Nihar

This a log of Nihar Pol's weekly status updates for Team B2: 'Sole Mate'; Spring 18-500 Capstone.

Weekly Status Report

May 3, 2018

This weeks' progress (so far):

• We gave our final presentation of Wednesday. Personally I felt happy with it, obviously I'm not sure about what the instructors thought.
• We have been looking at the imu data from our trials and have been trying to make sense of it. Reed and I are still grappling with the issue that the IMU data drifts. We are still trying to think about how exactly to extract meaningful data (esp. w.r.t. slope) from it.
• We have all been working on extracting features from the data sets. I worked on strike pattern detection and back posture detection primarily. I also helped Reed look into slope detection since that has become fairly difficult. Currently, we have 98.25% accuracy Heel Strike detection; 53.38% midfoot strike detection; 75.19 forefoot strike detection. We have 71.64% accuracy when detecting if back is bent forward; 37.2% accuracy when determining if back is neutral; 62.03% accuracy when detecting if back is bent backward.
• I also modified the IMU scripts to send data to the android phone in batches. Connor modified the app so that it puts the data into buckets (IMU/IR).

Next Steps:

• debug right shoe
• collect more data, esp. with Reed and Connor running.
• increase accuracy (if possible) of our feature - detection algorithms
• make app better, test it out with live data streaming.
• prepare for final demo
• start working on report
• make poster

Weekly Status Report

April 26, 2018

This week's progress (so far):

• Over the weekend I collected many data sets we need. These were all done on flat, level ground:
  • heel0 heel strike
  • heel1 heel strike
  • back_0 back bent backward, long stride, strike in front of knee.
  • back_1 back bent forward, forefoot strike.
  • curvy_path_f forefoot strike. Ran in a curvy path. Yaw should change, roll + pitch should not.
  • forefoot strike 0 (initheel) initially there was some heel strike. mostly forefoot strike.
  • forefoot strike 1
  • heavy_impact higher than normal landing force; heel strike.
  • midfoot strike
  • over pronation
  • supination
  • short stride many steps/second.
• Later in the week I collected a few more data sets. In these data sets, I varied strike pattern (forefoot, midfoot, heel). However, the surface was a steep hill which I ran up and down on, so this will give us a way to examine how to compute slope accurately. I haven't yet measured the hill incline, but I will do that tomorrow morning.
• We modified the scripts to write out the computed IR distances in mm (no raw data), so Connor adjusted the step detection to account for that.
• We started looking at the IMU data. So far we came up with a high pass filter to remove the drift that shows up in the IMU data. We set the cutoff frequency to 3, because that is the average steps/sec and that has shown good results. However, it does reduce the dynamic range of the data so we will have to scale it back.

Next steps:

• We have been looking at the data collected and are in the process of extracting features we need from it.
• Connor also gave me his bluetooth script so I will modify it to send data to the phone.
• Prepare for the presentation on Wednesday.

Weekly Status Report

April 19, 2018

This week's progress (so far):

• Over the weekend I finished integrating the scripts for three sensors together and adding timestamps (with microseconds) to the scripts. I also fiddled around to get the fastest sampling rate I could. The fastest possible are 67 samples/sec for the IR sensors, ~13 Hz for the load cell, and 300 Hz for the IMU. The IMU is much faster because it runs in C++ 11. Unless we overhaul the code and use a C library, those will probably be the fastest sampling rates. I inserted a delay into the IMU code so that it is around 125 Hz. With the timestamps the actual sampling rate can always be derived from the data.
• The weather wasn't too great but I did start collecting the final data sets. So far I just did three trials in which I varied strike pattern. Connor and Reed worked together to write a Matlab program that detects steps. I think Reed also already wrote translated python code.
• I calibrated the IR sensor—meaning, I constructed a piecewise function to translate the raw adc values into actual distances. The below plot shows both the three different functions I used as well as the final piecewise function that integrates all three of them.

Next Steps:

• Now that the scripts are ready, I will finish collecting the final data sets. Then I will work on impact , strike pattern , and slope computation.

Weekly Status Report

April 12, 2018

This week's progress (so far):

• On Friday we did our first running trials. At the time only the four IR sensors on the right shoe were confirmed working and collecting data. Although we were able to successfully collect 4 trials worth of data, the setup eventually failed because soldered connections broke. First two IR sensors failed and then the battery got disconnected. We resolved the issue this week by hot gluing all solder joints and double checking for weak connections. I spent time earlier this week making everything sturdy in general.
• Earlier this week I finally finished building both shoes. I installed both IMUs including the one that Reed had. I also installed all software packages necessary to run them. I have tested and confirmed that all desired functionality with respect to data collection is working for both shoes. This includes all 8 total IR sensors, the two load cells, and the IMUs. We experimentally determined that the maximum sampling rate for the IMU is 49 Hz. Personally I find this to be a little odd, but I haven't looked into it yet.
• Building the thing took much longer than expected. However, we finished building everything now. Also, the device is much sturdier now than it was even at the beginning of this week.

