Tag Archives: IR Follower

Charging Station System Integration – New Sunshade Testing

Posted 19 May 2017

While working with John Jenkins on the modulated IR beam idea, I decided to run some tests with the current 4-detector design, to see how the new sunshade with center divider was affected by ambient IR on a bright sunny day.  So, I disabled Wall-E2’s motors and then placed it at several different critical spots in the entry hallway.  At each location I used my little IR test generator to mark the beginning and the end of the test for that location, and then moved on to the  next one.  The locations are shown in the following photos, in the order that the tests were run.

Position 1: Near where Wall-E2 transitions from wall-tracking to IR beam homing

Position 2: This is where Wall-E2 has been winding up when it homes on the outside sunlight instead of the IR beam

Position 3: Here Wall-E2 should be firmly fixated on the IR beam

These results, combined with my earlier IR response tests with a single phototransistor are encouraging, because it is clear that at least in this case, Wall-E2 should have no difficulty discriminating between the ambient IR and the charging station IR beam.

I ran some homing tests, which Wall-E2 handled with ease; unfortunately God had already turned the lights out on this side of the world, so the tests weren’t in the presence of daylight IR interference.  I’ll do some more real-world discrimination testing tomorrow, and I am hopeful that the new sunshade-with-divider version will be successful, at least for this part of the house.

Stay tuned,

Frank

 

 

IR Light Follower for Wall-E2, Part XI – Center Divider Investigation

Posted 16 May 2017

One of the suggestions John Jenkins made during his visit, in addition to the idea of modulating the IR beam to suppress ‘flooding’ from ambient IR sources, was the idea of placing an opaque divider between the left and right halves of the detector array.  He thought that I might even be able to reduced the detector array from four to two phototransistors and still get good homing performance, assuming that each of the two detectors had sufficiently wide beamwidths to accommodate off-axis IR beam intercepts.  The current detectors have a +/- 12º beamwidth, but are arrayed in such a way as to provide well over 60º aggregate coverage.  To do the same thing with just two detectors would require parts with considerably wider beamwidths.  the TAOS TSL267 (another one of JJ’s suggestions) has an approximately +/-30º beamwidth at the half-response points, so they seem almost ideally suited for this application.  As a major added bonus, the TAOS parts feature a photo-diode integrated with an op-amp to address the dynamic range issue mentioned by John.  The diode operates in its linear range, and the op-amp amplifies the IR signal to useful levels.  The only fly in the ointment is that the op-amp gain isn’t adjustable, and its output limits at fairly low light levels – bummer!

In order to investigate the opaque divider idea, I decided to run some bench angular response tests using my robot’s 4-detector array with and without a center divider.  I printed out a copy of the compass rose graphic I had hanging around from my magnetometer project (one of my more spectacular failures) and set up a bench test with Wall-E2 and my little IR test source, as shown below.

Angular response test setup

Closeup of the ‘sunshade’ cowling (black rectangular opening) around the 4-detector array

Closeup of sunshade with divider installed

The results without the center divider were about as expected, with about +/- 45º coverage, as shown in the Excel plot below

Response vs angle for 4-detector array, without center divider

+/- 60 deg response vs angle for 4-detector array, without center divider

With the divider, the response appears to be about the same (note here that the ‘with divider’ response is flipped left/right from the ‘no divider’ case – oops!)

+/- 30 deg response vs angle for 4-detector array, without center divider

Note in the above ‘detail’ view that the response curves for the two center detectors (DET2 & DET3) seem to be very symmetric about the 0º point, as expected.  What this also shows is that just these two detectors could probably be used for homing, if off-axis beam detection weren’t a consideration.

+/- 30 deg response vs angle for 4-detector array, with center divider

The ‘with divider’ plot above shows a significant difference from the ‘no divider’ plot, right at the center.  In the ‘no divider case, the DET1 & DET2 responses are very nearly the same, but in the ‘with divider’ case they are significantly different, and almost perfectly anti-symmetric about the center.  This should allow more precise detection of small left/right deviations from the beam centerline, and therefore more precise homing.

Stay tuned!

Frank

 

 

IR Phototransistor Sensitivity/Dynamic Range Study

Posted 05/14/17

In a previous post, I mentioned that John Jenkins, mentor and old friend, had some ideas regarding my Wall-E2 robot’s problems with homing in on an  IR beam in the presence of ambient IR sources like overhead incandescent lighting and/or sunlight streaming in through windows and doors.   John pointed out that this was really a system dynamic range issue, and it was likely that, as currently configured, the IR phototransistors were running out of dynamic range (saturating) well before the 10-bit A/D’s on the AtMega SBC.  In order to get the detector sensitivity up to the  point where Wall-E2 could ‘see’ the IR beam far enough (1.5-2m) away to avoid hanging up on the lead-in rails (see this post and this post for details), I had to use a very high value collector resistor (330K), which reduces the dynamic range significantly.

In a subsequent email conversation, John suggested that the proper way to handle this problem was to reduce the collector resistor to the point where the detector doesn’t saturate under the worst case ambient IR conditions, and then add amplification as necessary after the detector stage to get the required sensitivity.  John’s point was that as long as the detector response is relatively linear (i.e., it’s not saturated), then there shouldn’t be any loss of information through the stage, so a later linear amplification stage will allow the desired IR signal to be detected/processed even in the presence of interference.  However, if the detector stage saturates due to interference, then it’s basically ‘game-over’ in terms of the ability to later pull the desired signal out of the noise.

