Minds and Machines Project
Team Name: DIPLS- Dan Z, Issac L, Pia P, Lin J, Sierra B
Slogan: "Not all intelligent machines are fast"
For a brief overview of the Wumpus World problem we were faced with, please see the links at the bottom of the page. Otherwise, below is our teams website laying out our process through the challenge from start to finish!
Day 1: We started by building the robot!
-Our objective for today was to decide on a design, and get as much of the building done as possible.
Materials:
Assembly was completed in 2.5hrs with precidence/ priority placed on getting the sensors about 1/3 in off the horizontal playing surface.
Day 2. Today the robot design was finished so the team worked on a sample code to try to figure out the turning for the robot.
Day 3:
Then came the logic scenarios and hypothetical input functioning that the team had to analyze in order to make the code for our robot. The team worked on this a lot; testing different inputs and each day a little more was added. Day 4, the team was able to get DIPLS Jr. onto the game board and confirm it was capable of turning.
When we first started programming we started with simple movement commands to be able to move the robot. At first, the robot couldn't move forward in a straight line so a few cable adjustments were made. Additionally, the group came to the conclusion that DIPLS Jr. is incapable of making 90 degree turns. Therefore, the team will have to make sure to watch DIPLS while navigating the board as to not go out of the boundaries of the board.
These videos below show DIPLS Jr. turning, and moving forward.
On the teams third lab day, DIPLS Jr. made it onto the board again to evaluate coding procedure and the team noticed repeated forward motion that would go off at an angle after only traveling a few inches.
Futhermore, the same problem was happening when DIPLS Jr. would go to make a turn to face a different grid square. DIPLS Jr. wasn't making 90 degree turns and with each turn, the angle consistency became less and less.
Therefore, on day 4 to analyze these problems, the team looked at the potential variables such as the movement of the USB cable that was hooked up to the robot and the affect an uncharged battery could have on DIPLS JR's turning.
As our first design for DIPLS Jr. had a leading ball bearing as the front tire centrally placed, the team decided this forward instability wasn't a way to navigate the board, but rather was more suitable to drag across the board.
As a result, DIPLS Jr. was flippped around and the sensor that was in the back was moved to the front of DIPLS Jr. so everytime DIPLS Jr. would go over a white line on the board it would know. Our logic behind placing the sensors in the front was supported by the idea that the team wanted DIPLS Jr. to be able to adjust to the new box he just moved into rather than realign based on the lines from the box he left.
The videos below show DIPLS Jr. traveling with his new design in much straighter lines, as well as stopping perfectly on the white lines.
DIPLS Jr.'s code consisted of an algorithm that marked each squares information based on the numerical input given. Then, it would try to eliminate false positives. Once the gold was found, DIPLS would find a recursive path to the gold, and then find a safe path back home the same way.
Day 6: The team switches from an array of integers within the coding for DIPLS to an array of tiles.
Benefits to using the array of tiles are:
DIPLS can keep track of a tile that's definitely the Wumpus, definitely a hole etc.
Easier to read and analyze the previous outputs of 1,2,5 etc.
Back to the drawing board for the code, DIPLS design is 100% finished with both sensors activated and working!
Days 7, 8& 9:
In the final stretch of the project window, the team finalized the logic and tested the code without the USB cable attached to the robot to make sure all the commands worked and the output movement was within the grid tiles.
The class competition takes place on Thursday, December 8th, 2016. Good Luck to DIPLS Jr. and all the other competitors, may the best code win!
Competition results: Of all the competitors, Sir Robin, DIPLS Jr., Whomp-Whomp, and W80W, Sir Robin was definietly the most intelligent robot present at the competition with the given game boards! Sir Robin was able to efficiently and concisely navigate through each grid square and logically make the right moves, killing the Wumpus when neccessary! DIPLS Jr. was able to navigate the board but running off a recursive loop, on the first board after obtaining the gold, DIPLS had to go through all 16 potential moves which took 10 minutes to get back home. When the correct route came up, DIPLS couldn't actually move it just kept running through the different options. ALso, in the first board, as shown in the video below DIPLS ended up in a loop going out of the grid lines.
After watching all the robot- computer- human interfaces the team still views DIPLS Jr. as mindless and without a conscience. There was no movement unless the human typed in directions which then, the computer relayed and fed to DIPLS Jr. Therefore, DIPLS Jr. also doesn't have free will either, it couldn't select a different route other than the one it was coded for if it didn't like that path. Additionally, I think having a mind means independence in terms of AI implication. For DIPLS case and all of the other robots viewed tonight, none of them were independent as each team had to show up to the competition. Lastly, the team realized the trial and error, work and time neccessary to create such machines and as a result we haven't neccessarily changed our opinion on the feasibility of AI, but now have more respect for the people who believe they can reach that level of intelligence with a machine and are dedicating that much time to the process.
Below is an additional video of DIPLS Jr. at the WumpusWorld competition showing him successfully reaching the gold and heading back but for some reason he stops one square next to home and thinks he's home. Therefore, all together DIPLS Jr.'s code and logic would need some tweeking. For starters, the group changed the logic after the first board to track the moves it made with the left bottom square being home instead of scanning through all the possible routes, so it would already know which one was safe. Additionally, the team ran out of time and kept running into loops when it came to putting the logic in for shooting the Wumpus, therefore DIPLS Jr. would need to be coded for that maneuver. Otherwise, I think the building of DIPLS Jr.'s robot turned out to be a very successsful design with the sensors in the front.
Website Links | Description | |
---|---|---|
1 | https://www.youtube.com/watch?v=lxrJKobzfaU&list=UUdH4jKoTt9Lr18RxrRqv4gQ&index=95 | Video from the 2011 RPI competition |
2 | http://www.cogsci.rpi.edu/~heuveb/Teaching/M&M/WumpusWorld/WumpusWorld.htm | Wumpus World webpage with rules and guildines |
*This page was designed for a group or RPI students for a final semester project reflecting the engineering process behind the Wumpus World Problem project in the Minds and Machines class at Rensselaer.
Designed by Sierra Blondeau
Contact GitHub API Training Shop Blog About © 2016 GitHub, Inc. Terms Privacy Security Status Help
.