1 / 18

Lab 3

Lab 3. 2 students: missing info from submissions? (emailed) 95 +/- 13.6 Remember, -10 per day late! Overall, really good. Homework 2. No submissions from 2 students? Overall Average: 58 Stdev : 6.3 Averages on each problem: 9.7 8.3 7.6 9.8 9.1 5.8 9.6. Problem 1. Problem 2

altessa
Télécharger la présentation

Lab 3

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Lab 3 • 2 students: missing info from submissions? (emailed) • 95 +/- 13.6 • Remember, -10 per day late! • Overall, really good

  2. Homework 2 • No submissions from 2 students? • Overall Average: 58 • Stdev: 6.3 Averages on each problem: • 9.7 • 8.3 • 7.6 • 9.8 • 9.1 • 5.8 • 9.6

  3. Problem 1

  4. Problem 2 Sum these 4 prob. distributions to get -> answer (which we can double check to see it sums to 100%) 3*1 + 2*2 + 3*14 + 51 = 3+4+42+51 = 100

  5. P(d|2): 1 is 20%, 2 is 60%, 3 is 20% We read a distance of 2 from YB YB

  6. P(d|2): 1 is 20%, 2 is 60%, 3 is 20% We read a distance of 2 from YB YB

  7. P(d|2): 1 is 20%, 2 is 60%, 3 is 20% We read a distance of 2 from YB YB

  8. P(d|2): 1 is 20%, 2 is 60%, 3 is 20% We read a distance of 2 from YB YB

  9. P(d|2): 1 is 20%, 2 is 60%, 3 is 20% We read a distance of 2 from YB YB 10.2+8.4+.2+.4 = 19.2 -> mult by 5.208333

  10. Problem 3 P(d|2): 1 is 20%, 2 is 60%, 3 is 20% We read a distance of 2 from YB YB 10.2+8.4+.2+.4 = 19.2 -> mult by 5.208333

  11. Problem 4 • This method isn’t great. First, we’re not observing after every step. Second, info from this one beacon only covers part of the space. It’s also a little suspicious that there’s no probability for being greater than 3 squares away. • We should look for the beacon on every timestep. If we don’t see the beacon, that can also help us update our belief state. • We could also have more than one beacon. Or add an additional sensor (e.g., range finder). Or take multiple (independent?) readings from the beacon.

  12. Problem 5 • Enlarge all boundaries by the maximum size of the robot. In this case, it’s the diagonal. • Really, instead of square S and G, you should just have points (since that’s the point of using a configuration space).

  13. Problem 6 • First, start with the configuration space. • Then, draw lines between all pairs of corners/robot that are visible to each other. • Many people lost points for • Confusing a “visibility graph” with the parts of the environment the robot can see • Using the workspace, rather than the configuration space • Only drawing lines from the center of the robot start location to visible corners • Missing some of the corners in the graph

  14. 6: A, B, C, D, E, F, G 7: A, B, E, F, D, C, G Each node gets expanded once (and only once) Therefore, will only “visit” a node once, and will not get stuck in a cycle In Depth First, will visit D and F because D will be added back to the queue of nodes under consideration

  15. Test Tuesday • 75 min • 1 page of paper, 1 side • Handwritten or typed

  16. Learning to Control a Low-Cost Manipulator using Data-Efficient Reinforcement Learning. M. Deisenroth, C. Rasmussen, D. Fox • http://www.cs.washington.edu/ai/Mobile_Robotics/projects/robot-rl/ • EM/Math • Timing, losing connection • Eliminate drift (w/ wind) • What about a more uniform room? Scaling to larger maps? • Automatic scaling was cool.

More Related