This wasn’t exactly what I wanted to hear, as adding the required post-detector amplification stage wasn’t going to be particularly easy – there’s not much left in the way of free real-estate on Wall-E2’s main platform (there’s plenty of room on the second level, but putting stuff there requires inter-level cabling and is a major PITA).  As I was thinking this, I had a flashback to similar conversations with John from 40-45 years ago, when we were both design engineers in a USG design lab; the usual outcome of such ‘conversations’ (to the uninitiated these might be mistaken for shouting matches) was that my circuit got ‘simplified’ by the addition of 50% more parts – but could then withstand a nuclear attack in the middle of a snowstorm at the South Pole!

Anyway, back in the present, John suggested I run some experiments designed to determine the ratio of IR field intensity to collector current for a the phototransistors I was using, so the proper collector resistor value to be computed to utilize the available detector dynamic range without driving it into saturation.  His suggestion was to not  use a collector resistor at all, but to simply connect the collector to +VDC through a current meter, and then expose the detector to worst-case ambient conditions.  I didn’t particularly like this idea, because I only have a manual non-recording multimeter, and I wanted to record the data for later analysis.  So, I decided to program up one of my small SBC’s with analog input capability (in this case, a Pololu Wixel), and set up to measure the voltage drop across a 10 Ω precision resistor in the collector circuit.  The hardware setup is shown in the following photo

Hardware setup for IR Phototransistor Response Experiment

Before and after collecting ‘field’ data, I made a collection run on the bench, using my IR LED test box, with the following results.  As can be seen in the plot, the maximum voltage drop across the 10Ω resistor was approximately 200mV, or about 20mA.

IR detector response bench test. Nose-to-nose with my IR test source

Bench test with IR LED test generator

 

To collect the ‘field’ data, I placed the SBC in different locations around the house, and recorded A/D values.  For these experiments I was using an example Wixel app that printed out the A/D values for all 6 analog inputs, scaled such that the maximum reading was equivalent to the SBC board voltage (3.3V) in millivolts.  In other words, the maximum A/D reading was approximately 3375, or about 3 mV/bit

Location 1: Looking out glass doors on south side of house

Location 2: Direct sunlight through window on south side of house

Location 3: Kitchen floor, looking toward entrance hall and atrium

Location 4: Entry hallway, looking toward atrium. This is the usual starting location for charging station homing tests

Location 5: Entry hallway near door to atrium

These results were not at all what I expected.  Either there is something wrong with the experimental setup, or the ambient light IR field intensity isn’t anywhere near as strong as expected.  If the data can be believed, the ambient conditions are significantly weaker than the bench-test conditions, which were basically nose-to-nose with my IR test LED.  So, if my planned double-check of the hardware doesn’t find any problems, then I’ll change the resistor value from 10Ω to 100Ω and repeat the tests.

17 May 2017 Update

After thoroughly reviewing the hardware setup, I concluded that everything was working properly, but that the 10Ω load resistor simply didn’t provide sufficient drop for reliable measurements.  So, I changed out the 10Ω resistor for 100Ω, and repeated the tests, with the following changes:

  • As before, I started the run by performing a ‘nose-to-nose’ test to verify proper operation, but this time I measured the detector current using a multimeter. The  result was about 15mA maximum current.
  • At each of the 5 locations, I started with a short nose-to-nose section (not shown in the plots) to make sure the detector was operating properly
  • At location 2 (the direct sunlight location), I physically oriented the detector for maximum response.
  • After location 5, I placed the detector on the charging station’s IR beam reflector boresight at about 0.5m distance, and physically oriented the detector for max response.
  • I calculated the detector current for each reading, and plotted that rather than the raw reading.  To calculate the detector current for each measurement, I subtracted the detector reading from the corresponding 3.3VDC supply voltage measurement and divided by 100.

The results of these tests are shown in the plots below:

 

These results are pretty interesting.  As I’m sure John would point out, the direct sunlight response of about 0.2 mA is about 20 times the value required to saturate the phototransistor with the current 330KΩ.  No wonder I was having interference problems – oops!

Stay tuned!

Frank

 

 

 

 

 

 

Charging Station Design, Part XIV – Modulated IR Beam Study

Posted May 10, 2017

A couple of weeks ago my old friend and mentor John Jenkins was visiting the area and stopped in for a couple of days.  Naturally I had to show him my Wall-E2 robot, complete with autonomous charging capability.  Just as naturally (if you believe in Murphy’s law), Wall-E2 refused to cooperate.  Instead of homing in on the charging station, it blissfully homed in on the sunlight streaming in through the atrium, bypassing the charging station entirely!

After John got through laughing, he mentioned that he knew of some other IR-following robotics projects where the designers used a modulated IR beam to allow the robot to discriminate between general IR background noise and the intended signal.  So, I decided to start a feasibility study to see if I could incorporate some sort of modulation into the design of the charging station, without completely overtaxing either the charging station, or the robot, or both!

I started with the idea that I could easily add square wave modulation to the IR beam by simply switching the LED off & on at some reasonable rate (say, 500Hz).  The problem occurs on the other end (the robot), as it will somehow have to a) detect the square wave signal, and b) still have the ability to home in on the modulated signal – i.e. still be able to determine signal variation across the four photo-transistor array.

To investigate,  I created a square wave generator using an Adafruit Trinket, and a square wave receiver using an Arduino Uno, as shown in the following photo

Adafruit Trinket SBC used for generating the square wave signal, and an Arduino UNO used as the receiver

Square wave signal from the Trinket

The first step was to see if the Arduino Uno was fast enough to accurately detect the square wave pattern, so I simply grabbed 100 samples from the input pin using digitalRead() with a delay of about 1/10 of the waveform period.  Then I plotted the data in Excel, as shown below:

Digital samples from Arduino Uno ‘receiver’

As can be seen from the plot, there are plenty of samples per cycle, well above the Nyquist rate for the dominant signal term.

