Build Log Day 80:
Today it was time for another milestone test. I took the Man to the test maze I built on day 65 and set it the task of sorting all the objects in the maze into either consumable or nonconsumable objects by bringing them to the appropriate collection boxes. It succeeded in performing the task on the first try but it was a little slow so I set it to run on an endless loop. I then spent the next couple of hours replacing the objects for it to sort. By mid afternoon it had achieved a much more stable balance when moving, only going down to three limbs every once in a while and thereby increased overall speed significantly. I once again used a voice command to stop the loop and once again it seemed like it wasn’t working but eventually it did stop, just like yesterday. I will have to work on some kind of signal system so that it can acknowledge that it has received a voice command.
Build Log Day 81:
I think the best way for the Man to acknowledge a voice command is to respond with some kind of sound. I tried installing a small buzzer but I couldn’t get it to work satisfactorily. It produces a sound but the volume is so low as to be barely audible. It turns out the electrical system I’m using runs on too low voltage. I have come up with an alternative solution however. The aeration bellows basically just circulates the air through a kind of chemical exchanger then pumps it back out again, using one single tube for intake and exhaust. If I can somehow modulate the exhaust air I think I can produce a sound that can be used for acknowledgement. However, I have to be smart about this because I don’t want to block the intake. I have some ideas for how to do it but I’m not sure which one I will go with at the moment.
Build Log Day 82:
I’ve been experimenting with different ways to produce sound and I find that if I stretch a rubber membrane across the air intake / exhaust pipe, it vibrates, creating a clear baaa sound. Unfortunately this membrane blocks the air intake. I tried a number of different ways of opening the membrane to let the air in. The best solution so far is to cut the membrane in half and attaching each half to some small Muscle actuators. This way I can make it stretch across the pipe opening for making sound and retract to the sides during air intake. I also wanted to attach a funnel to the output in order to project the sound further. I played around a bit with the idea of having an array of different funnels that can shape the sound waves to produce different sounds but in the end I settled for a simple amplifying cone. Instead of using different sounds, I programmed a set of combinations of long and short blasts of sound for different meanings. For example “ba ba ba“ means “message received” and ”baaa baaa” means “performing command” and so on.
Build Log Day 83:
Today I got back to working towards the main goal: finding and consuming biomatter. Since the Man is able to identify consumable biomatter and pick it up, the next task is to ingest it. The Metabola energy converter can only process small pieces so any object being fed into it has to be crushed or chopped. The most efficient way to do this is with two counter rotating rollers covered in blades. Unfortunately this is not possible with the actuators im using. Instead I created a kind of masher consisting of two jaws lined with cutting/ grinding surfaces, connected with a hinge. By repeatedly opening and closing this masher, any objects between the jaws will be mashed into a sort of pulp which is suitable for the energy converter. I attached a rubber sheet to the masher so that it forms a sort of flexible funnel, which leads the biomatter into the intake tube of the energy converter. When I tested it I quickly discovered that about 80% of the biomatter processed by the masher falls down to the floor. To solve this problem I attached another rubber sheet that encloses the masher jaws. There’s a slit in it that can be opened and closed by a set of Muscle actuators to let the biomatter in. I didn’t have time to test this addition, that will be my first task for tomorrow.
Build Log Day 84:
The first thing I did this morning was test the improved biomatter masher. It works well with over 99% of crushed biomatter going down the funnel into the intake tube for the energy converter. While working on it I had a stroke of genius. The masher with the rubber sheet enclosure is essentially a flexible funnel with an adjustable opening – just the kind of thing I would need to modulate the sound. I immediately set about modifying the Man accordingly. I took the intake tubes for both the energy converter and the aeration bellows and put them together in a larger tube that I attached to the output end of the masher. I also attached some extra actuators for controlling the opening in the rubber enclosure of the masher. I then mounted the whole thing on the B.R.A.I.N Box, just below the cameras, so it can easily project the sound in the desired direction. I spent the afternoon testing which sounds this setup can create and building up a list in the database.
Build Log Day 85:
I started the day with testing the masher in this new configuration. It ran into problems almost immediately, as some of the crushed biomatter dropped into the intake tube for air and clogged it up. It was a hell of a job to clean that out without getting any of it into the actual bellows. I decided to halt all testing and work on a solution for this problem. What I came up with is quite simple: a flap with a hinge at one end and a single actuator, which covers the air intake whenever biomatter is being processed in the masher. That’s all I had time for today.
