Updates History

This is a complete history of all updates posted on the home page of this site.

5/8/09

Our final demonstration was today. We accomplished obstacle detection and avoidance really well. Unfortunately, we had some problems with picking up barcodes from the ceiling due to latency issues with the barcode reader. The internal timing between the barcode and our software was still off. Our wall-following problems were alleviated by stringing 7 barcodes across the hallway at each point of interest. We hope that another student will pick up this project next year!

After our wheel got fixed, we found out that our sonar sensors had stopped working! Fortunately, it was a quick fix because it was a ribbon cable that was accidentally left disconnected internally while I-GUIDE was being repaired. The final presentation went well and and our final paper for the 2009 FCRAR has been officially accepted. The initial calibration routine has been fixed and we have chosen to simplify the wall follow algorithm slightly in order to maximize demo points. We are also using multiple barcodes across the ceiling (from one wall to the other) in order to better localize our self during the final demonstration.

4/28/09

This week our robot has managed to lose a wheel! The wheel came detached from the internal motor. We have to wait for Nick Schmidt to fix this before we can continue any experimental work on I-GUIDE. However, there are still bugs with our initial calibration routine that can be worked out in simulation. These calibration routine software bugs are what we worked on during the entirety of the lab period. There were also some small software glitches that we fixed at this time.

4/21/09

Today's time was mostly spent installing the new IR ring. We got the IR ring installed and running shortly after lunch. We cleaned up the code and cut our overall code size by almost 25%! We also programmed an updated wall following algorithm for testing. The basic concept for this wall following algorithm is as follows: 1.Store our distance to both walls at the center of the hallway; 2. If the right wall distance is not equal to the distance stored when we were in the center of the hallway, use the left wall to stay in the center of the hallway and vice versa for the left wall distance not being equal to the distance stored when we were in the center of the hallway; 3. If both walls fall out of calibrated range, then we want to try to stay equidistant from both the perceived left wall and the perceived right wall. We were only able to do a quick test run today and it seems like the new algorithm still needs some adjustments.

4/14/09

Our IR sensor ring was not complete for the laboratory today, so we developed a non-ideal hallway in the AriaMapper program and tweaked the wall-following algorithm. The non-ideal hallway is great for testing our algorithms, which detect doorways and open corridors. We are also able to test the robots ability to stay in the middle of the hallway with these static obstacles. We adjusted some of the wall-following parameters and tweaked the current wall following algorithm. However, it still is not stable enough to stay in the middle of the hallway. Because a large amount of reprogramming will need to be done, we decided it was better to spend our time tuning up the obstacle detection and avoidance algorithm. We were able to finalize our obstacle detection and avoidance algorithm. We now have implmented all aspects of obstacle detection, except for the bump sensors because they are not mounted on the robot yet.

4/7/09

Its been a couple of weeks since the last update; class projects and exams on top of senior project have gotten the better of us. The week of 3/24/2009 was spring break and we came into work on our senior project. Now that we have figured out how to get reliable readings from the ADC11 and IR sensors, we have to re-calibrate our sensors to verify that our software equations are correct to convert an analog voltage to a distance. We took data, came up with an equation, and re-programmed all of our software distance calculations.

The week of 3/31/2009 and 4/7/2009 have included massive code debugging and adjustment for real world situations. Essentially, we have the following real world problems: 1) We can scan a barcode off of the ceiling from the center of the hallway. However, we cannot stay in the middle of the hallway due to open doors. Some doors open inward, while others open outward. Both open and closed doors affect the current wall follow algorithm tremendously. 2) What is an acceptable distance to start avoiding people? Obstacle detection is very important but while avoiding obstacles it is very likely that we will miss important barcodes.

We are currently working through various solutions to the two problems stated above. We have been able to successfully detect doors that open inward, but still have problems with doors opening outward causing us to veer away from the center of the hallway. We have an algorithm finished in simulation to calculate our distance to the wall using two IR sensors. Unfortunately, in order to test in real life, we need our final IR ring put into place, which is currently being worked on by Nick Schmidt. The obstacle detection algorithm bugs are slowly being worked out so that we avoid obstacles and people at just the right time. We still need bump switches put into place so that the robot has a backup plan in case all else fails and we do actually hit something.

3/17/09

The IR Sensor issues we have been having caused us to lose an entire day's progress. We spent the entire day troubleshooting this IR sensor problem. It turns out that the IR sensors high input impedance coupled with the Analog to Digital Converter to USB (ADC11) was causing voltage spikes every time we sampled any of the ADC11's channels. We tried various solutions to fix this problem before we found the root cause (the high input impedance).

First we noticed that we were utilizing a 'single collection' mode for the ADC11 which was only intended for periodic sampling. The ADC11 also has a 'streaming collection' mode that seemed to be more suitable for our application because our application samples each channel that's in use once every software loop. After multiple hours of troubleshooting we determined there was a bug in the library provided from Pico Technology in one of the necessary functions for streaming mode. Finally after hooking an oscilloscope up and providing a voltage from a power supply we found that the spikes were due to the high impedance of the IR sensor. We fixed the problem by putting a 100k ohm resistor between each channel in use to ground.