Next, I moved the square wave signal to an analog input on the Uno, and reprogrammed to collect analog values vice digital ones, as shown below:

Analog readings with 500Hz square wave input

As can be seen, the analogRead() function is plenty fast enough to accurately reproduce the digital square wave signal, with about 10-13 samples/cycle.  So, at least in principle I should be able to detect a square-wave-modulated IR beam, and determine its ‘field strength’ (analog reading when the beam is ON) for homing purposes.

Stay tuned!

Frank

 

Charging Station System Integration – Part III

Posted 15 April 2017

In my previous post on this subject, I described some IR homing tests with and without the overhead incandescent lights, and the development of a ‘sunshade’ to block out enough of the IR energy from the overhead lamps to allow Wall-E2 to successfully home in on the IR beam from the charging station.  At the conclusion of that post, I had made a couple of successful runs using a temporary cardboard sunshade, and thought that a permanent sunshade would be all that I needed.

However, after installing the sunshade (shown below), I discovered that the homing performance in the presence of overhead IR lamps was marginal when the robot’s offset distance from the wall was more than about 50 cm.

Sunshade, oblique view

Sunshade, side view

Sunshade, front view

Apparently the IR interference was causing the robot to not respond to the IR beam until too close to miss the outer lead-in rail.  This issue was explored in an earlier post, but I have repeated the relevant drawings here as well.

 

Tilted gate option. The tilt decreases the minimum required IR beam capture distance from about 1.7m to about 1.0m

Capture parameters for the robot approaching a charging station

When the robot is ‘cruising’ at more than about 50 cm from the tracked wall,  the IR interference from the overhead lamps prevents the robot from acquiring the charging station IR beam until too late to avoid the outer lead-in rail, even in the 13º tilted rail arrangement in the first drawing above.

 

So, what to do?  I am already running the IR LED at close to the upper limit of the normal operating current, so I can’t significantly increase the IR beam intensity – at least not directly.  I can’t really increase the size of the ‘sunshade’ dramatically without also significantly affecting the IR beam detection performance.  What I really needed was a way of increasing the IR beam intensity without increasing the LED current.  As it turns out, I spent over a decade as a research scientist at The Ohio State University ElectroScience Lab, where I helped design reflector antenna systems for spacecraft.  Spacecraft are power and weight limited, so anything that can be done to improve link margins without increasing weight and/or power is a good thing, and it turns out you can do just that by using well-designed reflector dishes to focus the microwave communications energy much like a flashlight. You get more power where you want it, but you don’t have to pay for it with more power input; the only ‘cost’ is the insignificant added weight of the reflector structure itself – almost free!  In any case, I needed something similar for my design, and I happened to have a small flashlight reflector hanging around from a previous project – maybe I could use that to focus and narrow the IR beam along the charging station centerline.

LED flashlight reflector

So, using my trusty PowerSpec PRO 3D printer and TinkerCad, I whipped up an experimental holder for the above reflector, as shown below

Experimental 3D-printed flashlight reflector holder

Reflector mounted on experimental holder

IR LED mounted on reflector

A couple of quick bench-top tests convinced me I was on the right track; At 1m separation between the IR LED/reflector combination and the robot, I was able to drive the robot’s phototransistors into saturation (i.e. an analog input reading of about 20 out of 1024 max), where before I was lucky to get it down to 100 or so.  However, this only happened when I got the LED positioned at the reflector focal point, which was tricky to do by hand, but not too bad for a first try!

Next, I tried incorporating the reflector idea into the current charging station IR LED/charging probe fixture, as shown in the following photo. This was much closer to what I wanted, but it still was too difficult to get the IR LED positioned correctly, and this was made even more difficult by the fact that I literally could not see what I was doing – it’s IR after all!

New reflector and old charging station fixture designs

However, the reflector focusing performance should be (mostly) the same for IR and visible wavelengths, so I should be able to use a visible-wavelength LED for initial testing, at least.  So, I set up a small white screen 15-20 cm away from the reflector, and used a regular visible LED to investigate focus point position effects.  As the following photos show, the reflector makes quite a difference in energy density.

Green visible LED, hand-positioned near the focal point

Pattern without the reflector

Next, I used my Canon PowerShot SX260HS digital camera as an IR visualizer so I could see the IR beam pattern. As shown below, the reflector does an excellent job of focusing the available IR energy into a tight beam

IR beam visualized using my Canon PowerShot SX260HS digital camera

IR LED, without reflector

Next, I made another version of the reflector holder, but this time with a way of mounting the LED more firmly at (or as near as I could eyeball) the reflector focal point.

Reflector holder modified for more accurate LED mounting

With this modification, I was able to get pretty good focusing without having to fiddle with the LED location, so I set up some range tests on the floor of my lab.  With LED overhead lighting (not incandescent), I was able to get excellent homing performance all the way out to 2m, as shown in the following photos and plots

Range testing the IR reflector in the lab. Distance 2m

IR Detector response vs orientation at 2m from reflector, in the lab

IR reflector beam pattern at 2m, visualized using digital CCD camera

After this, I decided to try my luck again out in our entry hallway, with the dreaded IR interference from the overhead lighting and/or sunlight.   I installed the lead-in rails in the ’tilted’ arrangement, and then performed a response vs orientation test with the robot situated about 2.5m from the IR LED/reflector assembly, in natural daylight illumination with the overhead incandescents OFF.  This produced the curves shown in the plot below.