Build Log Day 86:
This morning I started testing the full process of consuming biomatter including picking it up, feeding it into the masher, mashing it to a pulp then sending it down into the energy converter. A first everything was working well but after about an hour the Man started acting strange. It was making weird sounds and repeating a kind of convulsive movement, sort of like a wave was going through its body. After acting like this for a couple of minutes it opened the masher to maximum opening and a slurry consisting of semi converted biomatter shot out through the intake tube, flowing out onto the ground, some also spilled onto the Man. I immediately shut down the Man then started cleaning up the mess. Once I had everything cleaned up I could analyze what happened. My conclusion is that some of the biomatter had been left out for too long and become fermented. The energy converter had a strong negative reaction to the chemicals produced during fermentation which caused it to expel its entire contents as part of a built in damage prevention routine. This expulsion was what caused the Man to convulse and make sounds. Clearly there needs to be some kind of chemical sensor to detect if the biomatter has fermented and thereby avoid ingesting it in large enough quantities to cause such a bad reaction.
Build Log Day 87:
I’ve been researching appropriate chemical sensors. Different types of biomatter will react differently when they go bad so I need one which covers a broad range of chemicals. However, most of the ones I can find are either specific for a certain type chemical or require manual adjustment. After a lot of searching I’ve found a combination that I believe will work. I ordered a gas sensor from the company Olfactory Inc. which can detect a fairly wide range of gases without manual adjustment. The problem is that the detection range is rather limited so I’ve complemented it with some chemical receptors that have to come into contact a material to detect what it is. Each such can only detect a single type of chemical but since they are small I can arrange them in little sensor arrays so as to cover a wider range. I ordered with express shipping, hopefully I can get the sensors tomorrow.
Build Log Day 88:
The chemical sensors arrived this morning just like I had hoped. I installed the gas sensor between the masher and the cameras. It pokes out a bit from the rest of the B.R.A.I.N Box which gives the whole thing a rather characteristic appearance. I then a built several sensor arrays from the little chemical receptors and attached them all on a rubber wedge. I attached this inside the masher so the crushed biomatter can pass over its surface and the chemical receptors can detect what type of chemical is inside. I spent the afternoon writing a detection routine. The basic concept is to first hold an object up to the gas sensor. If no bad gas is detected, the object is fed into the masher. During mashing the chemical sensor arrays are constantly active to detect any unexpected chemicals. If an unwanted chemical is detected, the masher will open and all contents will be expelled. If no unwanted chemicals are detected, the mashed biomatter will be allowed down into the energy converter. The only question is, how to define good vs bad chemicals.
Build Log Day 89:
Last night I kept thinking about how to define wanted or unwanted chemicals. I came to a conclusion that should have been obvious right from the start: training. I have spent most of the day feeding different types of fresh biomatter to both the gas sensor and chemical sensor array, and registering those in the database. I will leave sample of all of them out for a couple of days then do the same thing again.
Build Log Day 94:
I figured five days would be enough for the biomatter that I left out to go bad. I’ve spent most of the day feeding it to the chemical and gas sensors, to register the chemical compositions for “bad” or “unwanted” in the database. For each sample let the Man expel it from the masher so as to avoid any adverse reactions. I ended the day with running a few tests, giving the Man either fresh or bad biomatter and seeing how it would react. I went well in most cases but I had to tweak the settings for a few of them.
Build Log Day 95:
This morning I started a big test of the entire biomatter identification and consumption system. I had put a mix of fresh and old biomatter, as well as some nonconsumable object on a big table. I then tasked the Man with consuming the fresh biomatter and leaving everything else, be it old biomatter or some other object, on the table. At first it was going really well but after about an hour it all went wrong. The Man started expelling liquid waste from the bladder, spraying it onto the table and thereby contaminating everything on it. I had to shut the test down and discard all the biomatter then clean up the nonconsumable objects. To be honest, this happened because I was so busy with the masher and the chemical sensor and so on that I completely forgot about the liquid waste from the Kidney purifiers and apparently the subroutine for emptying the bladder triggered on its own. After cleaning everything up I also checked the solid waste hose from the energy converter. It was so full it was about to explode. I emptied out the stinking waste and wiped the valve with some tissue paper so that the Man wont smell. This is clearly something I will have to deal with before I can call this project complete.
Day 95 Continued:
My strategy for dealing with the waste is to have the Man expel it in a certain location so as to avoid contaminating anything. I have created a small routine where it will break from whatever it is doing and move to the corner of the room for outputting the waste, then go back to the task. I also entered the object type “bucket” into the database. If it can identify a bucket in the corner, it will aim the waste output s into that. For this routine to work the B.R.A.I.N has to know when the waste container is full. For liquid waste I had the foresight to install a bladder full sensor but the solid waste tube completely lacks any control, it just has a manual valve. I rectified this problem, installing a full sensor on the tube and replacing the valve with one controlled my a muscle actuator. I also installed a length of rubber tube on the output from the bladder, hopefully that will create a controllable stream rather than having liquid waste spray out randomly. Now I just have to wait until either waste container fills up to test it.