Plan going forward:

• We are on a tight squeeze now because we are very behind in our timeline. We only have two weeks left to collect the data and analyze it. I am going to collect the datasets either tomorrow or Saturday. We will start analyzing the data on the same day. At least the simplest data analysis parts including detecting steps, strike pattern, slope / incline, and impact computation should be done by early next week.

Weekly Status Report

April 5, 2018

This week's progress (so far):

• The main thing I accomplished so far this week was creating version 2 prototype of our shoe design. I replaced all the metal components with nylon parts. I also reduced the amount of wood used. I moved the case to the rear. I replaced the heavy 2200mAh battery with a 500 mAh battery which is much lighter. I doubt the overall thing will get lighter than this so for now we will stop iterating on this part of the design and move on to the data stuff.
• I started working on duplicating the right shoe onto the left shoe. I haven't finished yet, but so far I have the IR sensors mounted. I have to finish the mounting and wiring inside the case.
• One of the step-up dc power boards failed, so I bought a different one. I think it failed because some solder fell on it and shorted some thing; it doesn't work anymore (I measured no voltage when I plug in input). I rewired it with the new one and it works well. I ordered a spare just in case.
• I worked with Reed a little on looking at some of the data we collected last week. So far he has been able to get the data through matlab. Not sure what else he has looked at.
• We demoed, obviously, this week. At time of demo, the power board had not been working so we had to use usb power. We were still able to demo, though.
• I had actually planned to do some actual running data collection today (Thursday). However, because of other coursework I pushed it to tomorrow. Hopefully tomorrow we will get some good data that everyone can look at. I will also finish the left shoe by then.
• I have thought some more on the data collection part. I put together a table of our various strategies for how we are going to approach the different characteristics we are hoping to extract. To see it, please look at pages 7 and 8 of the below document "Project Log".

Next steps:

• Finish building the left shoe to be identical to the right shoe
• Start leaving the project at school so anyone can work on it anytime.
• Build all of our datasets.
• Bridge with Connor and start sending relevant data from the pi to the app.

Weekly Status Report

March 29, 2018

This week's progress (so far):

• I finished mounting the two IR sensors to the shoe (in addition to the load cell from last week). I went and got a case for the pi + other chips and I mounted everything inside it. This includes the pi, the ADC, load cell amp. Everything fits in the case and it actually closes.
• Pictures are attached of the current prototype.
• I showed the prototype to the group and they as well as the TA's + instructors suggested we make it lighter if possible. I ordered nylon angle brackets and screws to replace the steel ones currently in use.
• On Thursday (today) we collected 70 seconds worth of walking data. We are just trying to get a feel for what the data is looking like. We also successfully got the pi to work with the new, lighter batteries. It automatically starts collecting data on bootup.
• All the remaining parts we ordered came (except for IMU which Reed ordered later).

Next steps:

• Modify the prototype to have the case mounted in the rear to reduce the moment of the foot.
• Replace steel with nylon components to make it lighter.
• Build the left shoe to be identical to the right shoe
• Integrate Reed's IMU into the setup (he's not quite ready for it yet)
• Get the left and right shoe pis to communicate with each other and synchronously collect data.
• Prepare for the demo on Monday. We need to do research regarding what algorithms we will use to process the data.
• Collect more data to build our datasets.

Here is a chart of IR sensor data for the 70 second walk we did. Note that it was not continuous walking. Y axis is sensor voltage which is inversely proportional to distance.

Weekly Status Report

March 23, 2018

This week's progress (so far):

I created a script and edited some files on the Pi so that when it boots up with my power bank, it automatically runs whatever program I want it to. My intended method of data collection for now is to write the data to a file.

I placed orders for what should be pretty much all the remaining parts.

I began mounting sensors to the shoe. So far I have finished mounting the load cell in a cavity inside the shoe that I made. The load cell actually sits between the bottom of the shoe insert and a metal plate I placed under it. I positioned it such that the top of the load cell is level with the rest of the sole, so there isn't any odd feeling when wearing the shoe. I have also begun the process of mounting the IR sensors. I have attached pictures of what I have so far.

Next steps:

- Finish mounting the sensors to the pi

- Obtain a case and mount the pi + chips inside it.

- Collect some preliminary data with the 4 IR sensors + load cell (+ IMU once Reed has it going)

- Once the rest of the parts we ordered come in, I can finish modifying the left shoe and then we can focus on data collection and processing.

I should also mention that I sent a message to the group and Ben (TA) during the Spring Break regarding my reflection of Connor's contribution to the group up to that point. TA Ben helped us handle the situation and he spoke with everyone one-on-one. After the break it seems that Connor has found something he wants to work on, but obviously we still want to make sure everyone feels comfortable and that we are all communicating with each other. I have included a summary of the message here for reference:

In my opinion, I find Connor's contribution to our group to be less than satisfactory.

When Reed and I were working on the design document due the Thursday before Spring Break, I was surprised that Connor was not working on the document when we were. The total work that Connor put into the document consisted of copying and pasting the rubric from the rubric document into a google document and possibly copying and pasting some bullet points from our slide presentation. Connor contributed no meaningful content to the document, however. Connor made no sketches, schematics, nor drawings that went into the document. Connor did not write any prose or even work on any of the sections of the document. Connor did not even inject any meaningful ideas into the creation of the document itself. When I asked Connor what he was working on, he did not respond quickly and even when he did respond he stated that he was not available to work on the document.