Robot response vs orientation test setup, 2.5 m from tilted lead-in rails & LED/reflector assembly

IR detector response vs orientation test, 2.5 m from IR LED/Reflector assembly

In the above Excel plot, the individual detector response minimums can be clearly seen, with minimum values in the 200-300 range, and off-axis responses in the 800-1000 range.  This should be more than enough for successful IR homing.

After seeing these positive responses, I ran some homing tests starting from this same general position.  In each run, the robot started off tracking the right-hand wall at about 50 cm offset.  One run was in daylight with the overhead lights OFF, and another was in daylight with the overhead lights ON.  As can be seen in the videos below.

Both of the above test runs were successful.  The robot started homing on the IR beam almost immediately, and was successfully captured by the lead-in rails.

So, it is clear the reflector idea is a winner – it allows the robot to detect and home in on the IR beam from far enough away to not miss the capture aperture, even in the presence of IR interference from daylight and/or overhead incandescent lighting.

Next step – reprint the IR LED reflector holder with the charging probe holder included (I managed to leave it out of the model the last time), and verify that the robot will indeed connect and start charging.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Charging Station System Integration – Part II

Posted 07 April 2017

After getting the wall-following mode re-implemented in the ‘new-improved’ Wall-E2 operating system, I re-enabled the IR homing part of the code, only to discover that Wall-E2 thought it could see the IR homing beam everywhere, in the atrium, even though the IR homing LED wasn’t even in the room (and was turned OFF, besides)!  This didn’t make a whole lot of sense, until I realized Wall-E2’s sensors were ‘seeing’ IR from the overhead floodlamps.  I confirmed this by having my ‘lovely assistant’ wife turn the atrium lights on & off and monitoring the detector outputs.  Same thing for the entry hallway (see data below)

I thought I had solved this problem way back when I first started working with the IR detection/homing idea back in October of last  year.  In that initial post where I investigated the OSEPP ‘IR Light Follower’, I found I had to turn OFF the LED track lights in my lab in order to avoid swamping the detector array.  Over the next month or so this project evolved into a custom designed array of Oshram SFH309FA-4 phototransistors (see http://fpaynter.com/2016/11/ir-light-follower-wall-e2-part-vi/), and as part of this I discovered that the Oshram devices weren’t at all sensitive to the LED track lighting in my lab.  From this I concluded (erroneously as it now turns out!) that I didn’t need a ‘sunshade’ at all, which allowed for a much simpler installation on the robot (see http://fpaynter.com/2016/11/ir-light-follower-wall-e2-part-vii/).  Finally, when this module was transferred from the 3-wheel test bed to the Wall-E2 and combined with the charging status display panel, we arrived at the ‘final’ arrangement shown in the following photo

IR detector module (red) shown beneath charging status display panel (blue)

This worked great for all my in-lab IR homing tests with my all-LED overhead track lighting, but as soon as I started testing out in the rest of the house, the overhead incandescent and halogen lamps swamped out the detectors. So, it was back to the drawing board – again :-(.

The first thing I did was to confirm that the IR detectors were indeed getting swamped by the overhead lighting, and that it was possible to prevent this by some sort of shielding.  I started with a wrap-around cardboard shield that completely covered the IR detector module, as shown below, and placed the robot on the floor of the entry hallway in a normal ‘wall-following’ position

Sunshade V0. Not very practical, but does set the baseline for the rest of the tests

Test position for 07 April 2017 sunshade tests

 

With this setup, Wall-E2 was pretty much deaf to the overhead lamps, showing no reaction to cycling the lights.  Of course, this configuration would also completely prevent it from detecting the charging station IR beam, but….

Next I modified the above V0 cover to allow a bit more visibility, as shown below:

Sunshade V1 oblique view

Sunshade V1 front view

This version is still a bit too restrictive for normal IR beam detection/homing, but was worth a shot for testing purposes.  With this setup, the overhead lighting is noticeable, as shown below.

Sunshade V1 at sunshade test position. Lights ON four times for 5, 5, 5 and 10 sec

As can be seen from the plot, the V1 sunshade responds to the overhead lamp IR with an average A/D reading about 700 out of 1024.

Next, I modified the sunshade again so that it would just allow the IR detector module to ‘see’ straight ahead, on the theory that that would be the minimum requirement for successful IR beam detection/homing.  With this setup, the overhead lamp response was stronger, but still not completely overbearing; the IR detector response to the overhead lights averages A/D values of about 600 out of 1024.

Sunshade V2 oblique view

Sunshade V2 front view.  Note IR detectors visible from directly in front

IR detector response to overhead lamps. ON three times for 5 sec each

When I ran a homing test with the V2 sunshade in my lab (with LED overheads, not incandescent), the robot homed successfully from about 1 m away, with the following IR detector/homing performance

In-lab (LED overhead lighting) charging station homing performance with V2 sunshade

In the above homing test, the IR detector values started out at approximately 900, 200, 450 and 750, respectively.  This means I could set the IR beam detection threshold at, say, 400 and still have a 200-count noise margin in both directions (200 down from the overhead IR flooding value, and 200 up from the typical 1-meter IR beam intensity).  From the movie it is obvious that the minimum IR reading is switching back and forth between at least two detectors, so I think it is safe to say that IR homing would occur even in the presence of a 600 unit noise level, as the 200 or lower reading switched between detectors.

Next, I moved the charging station into the entry hall so I could test IR homing performance with the overhead lights on and off.  My prediction was that Wall-E2 should be able to home successfully in both cases.  For the test, the charging station lead-in rails were oriented in the preferred ’tilted’ orientation, as that should give the best homing performance, as shown in the photos below:

Hallway IR homing test setup

Robot shown in captured position. Note lead-in rail ’tilt’ relative to the wall

I made three runs; Lights ON, lights OFF, and Lights ON, as shown in the three video clips below:

 

As can be seen in the videos, Wall-E2 successfully homed to the charging station with the overhead lights OFF, but not with them ON – bummer!