Build Log Day 96:
Before the liquid waste disaster, the big test I was running yesterday worked well enough that I think I can call it a success. That only leaves one task for the Man to complete before I can finish the project: it must find and consume water in order to replenish the hydraulic fluid since some of it is lost during the purification process. To this end I have added water into the database and started teaching the B.R.A.I.N to identify it. This is trickier than most biomatter because water doesn’t have a distinct shape or color. For this reason I’ve programmed a three step method. I started by having it identify suitable containers like bottles, buckets and so on. It then checks the container for a clear liquid. If there is a clear liquid inside it will use the chemical sensors to identify it as water or something else. Water will be consumed but anything else will be expelled.
Build Log Day 98:
The last two days have been spent training the Man to identify water according to the method I described above. It now works pretty well so I ran a test. I put several container in the test chamber maze filling some with water, some with other clear liquids such as alcohol or some weak acid, and left some of them empty. Results are good, the man consumed all the water, leaving the other liquids in their containers. I even raised the difficulty slightly by filling some containers with alcohol mixed with water at a very low concentration. The man was no longer able to detect the alcohol with the gas sensor and would try to consume it, only to expel it within a few seconds. I call that a success. I think it’s time for a large scale test.
Build log day 99:
I spent most of the day setting up my final, large scale test of the Man’s abilities. I substantially enlarged my test maze and built several new walls and other obstacles. I then distributed a large amount of consumable biomatter, bottles of both water and diluted alcohol, and a bunch of nonconsumable objects, all over the maze. I finished off by setting waste buckets in the four corners of the room. I also purchased and installed a small Remote Data Logging Unit so I could monitor the Man’s condition throughout the test without having to interfere. It was late afternoon when I finished setting everything up. I gave the man its full set of tasks and started the test. Let’s see how it goes.
Build Log Day 103:
I thought I would be able to run the large scale test for at least 10 days but apparently that is not the case. The first day went as expected and the Man was performing well, consuming biomatter and water, and leaving all other objects in their place. On the second day however, I started seeing values on the remote data logger that are consistent with a decrease in energy reserves. This was strange because the Man was consuming enough biomatter to cover its energy need, as if something unexpected was draining energy from the system. The values kept getting worse during day three, and at 72hours, 47 minutes after test start the Man collapsed. At this point energy levels were so low it was no longer able to move any of the actuators except for the bellows and hydraulic pump. I shut the whole thing down to stop it from completely emptying the last reserves. I’ve started analyzing the data logs to see what went wrong.
Build Log Day 104:
The result of the analysis is this: apparently the B.R.A.I.N is unable to fully process all the sensor data on the fly. Instead it processes what it needs for the task at hand and stores the rest in a big memory buffer. Theoretically it should be able to empty the buffer when there is less incoming data, but since the Man is constantly performing tasks it never gets a chance. Over time the buffer fills up and in order for it to not become completely full, the B.R.A.I.N must devote extra processing power to it. This draws more current than normal operation, draining the energy reserves faster than expected. Increasing the energy intake won’t work, it will just delay the problem. What’s needed is to completely clear the buffer periodically. The most efficent way to do this is to minimize all other activity so that the B.R.A.I.N can devote most of it’s processing power to clearing the buffer. For this reason I have programmed a rest period into the Man’s overall behavior pattern. I’ve given it a couple of hours each day where it simply lies down on the floor, limbs not moving, with the cameras turned off. The only remaining tasks are simple automated ones including running the bellows and energy converter. As a bonus this rest period will let all the actuators cool down completely. Tomorrow I will restart the large scale test.
Build Log Day 125:
I’ve been running the large scale test for 20 days since implementing the buffer clearing rest period. Everything has been working well, the Man is consistently able to consume fresh biomatter and pure water, and discard everything else be it nonconsumable objects, old biomatter, or alcohol diluted in water. When the waste fills up it goes to one of the corners to empty it out into the buckets there and it will take periodic rests. All values look good; except for minor variations, energy levels have been stable throughout the test. I never had to interfere with the Man directly, the only thing I had to do was refill the test maze with water and biomatter, and empty out the waste buckets. I count this as a resounding success. At the moment I don’t know what to do with the Man at this point but I will keep it running in the test chamber until it runs out of biomatter, after that we shall see. For now I’m closing down this log.
Leave a Reply