depth P, it won't get to depth. - Dive with just P doesn't overshoot depth at 0.0 deg, but does with -20 deg. Timeout should probably be RISE. - Files depth/pitch updated with tuned defaults. Modified 2017-11-06 by Troy - Added acronyms to print statements (when running this with XBee you don't know what mode the hardware is in because of signal dropping out) - Fixed print error with setpoints by printing out the setpoint variable, not the outerloop getCommand (command is sent every time hardware dives, finds neutral, etc.) - Set the pitchCommand input to 0 on the "find neutral" command, it was sending a non-zero pitch command - Created a class for the StateMachine Modified 2017-11-14 by Troy - Changed tare to void function (functions does not return anything when called) - Added a "subclass" in the cases for the Neutral Finding Sequence (Sinking, Slowly Rise, Check Pitch (and save positions)) Modified 2017-11-20 by Troy - Modified StateMachine class to separate keyboard inputs from FSM - Added Neutral Finding sub-statemachine - Verified both state machines are working with hardware on desktop - Added class to save neutral battery and buoyancy engine positions to neutral.cfg file Modified 2017-11-21 by Troy - Need to check remove ConfigFileIO and place saving functions into config_functions.cpp (rename ConfigFunctions.cpp) - Added multi-dive (and multi-rise) states to FSM (may incorporate single dive states into states) - Need to double-check behavior of sub-FSM but bench-top testing showed correct behavior for transition from sinking, to slowly rise, to check pitch - Need to add a check for the multi-dive sequence file (sequence.cfg) being loaded on the MBED Modified 2017-11-21 by Dan - removed blocker and set up a loop rate timer that runs the state machine and keyboard at 10 Hz. - work inside StateMachine and particularly in the findNeutralSubState. Modified 2017-11-22 by Troy - Placed config_functions into ConfigFileIO and added the ability to write the neutral positions to the depth & pitch files - Streamlined sub-FSM and Find_Neutral state - PID depth and pitch gains have been made into class variables so that they can be resaved to the depth/pitch files - Added the ability to save depth & pitch gains and neutral offsets (configfile writes entire file at once) - Minor formatting fixes - Fixed issue with keyboard function, the keyboard function was continuously active instead of checking pc readable in sit idle - Bench tested neutral finding sequence and multi-dive sequence, so far so good - Question: Why do we want the keyboard and FSM to run together, specifically allowing the FSM call the keyboard? Modified 2017-11-22 by Dan - added ability for substate NEUTRAL_CHECK_PITCH to move the battery toward level, then saves offsets. - added new entry state NEUTRAL_FIRST_PITCH that also moves the battery toward level, but doesn't save. Modified 2017-11-27 by Troy - Fixed exit condition (restart at NEUTRAL_FIRST_PITCH) - Pool tested, needed to run timer for longer than 5 minutes (neutral finding sequence timed out) - IMU fell off tape Modified 2017-11-28 Rev A by Dan/Troy - Modified NEUTRAL_FIRST_PITCH to save the battery offset into the Pitch Outer Loop before beginning neutral finding sequence before sinking - Pitch Outer Loop now runs NEUTRAL_SINKING and NEUTRAL_SLOWLY_RISE states (instead of batt fixed position control) - NEUTRAL_SINKING has been changed to a 5-second timer (vice 10 seconds) - NEUTRAL_SLOWLY_RISE has been changed to move the piston 2 mm each interval (vice 1 mm) Modified 2017-11-28 Rev B by Dan/Troy - Changed pitch rate margin from absolute value < 0.5 deg/sec to 5.0 deg/s - Changed sink timer to move piston 5 mm at a time instead of 10 mm - Find level motion will now run every 10 seconds instead of 5 seconds - Created an integer array to do a quick check of what states Find Neutral sub-FSM ran through (press "c" to see states) - FLOAT_BROADCAST now uses limits of battery and BCE from LinearActuator.getPosition Modified 2017-11-28 Rev C by Dan/Troy - State FIND_NEUTRAL sends pitch offset from file as first commanded - NEUTRAL_FIRST_PITCH and NEUTRAL_CHECK_PITCH are now moving the battery mass with getSetPosition_mm command instead of the getPosition command to allow the sub-FSM time to adequately move the battery mass - Added "else if (substate == NEUTRAL_CHECK_PITCH)" to the NEUTRAL_FIRST_PITCH/NEUTRAL_CHECK_PITCH(NFP/NCP) state to confirm that NEUTRAL_CHECK_PITCH is working - Added "else" to the NFP/NCP state as an error catching mechanism - Modified NFP/NCP state to use pitchLoop().getPosition() instead of imu().getPitch() to use the filtered readings - Added several "isSubStateTimerRunning = false;" statements to the sub-FSM exits to make sure the one-shot actions work in the next state! Definitely a fix. - Reworked the exiting sub-FSM to be a little clearer. - Moved the sub-state logger into all the sub-state transitions ... this should be done in FIND_NEUTRAL on sub-state changes instead (less copy/paste). */