So, further testing will be required to determine the particulars of the two failed runs, and what – if anything – can be done about it.

After moving my laptop out into the hallway so I could capture telemetry from the robot while also filming the runs, and checking everything out, I made two successful IR homing runs – one with the overhead lights ON, and another one with them OFF.  I captured telemetry from both runs, and was clearly able to distinguish between the lights ON and OFF scenarios.  The videos and the telemetry plots are shown below:

08 April 2017 Hallway Test2 with V2 Sunshade – Lights OFF

08 April 2017 Hallway Test2 with V2 Sunshade – Lights ON

From the above plots, the lights ON & OFF conditions are readily recognizable.  In the ‘OFF’ condition, the ‘background noise’ is at about 600-700, and the IR beam is at 100-200.  In the ‘ON’ case, the noise level is higher (lower count), at about 150-250, but the IR beam is lower too – around 50-150.  I guess this makes some sort of sense, as the IR energy from the charging station beam is a separate, additive source relative to the overhead lighting.  Also, both plots show good motor response curves – the left & right motor speeds are obviously being adjusted rapidly in response to IR detector changes.

So, at this point I’m pretty convinced that the V2 sunshade is working, so I plan to print up a permanent version that will (hopefully) simply slide on to the front of the existing charge status display panel.

Stay tuned!

Frank

 

 

 

 

 

Charging Station Design, Part XI – PID Tuning

Posted 14 March 2017

As I was doing the IR homing tests described in my last post, I noted that Wall-E2 wasn’t all that great at homing; in particular it missed the opening in the lead-in rails on several occasions, instead hanging up on one side or another.  This was a bit mystifying to me, as I thought I had the homing code working very well with my old 3-wheel robot (see ‘IR Light Follower for Wall-E2, Part X – More PID Tuning‘).  At the time, I decided to use one of my best scientific research tools and simply ignore the problem, hoping it would either go away, or my subconscious mind (by far smarter than my conscious one!) would figure it out in the shower or while drifting off to sleep.

And, in fact, last night while drifting off to sleep, I remembered that the PID tuning for the 3-wheel robot had to take into account the fact that even small differences in drive wheel speeds get magnified by the free-castering front wheel.  Wall-E2, with its all-wheel drive behaves entirely differently, and since small wheel speed differences don’t get amplified into big directional changes, the PID tuning parameters appropriate for the 3-wheel version are too passive for the 4-wheel one.

So, I went back to the drawing board (again!) for PID tuning for the 4WD Wall-E2, starting with the current 3-wheel parameters as the ‘too passive’ baseline.  From my previous article, the final PID parameters were P = 0.1, I = 10, D = 0.2, with the input scaled by 1, 5, or 10 depending on IR beam signal strength.  My initial thought is that the 4WD robot needs a lot more ‘D’ (differential) to increase its turn rate with respect to IR beam heading changes, so I tried a couple of runs with the D value increased from 0.2 to 2 (factor of 10 increase).  As the following video shows, this did seems to increase Wall-E2’s agility somewhat, but still not enough to overcome even minor initial heading offsets, especially to the left.

After some more testing with different values of P,I,D, I began to wonder if I might be having problems with the basic IR detection hardware and homing software, independent of the PID tuning issue.  When I did the previous PID study, I also captured the raw detector data, which allowed me to determine how the PID tuning and the basic detection hardware/software were interacting.  I may have to go back and do that again with the 4WD setup

Stay tuned,

Frank

 

 

Charging Station Design, Part X

Posted 11 March 2017

The last few days have been spent partially implementing the software structure and state design described in my ‘Wall-E2 Operating Mode Review‘ of 06 March.  Up to this point, the only thing Wall-E2 knew how to do was wall-following, but there were (and still are) a number of ‘sub-modes’ associated with wall-following.  Now all of this code has to be grouped into the ‘Wall-Following’ state/function under the main program, and the new ‘Charging’ and ‘IR Homing’ state/functions added.

For purposes of testing, I decided to basically comment out all the wall-following code and concentrate on the IR Homing and Charge modes.  To do this, I added the new operating mode determination and top-level case switching code at the top of ‘loop()’, and left the MODE_WALLFOLLOW: case un-implemented.  the MODE_CHARGING code is all new, but the MODE_IRHOMING code was copied from my previous work with the 3-wheel robot last November.

After getting most of the MODE_CHARGING and MODE_IRHOMING code implemented, I ran some bench tests to see how well things work.  After finding and fixing a number of typos, logic errors, and downright goofs, I was able to video a successful IR home to charge, followed by a charge-completion disconnect (instead of waiting for a full charge termination, I tied execution to manual charge plug disconnect).  Here is the video

 

A couple of side notes from the above video

  • It is evident that the IR homing with PID works very well, to the point where the lead-in side rails almost aren’t needed (but only ‘almost’, I fear)
  • The fixed charging station fixture isn’t quite high enough.  In order to more accurately line up with the center of the charging port, I had to raise the fixture by about 5mm
  • The disconnect routine backs up far enough, but doesn’t quite make a complete 90º turn.  I should be able to tweak the turn duration a bit to make that happen.

