Close

Dog Trials (more like...travails)

A project log for A Halo For Lucy

It's not what you think.

Bud BennettBud Bennett 04/16/2019 at 17:342 Comments

The days are not sequential. There were days between where nothing happened.

Day 1:

I started out by just securing the prototype to Lucy’s head with a long rubber band. She immediately shook it off and it went flying across the room. 

The next approach was to run two short steel wires from her collar to the back of the unit. She shook it again and it would flop down sideways or backward. It would not remain on her head.

Another version used a stiff nylon strap from her collar that ran under the prototype. Same problem as above.

Clearly, I needed some help from from somebody with knowledge and skill in the physical world...my Wife (who can sew).

Day2:

Whenever you have more than on person on a project progress slows to a crawl. The Wife’s priorities aren’t the same as mine and she has veto power over the activities upon which I depend. We both agreed that some sort of bonnet would be best. The first attempt was more like a hood than a bonnet. I had no input as to the color. It had two holes for Lucy’s ears, but quite a bit of material on the top, bottom and sides. Lucy had major issues with it when we put it on her without the electronics attached: she shook her head repeatedly, scratched it from the sides with the front paws and then sat down and used her rear paws to really do a job on it. The bonnet stayed put, but looked uncomfortable to me. I thought her issue with that particular bonnet was that it rubbed against her ears and that was annoying.

Day3-Day5:

I suggested to my Wife that a better approach might be an upside-down jock strap (but not in those terms.) Basically, two ribbons with a connecting piece of cloth to run between Lucy's ears where the electronics could attach. This was produced with a bit of grumbling. The ribbons connected under Lucy’s chin with pieces of mating Velcro. Again, I had no input for the color - Pink! I glued a softer piece of foam pad below the electronics to increase the comfort level.

Lucy got used to this headdress pretty quickly, first without the attached electronics and later with, but still executed the obligatory shake rattle and roll. All of the abuse took its toll on the electronics. One of the speakers went flying across the room — the speakers have a great big gob of hot-melt glue as reinforcement now. I’m surprised that’s the only thing that broke so far.

The first time that Lucy wore the complete package with battery and operating electronics was in a bright room. Every time she would shake her head the position of the sensors on her head would change — either sliding slightly backward or moving one half inch to one side. Many times the hair on her ears, or the ear itself, would block the outer sensor and the unit would be inhibited (objects less than 50mm inhibit the system.) 

There was a five minute interval where everything worked and she followed me around the house. It was bright so she could see fairly well, but when she drifted too close to an object or wall the unit would flash a detection event and I could tell that she noticed the beep from the speaker.

Later on in the evening when it became darker I put the powered prototype on her again. The same issues with alignment and head shaking persisted, but there was a short period when it seemed to be working correctly. I tried to inform her as to the function of the unit by bringing my hand toward her to activate the speaker and then touching her head from that side. Since it was dark, Lucy was moving much more slowly and appeared to respond to the audible feedback when she got too close to an object or wall. Unfortunately, this prototype doesn’t mute when it’s not moving or tilted so I had to prevent her from lying down or sitting close to an object that caused the unit to keep beeping for no good reason. The entire session came to an end when she shook her head and disconnected the solder joint from the negative terminal of the battery.

Day 6:

Back to the old bonnet style. The jock strap version did not keep the prototype in the correct position after Lucy's head shake. The larger bonnet is better suited for the larger electronics package we're using now. I still need to readjust it a bit after she shakes it, but in general is is a better temporary solution.

At night I rearranged the soft furniture in the living room and turned down the lights. Lucy was fitted with the bonnet and electronics -- the 14500 Li-Ion  battery hung by rubber bands from her collar. She was an unwilling participant. I called her from around the room and observed her progress negotiating the maze. The unit pretty much performed as expected, but it was difficult to determine if Lucy was understanding what it was doing. She seemed to pay attention to the feedback from the sides that warned her of a nearby object, but she occasionally bumped into something head on.

I think the problem may be that the unit would beep at the ground whenever she lowered her head so she was ignoring that data after a while. The unit had detection thresholds of 500mm/front and 400mm/sides.

Day 7 - The Dog Whisperer:

We invited a local dog trainer (dog psychologist?) to observe Lucy with the prototype and give us some tips on how to train her to understand and use it. We decided to temporarily disable the front sensor since it was detecting the ground when Lucy pointed her head in that direction, which was often. We thought that this would confuse her initially, so for training purposes its gone. 

The trainer suggested that when I train her to the prototype I have a short leash connected to her harness and lead her around the training space populated with cones or empty cardboard boxes. When she bumps into a box  I should gently pull her back with the leash and give her some verbal feedback. Training sessions should be no more than 5 minutes in duration.

Going Forward:

This prototype was made to prove out the concept and identify changes or improvements in both the hardware and software. It has done its job in that regard. I'll be moving on to the second pass prototype after a couple more sessions with Lucy.

Discussions

Bud Bennett wrote 04/20/2019 at 19:39 point

No we don't use a clicker, even though the trainer recommended it when she was a pup. She responds pretty well to voice commands.

  Are you sure? yes | no

Gerben wrote 04/20/2019 at 18:50 point

Pretty exciting. I hope see get's it eventually.
Good idea to get a dog trainer involved. Did you ever do clicker-training with Lucy?

  Are you sure? yes | no