3/10/09

We have almost finalized our wall follow algorithm. We have tuned it so that it drives relatively smoothly down the center of the hall. You can see our rough first pass at the algorithm (it's much better now) in the video section of our gallery page here here. This video was taken before we realized we are receiving poor IR sensor readings. (Note: Our wall follow algorithm still actually functions :P) There is a problem with the hardware interface between the IR sensors and the ADC to USB, which we are still working out. The first generation of the barcode scanner holder has been finished. We have yet to mount it to the actual robot due to the problems with the IR sensor interface. We plan to work on our project pretty heavily over spring break to catch up with the problems that we have been having.

2/24/09

We finally have our IR sensors and all connectivity cables in our possession and we figured out how to hook them up. Originally, the IR sensors arrived without cables. We ordered ten but they only shipped five. Once we got all of the cables and hooked up the IR sensors, we discovered that the IR sensors part number had been mislabeled on all ten IR sensors. They were actually the updated version of IR sensor which has a different pinout. Luckily no IR sensors were permanently damaged during initial testing. We now have two IR sensors mounted on the Pioneer along with the IR sensor voltage regulator circuitry. The ADC to USB is hooked up on the Pioneer allowing us to easily interface the IR sensors with the laptop. Other than a few software bugs, we are ready for experimental testing of the wall follow algorithm.

We also found out how to operate our long range barcode reader in hands free mode. The user manual stated that we needed to have the proprietary stand to place it in 'hands-free' mode. When we called tech support they told us we had to buy a proprietary $100 stand because it had a magnet inside it. Bingo! We played around with a magnet which we had in the department and in only a few minutes we found where we needed to place it to put the scanner in hands free mode.

Our software platform has a built-in obstacle detection algorithm. This algorithm will be modified and utilized for our obstacle detection algorithm. You can see our Project Update presentation on our Deliverables page here. You can also see some photos of the current progress with two IR sensors, ADC to USB, and external circuitry hooked up on our Gallery page here.

2/9/09

Everything is still on schedule to finish up the project by spring break. We have decided to use a low dropout voltage regulator for the monitor protection circuitry. We have also eliminated all feasible options to implement a fix for the poor sonar sensor readings in Jobst Hall. We cannot change the firing order or firing frequency because those are handled at the microcontroller level on the Pioneer 3. Also, the extensive bad data does not allow for simple logic fixes. While a complex algorithm could be developed to make these sonar signals usable, we will choose to continue implementation of the IR sensors for our wall follow algorithm. The ADC to USB is now fully integrated and the IR sensors and the bump sensors will utilize this device to interface to I-Guide's internal program. The barcode reader we received was unfortunately the incorrect reader. It is currently being sent back to the manufacturer so we have to wait for the correct one to come back. However, once we receive the correct reader the implementation will be simple. I-Guide's program will see the barcode simply as a keystroke. I-Guide's abstract will be submitted to The Florida Conference on Recent Advances in Robotics (FCRAR) by Saturday, February 15th.

2/3/09

We have decided to implement I-Guide without the use of a compass or gyroscope. The tile floors in Jobst Hall are an ideal environment as far as wheel slippage is concerned. We can use a combination of the sonar sensors, IR sensors, barcode readings, and rough dead reckoning for localization. We received the ADC to USB, long range barcode reader, IR sensors, and our monitor. We have already integrated the ADC to USB with our existing program. We are currently working on protection circuitry for the monitor and various fixes for the poor sonar sensor readings we receive in Jobst Hall. The Florida Conference on Recent Advances in Robotics (FCRAR) abstract has been submitted to our advisor Dr. Schipper for review.

1/27/09

We have just completed our first lab day for the spring semester of 2009. We have had to shuffle our schedule slightly because we have not received any of our parts yet. Feel free to view our updated schedule here. Also, we are still performing a cost / benefit analysis on a few components for I-Guide, such as the indoor / outdoor compass, gyroscopic correction system, and rear sonar sensors. We are also researching the possibility of implementing I-Guide without utilizing a compass or gyroscope. We are currently working on a paper for submitting our project to the Florida Conference on Recent Advances in Robotics (FCRAR).

12/9/08

We are now sponsored by Northrop Grumman! We have presented our first semester's work and planned out next semester's work. We also found out that the Pioneer 3's sonar sensors do not work well in our environment. Over winter break we plan to research various methods for making the sonar sensors more reliable. We will also complete the obstacle detection and avoidance algorithms that we did not finish this semester.

11/13/08

The development HID, a simple joystick, is now up and running in simulation. Basic program layout has begun. Subsumption vs blackboard architecture types are still being researched. Also, other tour robots and localization methods are being researched. The most viable option at the moment for localization appears to be a combination of a digital compass, sonar sensors, and RFID tags.

11/4/08

A lot of progress has been made in the first two weeks of this project. The Pioneer 3 has been chosen over the Magellan from RWI as the platform for the robot. Complete MobileSim maps of the 2nd and 3rd floors of the ECE Department have been completed. Flowcharts and pseudocode for the basic navigation algorithm are also complete. Currently working on a joystick interface to use for the development HID and researching localization methods and hardware.

10/13/08

Website is up!