So, at this point, I think I have most of the new parts of the operating system working, although some parts are hard to test due to the long charge times.  Here’s what I think remains to be done:

  • Reconnect and test the LIDAR and ping sensor hardware; this will be required to test wall-following and the charge station avoidance sub-mode
  • Reprint the charging station fixture with 5mm more height
  • Set up the full lead-in rail arrangement and confirm proper homing/engagement, along with proper dis-engagement, and proper avoidance when the battery isn’t low.
  • re-implement and test the wall-following code in the new structure, as MODE_WALLFOLLOW.  This should be just a bunch of cut-and-paste operations.
  • test the various ‘normal’ state transitions; wall-follow to IR homing to charge monitoring to wall-following
  • test the wall-follow to IR homing to charge station avoidance transition.
  • fully test the end-of-charge scenario.

12 March Update

I reconnected the LIDAR and ping sensor hardware, and confirmed (not without some wailing and gnashing of teeth!) proper operation.

I reprinted the charging station fixture with a 5mm pedestal, and transferred the LED/charging plug from the old fixture to the new one.

Set up the lead-in rails on my benchtop so that I could test both the ‘hungry’ and ‘not hungry’ IR homing scenarios.

Tested the ‘not hungry’ scenario by setting the ‘full’ threshold back down to 7.4V (50% charge, per http://batteryuniversity.com/learn/article/lithium_based_batteries).  The following video shows one of the test runs:

 

13 March Update:

After a bunch of software and hardware bugfixes, I think I finally have the ‘home to charge’ scenario working, as shown in the following video clip and edited telemetry capture:

The significant parts of the video are:

  • The POST test in the first few seconds
  • The successful IR homing operation
  • The successful charger plug engagement.  When the robot senses the charger plug, it commands the motors to stop.
  • Just after the charger plug engages, the charger’s battery relay is enabled, which disables power to the motors (note the red motor controller power LED goes OFF).
  • For this test, the BATT_CHG_TIMEOUT_SEC parameter was set to 10 seconds.  Rather than wait the entire 10sec, all but the first few and last few seconds were clipped.  Note that after the 10sec timeout, the robot disables the charger relay which re-applies motor power to the robot (note the red motor controller LED is re-enabled), and the robot backs off the charger plug.
  • The robot now backs completely clear of the lead-in rails, and turns away from the nearest wall before going back to wall-following mode.

I have also posted an edited version of the telemetry captured during this test.  As you can see, the robot transitions to IR homing mode, successfully homes on and engages with the charging plug, monitors the charging process, an then executes the ‘ExecDisconManeuver()’ to disengage from the charger and back out of the charging station area.

 

Stay tuned,

Frank

 

Wall-E2 Operating Mode Review

Posted 06 March 2017

At this point in the 4WD robot’s development, I think it might be time to review all of Wall-E2’s different operating modes, and what external sensor conditions determine which operating mode is active.  Previous to this point, the robot had only one operating mode – the ‘Wall-Following’ mode.  Now that the new battery pack has been added, along with the new capability for homing in on an IR beam to connect to a charging station has been added, the robot operating system must be enhanced to manage the additional operating modes.

Operating Modes/States:

  • Wall-Following:  This is the ‘normal’ operating mode, active when nothing else is going on (no IR beam present, no charging voltage present)
  • IR Beam Homing: Occurs when an IR beam is detected.  In this mode, a PID controller determines wheel speeds in an attempt to home in on the IR source.  This mode has two distinct sub-modes, depending on the battery state of charge.  if the battery is relatively full, then the front obstacle avoidance distance is increased such that the obstacle avoidance turn will occur before the robot gets captured by the lead-in side rails.  Otherwise, the obstacle avoidance distance is disabled entirely, allowing the robot to be captured and the charging connection to be made
  • Charging:  Occurs when the robot is connected to the charger.  In this mode, the charging status lines are monitored to detect end-of-charge (EOC).  When EOC is detected, the robot backs straight out of the lead-in rail area, and then turns 90º away from the nearest wall and transitions to wall-following mode.

Here is a draft state-transition diagram for the system

As can be seen from the diagram, the system starts out in the POST (Power On Self Test) state, and transitions to one of the other states depending on sensor input.

  • If no IR beam is detectable, and the charging station is not connected, then the system transitions to the normal ‘Wall-Following’ state.  The wall-following state continues until either an IR beam is detected, or a physical connection to a charging station is detected.  The normal exit condition from the wall-following state is via detection of an IR beam, which causes transition to the ‘IR Homing’ state.
  • In the IR Homing state, there are two possible behaviors; if the battery is more than 3/4 full (not quite sure how I’m going to define that, but…), then the robot will actively avoid the charging station by performing an obstacle avoidance 90º degree turn at a front distance larger than the extent of the charging station lead-in gate, and transitioning back to the ‘Wall-Following’ state.  If the battery is less than 3/4 full and a charge is desired, then the robot will continue to home in on the IR beam until a physical connection to the charging station is detected, whereupon the system transitions to the ‘Charging’ state.
  • In the Charging state, the system monitors the charging status signals, waiting for both ‘Fin1’ and ‘Fin2’ signals to become TRUE.  When this happens, the robot backs out of the charging station lead-in gate area, executes a 90º turn away from the nearest wall, and transitions back to the ‘Wall-Following’ state.

Here is a first cut at a system software structure chart that incorporates the above modes/states

Wall-E2 Charging Station Design, Part VIII

Posted 18 January 2017

In my last post on this subject, I had discovered two major problems with my current strategy to free Wall-E2 from the need for human assistance.  The first problem was how to get Wall-E2 disengaged from the charging station, and the second was how to transition from charging power to on-board battery power without going through a power-cycle reboot.  As noted there, I decided there were two things I needed to do – install a MOSFET switch in the coil circuit so I could switch the relay back from ‘charge’ mode to ‘run’ mode before the external +5V charging voltage disappeared, and to connect the external +5V charging voltage through a blocking diode to the Arduino Mega’s +5V buss so the controller could continue to operate while the batteries were charging.