Instead, he claimed that he was busy traveling or doing something else and thus could not work on the document because the said event took longer than he expected.

When I submitted the design document on Thursday night, I did not feel comfortable at all. I knew that it was incomplete. The document was nowhere near as detailed as it needed to be. Even though we received a B grade on the assignment, I genuinely felt the document is worthy of only a C grade. It is missing an entire person’s worth of ideas, research, and work.

One of the most frustrating things I had to deal with when working on both the design presentation and the design document was trying to figure out what exactly Connor was going to work on. Not once has he communicated to the group that he is going to take ownership of any part of our project. When I made the timeline and alerted the group about it, nobody edited it or responded regarding what parts of the project they wanted to work on. In the design document, I did not know what part of the project he wanted to work on, so I have placed items for which he is responsible in the Team member responsibilities section.

It turns out, however, that he has a history of contributing only partially to the group’s efforts.

For the very first design presentation, he made (I am pretty sure) zero slides.

I was ok with that, however, given that it was the first presentation and our idea was still premature. What really disappointed me, however, was that I could tell that he had not even read the presentation beforehand. He was reading off the slides when we presented.

For the second design presentation, he worked on no more than two slides. He made just one or two of the slides and they had just to do with the risks and the demo day. These are important slides, but they are not the core of the presentation. When I asked him, "Connor, are you doing something?", he did not respond.

Furthermore, he did not come to class twice around three weeks ago. He provided no reason as to why he did not show up to class.

In general, an entire person’s worth of effort and ideas is missing from our project. I don’t want to point fingers, but it is true that his contribution to this group is nowhere near that of what it should be. He needs to contribute both ideas and actual work, especially when there is work in front of him to do. He cannot just sit back and let other people do the work for him.

Weekly Status Report

March 8, 2018

All the parts have come.

We ordered some additional usb cables so that we don't get bottlenecked trying to communicate to the Pi.

The issue we dealt with this week had to do with installing libraries to get the Pi to work with the sensors. To install the libraries, we needed to first be able to ssh into the Pi via USB. Unfortunately we had only one USB TTL cable to use so we had to share it. We have ordered some more so that after the Spring break we shouldn't have any issues. The other issue was that we needed to connect the Pi to the internet, but the CMU wi-fi is different from normal wi-fi because it has authentication.

I was able to get it working on my own wi-fi and I installed the libraries. I soldered on wires to hookup two IR distance sensors (the Sharp ones; 2 - 15cm) to the ADC. I also got the ADC to work using the I2C libraries.

I modified some code to output the data from the IR sensors. I tested the sensors and verified that they can sense distance between 1 ~ 20 cm. I also tested them alongside each other to see if they would encounter cross-talk. I tried but was not able to distort the output of the sensors, even when I placed them right next to each other. I tried on both reflective and opaque surfaces.

The one thing I noticed, however, was that the voltage output curve varies from sensor to sensor. The maximum output of one sensor is much less than the other one, although the shape of the curve is the same. So, we will need to calibrate the data to take this into account.

Next steps:

• Do some more testing and measuring to better determine whether there is any cross-talk to worry about.
• We also have two more distance sensors that we had ordered, I will try to get those up and running, too.
• Continue working toward getting the other sensors (load cell) to work with the pis and then collecting data.

Weekly Status Report

March 1, 2018

All the parts have come except for the running shoes. We emailed Ben and are awaiting his response.

Reed finally got his raspberry pi working and gave me information on how to get mine working. I haven't tried yet, but I will soon.

The main event of this week was the second round of the design presentations. We presented on Monday and tried to address most of the concerns raised in the first design presentation. We have modified our idea slightly so that we are collecting distance measurements to compute the impact force rather than force-sensitive resistors or load cells. We are planning on securely bolting the IR distance sensors to the sides of the shoe.

The main concerns that were voiced by the professors on Monday in response to our design presentation were:

1) adding weight to the shoe or leg which may negatively impact the runner

2) using cables that could detach or break while running

3) using an OLED display mounted on the shoe that could potentially also impact the running form because the runner must look at it

4) handling cross-talk between IR sensors that may be positioned alongside each other and are blasting IR

We will keep thinking about these concerns as we work on the design document.

Next steps:

• work on the design document (due Monday)
• Figure out what happened to the running shoe
• Continue working toward getting the sensors to work with the pis and then collecting data.

Weekly Status Report

Feb 23, 2018

We ordered parts last week and they came. We soldered the headers to the three Raspberry Pis. We tried to get one of the Pis working, but were unable to. We will take a look next week once the design presentation is over.

We spent much time putting together some of the key components of the design presentation. We revised our idea to have collect more accurate data using IR distance sensors rather than just load sensors or force-resistive sensors. We put together a timeline and a block diagram.

Next steps:

• finish the design presentation. We need to make sure it addresses all the concerns raised by the professors.
• get the Pis working. We want to get data from the sensors soon.