Bruker:Vegardli

Fra Robin

(Forskjeller mellom versjoner)
Gå til: navigasjon, søk
(Added deception picture)
 
(5 mellomrevisjoner ikke vist.)
Linje 1: Linje 1:
Pathfinding using genetic algorithms and hardware acceleration for trajectory cost calculations.
Pathfinding using genetic algorithms and hardware acceleration for trajectory cost calculations.
-
= State of the project (22.10.2015) =
+
= State of the project (16.01.16) =
-
* Essay: Done
+
* <tt><++++++++++></tt> Essay: Done
-
* GA Application: Partly done
+
* <tt><++++++++--></tt> GA Application: Mostly done (~900 lines)
-
* CPU Evaluator: Mostly done
+
* <tt><++++++++++></tt> CPU Evaluator: Done
-
* GPU Evaluator: Not started
+
* <tt><----------></tt> GPU Evaluator: Not started (~1100 lines)
-
* HDL Evaluator: Mostly done
+
* <tt><+++++++++-></tt> HDL Evaluator: Mostly done (~2300 lines, many of which are not mine)
-
* Thesis: Not started
+
* <tt><----------></tt> Thesis: Not started
== Essay ==
== Essay ==
Linje 13: Linje 13:
== GA Application ==
== GA Application ==
-
Basic GA operations implemented, some results have been achieved. Needs further tweaking, trying different strategies and making it multithreaded.
+
Mostly done.
 +
 
 +
Successful demonstration of diversity algorithm under deception:
 +
 
 +
[[Fil:vegardli-divergence-success.png]]
== CPU Evaluator ==
== CPU Evaluator ==
-
Mostly done, maybe make it threaded.
+
Done.
-
Also comes with visualization.
+
Also comes with visualization. Generation 0, 5 and 10 of a typical run:
-
[[Fil:Vegardli-trajectory-visualization.png]]
+
[[Fil:vegardli-visualization-screenshot-2-gen-0.png]]
 +
[[Fil:vegardli-visualization-screenshot-2-gen-5.png]]
 +
[[Fil:vegardli-visualization-screenshot-2-gen-10.png]]
== GPU Evaluator ==
== GPU Evaluator ==
Linje 30: Linje 36:
== Thesis ==
== Thesis ==
Not started
Not started
 +
 +
= Future work plan =
 +
== January ==
 +
* Implement diversity methods in application
 +
* Establish communications between application and HDL model
 +
 +
== February ==
 +
* Run experiments to find performance and limits of model
 +
* Start focusing on writing
 +
 +
== March ==
 +
 +
== April ==
 +
 +
== May ==
 +
* Hand-in

Nåværende revisjon fra 3. mar 2016 kl. 14:49

Pathfinding using genetic algorithms and hardware acceleration for trajectory cost calculations.

Innhold

State of the project (16.01.16)

  • <++++++++++> Essay: Done
  • <++++++++--> GA Application: Mostly done (~900 lines)
  • <++++++++++> CPU Evaluator: Done
  • <----------> GPU Evaluator: Not started (~1100 lines)
  • <+++++++++-> HDL Evaluator: Mostly done (~2300 lines, many of which are not mine)
  • <----------> Thesis: Not started

Essay

Handed in and awaiting feedback

GA Application

Mostly done.

Successful demonstration of diversity algorithm under deception:

Fil:vegardli-divergence-success.png

CPU Evaluator

Done.

Also comes with visualization. Generation 0, 5 and 10 of a typical run:

Fil:vegardli-visualization-screenshot-2-gen-0.png Fil:vegardli-visualization-screenshot-2-gen-5.png Fil:vegardli-visualization-screenshot-2-gen-10.png

GPU Evaluator

Not started

HDL Evaluator

Mostly done, need more timing info and integration with GA application

Thesis

Not started

Future work plan

January

  • Implement diversity methods in application
  • Establish communications between application and HDL model

February

  • Run experiments to find performance and limits of model
  • Start focusing on writing

March

April

May

  • Hand-in
Personlige verktøy