So, I made the above changes to the charger module, as shown in the image below (changed areas highlighted). For reference, the ‘original’ schematic has also been included

Dual Cell Charging Module with changes highlighted

Dual cell balance charger. Note the two Axicom relays ganged to form the required 3PDT switch

As can be seen above, there are two major changes

  • Added a IR510 n-channel enhancement mode MOSFET to control relay coil current via a new ‘Coil Enable’ signal rather than directly from the external +5V line.  A 100K pullup was added so that the default configuration of the MOSFET switch was ‘ON’.  A LOW signal on the ‘Coil Enbl’ line will switch the MOSFET to ‘OFF’, thereby switching the 3.7V cells from ‘charge’ (parallel) to ‘run’ (serial) mode.
  • Removed the blocking diode from the +7.4V ‘Robot +V’ line, and added a separate ‘Chg +5’ 2-pin terminal with a blocking diode.

After making these changes,  I set up an experiment where I could simulate the process of connecting the robot to the charger (thereby switching the batteries from the ‘run’ (serial) to ‘charge’ (parallel) configuration, with Arduino power being supplied by the external +5V line, and then disconnecting it using the new MOSFET circuit to switch the batteries back to ‘run’ (series) configuration before mechanically disengaging the external +5V plug.

Before conducting the experiment, I wanted to confirm that the reboot problem still existed.  I set the robot up a small block so the wheels were off the ground, turned on the main power switch, waited while the robot booted up and the wheels began to turn (the program was configured for continuous half-speed motion), and then engaged/disengaged the external +5V charging plug.  As expected, when the plug was engaged the wheels stopped turning but the Arduino continued to operate (confirmed by noting that telemetry readouts continued uninterrupted).  However, when I disengaged the plug, the wheels started turning again immediately, and the telemetry readouts continued unabated, indicating that no power-cycle reboot had occurred!  I ran this experiment several more times with the same result – after making the above changes, I could not get the robot to reboot in either direction – from ‘run’ to ‘charge’ mode or from ‘charge’ to ‘run’ mode. Amazing!

OK, so what caused the different behavior?  When I ran this same experiment before the changes, I saw an approximately 50msec gap between the time the external +5V line went away, to the time when the Arduino +5V line was again stable due to the +7.4V power via the regulator block. AFAIK, there were only three significant changes made to the charging module:

  • The IR510 MOSFET was added to enable/disable the coils under computer control, with a 100K pullup to keep it in the ‘ON’ (low resistance) mode by default.
  • The external +5V line and its associated blocking diode was removed from the ‘Robot +V’ terminal to a new ‘Chg +5’ terminal
  • The blocking diode between the battery stack and the ‘Robot +V’ terminal was removed

So, what’s the deal here?  AFAICT there are only two possibilities:

  1. I didn’t/don’t fully understand the mechanism producing the 50msec power gap in the previous configuration
  2. I don’t fully understand why the current configuration doesn’t have a 50msec power gap

Previous Configuration:

I am 100% certain that I observed a consistent, repeatable power-cycle reboot when switching from external +5V to internal battery operation, and I got the 50msec number from scope measurements .  To make the scope measurements, I triggered the scope trace on the falling edge of the external +5V line, and measured the time lag from that trigger to the time that internal battery voltage was available to the Arduino.  The reboot phenomenon was verified by noting the long (5-10 sec) delay between the time the external power was removed to the time when the motors started running again, and by watching the interruption on the Arduino serial port.  The 50msec or so gap is consistent with the idea that it takes some time for the relay coil field to collapse after external power removal, plus the time required for the relay contacts to physically move from the ‘engaged’ to the ‘disengaged’ contacts.  During this interval, the only thing powering the Arduino is the charge left in the 680 uF power supply filter cap.  Assuming a current drain of around 100mA, it would take only about 5-10msec to cause a 1-2V drop on the +5V buss.  With the measured current drain of about 68mA, I measured about 30msec for a 2V drop using my trusty O’scope, so this all tracks.

Current Configuration:

In the current configuration,  after the 2V drop, the voltage drops only very slowly, so the current drain must also drop significantly – could that be the answer?  Maybe most, if not almost all of the measured current drain is the coils themselves – so that after the blocking diode gets reversed, the drain out of the filter cap goes down by an order of magnitude or so, thereby letting the Arduino live on until the internal battery takes over?  Lets see – the specs for the Axicom V23105A5476A201 relay show 30mA for the coil current, times two relays gives about 60mA total.  The measured current with the relays engaged was about 68mA, meaning that when the diode blocks, the drain from the cap goes from 68 to 8mA, meaning an additional delta of 1V (from about 4.5 to about 3.5) should take 680X10-6/8X10-3 = 85msec, which is reasonably close to what I’m seeing on my O’scope.

That’s my story and I’m stick’n to it!

OK, so now my story is this:  In the previous configuration, the Arduino 6800uF filter cap supplied relay current all the way down to zero volts, which meant that the voltage across the cap (and consequently, the Arduino working voltage) dropped to below 3V in less than 20msec, well less than the time required for the internal battery source to take over operation.  In the new configuration, the blocking diode between the external +5V supply line and the Arduino isolates the Arduino from the charging circuit after a drop of about 2V.  After this, the Arduino is powered solely by the 6800uF cap, but because the Arduino’s current drain is much smaller than the 60-70mA required by the charging circuit, the cap can power the Arduino for another 100msec or so, which is plenty of time for the internal power source to come on line.

One implication from this story is that the MOSFET circuit may not have been required at all.  As evidence, the gate of the MOSFET is tied to external +5 through a 100K resistor, so by default it is ON (low drain-source resistance) all the time, unless deliberately switched from the Arduino.  During all this testing, that control line  has been left open, meaning the MOSFET is just sitting there, doing its best to emulate a short length of wire. However, I’m reluctant to take it out or deliberately short around it for three very good reasons (actually only one good reason, and two not-so-good ones); first, it is just barely possible that the MOSFET actually turns OFF at some point in the process, maybe hastening the relay change from energized to de-energized (that’s a not-so-good reason).  Second, it is a major PITA to disassemble the robot down to the point where I can access the MOSFET and install the short (another not-so-good reason). Finally, even if I do properly understand what is going on now, it is still possible that increased Arduino loads in the future will cause the reboot problem to re-appear; in this case, being able to de-energize the relays before disengaging from the charger will be a life-saver (that’s the good reason).  In addition, I’m unwilling to screw around with something that appears to be working just like I want it to (in other words – “if it’s working, don’t screw with it!!”)

Where to from here?

As it stands, I have a robot that can be charged through its front-mounted external power jack, and should be able to (assuming appropriate information availability) switch to internal battery power and disengage itself from the charging station.  Now I need to actually implement the entire solution, generally as follows:

  • Confirm that the proposed engagement/disengagement strategy will actually work.  To do this, I’ll need to modify the operating software to
    • recognize when the external power plug has engaged and is supplying power
    • switch back from external to internal power
    • disengage from the external power plug.
  • Build up the fixed portion of the charging station, including mounting the IR LED and supplying 5V power
  • Simulate the entire IR track/side-rail capture/engagement/disengagement cycle on the bench
  • Modify the operating system to implement the required additional tracking/movement modes

Independently of the above, I need to revisit the issue of how the charging station connects to the robot.  Originally, the idea was to connect via an array of contacts on the underside of the robot.  These contacts would mate with spring contact fingers on the top surface of a raised section of the fixed charging station, which would also contain status LEDs for the two embedded Li-Po chargers.  Unfortunately, I have been unable to come up with contact fingers appropriate for the application, despite trying three different contact finger ideas (EMI shielding finger stock, TE connectivity spring contact fingers, and Mill-Max spring-loaded contacts).  Fortunately, in the meantime I was able to successfully demonstrate automatic external charging power connection using a guide funnel for the front-mounted external power jack and a semi-flexible probe tube with the mating external power plug mounted at its end.

The advantage of using the front-mounted jack for automatic power connection is that all I have to do is to get that one jack/plug pair engaged/disengaged, as opposed to the nine underside contacts.  This is a huge simplification of the problem, and one that I have already demonstrated to be feasible.  The major disadvantage of this option is that all the charge-related decision making will have to be done by the robot, as the fixed part of the charging setup won’t know what is going on at all.  If I want to monitor charging status, I’ll have to do that via the onboard Arduino.  In the previous configuration (pre-MOSFET) this disadvantage was compounded by the fact that charging power could only be removed by physically disengaging the external power plug, which could only be done by some external physical mechanism since (by definition) the onboard wheel motors weren’t available during the charging process.  Since I now have a way around that dilemma (i.e. the robot can now unilaterally switch from external to internal power by means of the MOSFET switch and then use the onboard motors to effect physical disengagement), this huge problem goes away entirely.  I still have the problem of how (or if) to display charging status, but this is trivial compared to the problem of physically disengaging the charging plug.

If I decide to abandon the underbelly contact array idea, then I can re-purpose the 8-pin header on the charging module to route charging status information to the Arduino instead of to the underbelly contact array. This header is shown in the image below:

Charger module 8-pin male status/control header

As shown, there are 3 status pins for each cell charger, a ground line, and the new ‘Coil Enable’ line.  The status pins show whether or not the charger is receiving power (PWR), and whether the cell is still charging (CHG) or has finished (FIN).  In the original charging station design, the six status lines were brought out to individual LEDs via the contact array. If the underbelly array strategy is abandoned in favor of the single front-mounted connector, then these LEDs will have to be mounted somwhere/somehow on the robot itself.  Originally I was thinking that each status line would consume an Arduino Digital I/O pin, but now I’m not so sure.  All of these lines are actually already ‘powered’ from the charger modules themselves – all that is required to illuminate the CHG and FIN LEDs is +5V – the status line is tied to an open-collector output through a limiting resistor.  The PWR status line is simply the +5V power to each cell charger, so a limiting resistor is required.  All the required signals and connections are available, so all that is needed is some sort of mounting arrangement on the robot – perhaps it could be integrated into the mounting for the front-mounted IR phototransistors in a manner similar to the ‘backup light’ mount at the rear?

‘Backup Lights’ mounted to rear of the robot

IR phototransistor mounting at the front of the robot

there would be more LEDs (7 vs 4) but each LED would be much smaller (3mm vs 5mm).  On the same 56mm panel, 7 LEDs could be spaced 6mm apart, with 6mm spacing on the ends, something like the following

Prototype for a charging status LED panel

And, with my trusty PowerSpec PRO 3D printer I printed out a full-scale feasibility assessment panel in just a few minutes, as shown below. Of course much more work would be required to make this into a fully functional panel, but just this piece shows that all 7 LEDs can be accommodated in a panel that is generally the same dimensions as the IR sensor module.

Charge status LED panel full-scale feasibility model

Charge Status Display Panel Update

After the normal number of trials, I came up with a charge status display panel that could be co-mounted with the current IR detector assembly, as shown in the following images:

Stay tuned!

Frank