daneshtarapore
commented
5 years ago See if this thread helps: https://www.argos-sim.info/forum/viewtopic.php?t=178 It seems we have to split the arena into multiple parts, and keep track of robots and objects in each part (see example https://github.com/ilpincy/argos3-examples/blob/master/experiments/multiple_engines.argos).
The thread also has some tweaks to speed up the argos simulation.
I also came across the MPGA example in https://www.argos-sim.info/examples.php (see last example). It seems interesting, but would take some work to redesign the argos-sferes interface.
Kind regards, Danesh.
On Thu, Jun 27, 2019 at 3:39 AM bossdm notifications@github.com wrote:
I have run code with multiple physics engines now successfully and this greatly speeds up the code.
However, the manual setting of the robots is causing some issue:
./bin/behaviour_evolcvt10D` experiments/Gomes_walls_and_robots_std.argos [INFO] Using 4 parallel threads [INFO] Chosen method "balance_quantity": threads will be assigned the same [INFO] number of tasks, independently of the task length. [INFO] Using random seed = 1 [INFO] Using simulation clock tick = 0.1 [INFO] Total experiment length in clock ticks = 1200 [INFO] Loaded library "./lib/libnn_controller.so" [INFO] Loaded library "./lib/libevolution_loopfunctionscvt10D.so" [INFO] The physics engine "dyn2d_0" will perform 10 iterations per tick (dt = 0.01 sec) [INFO] The physics engine "dyn2d_1" will perform 10 iterations per tick (dt = 0.01 sec) [INFO] The physics engine "dyn2d_2" will perform 10 iterations per tick (dt = 0.01 sec) [INFO] The physics engine "dyn2d_3" will perform 10 iterations per tick (dt = 0.01 sec) [FATAL] Dynamics2D model id "thymio1" not found in dynamics 2D engine "dyn2d_0"
The error can be avoided by commenting line 292 in the base_loop_functions.cpp which calls the reset_agent_positions(). However then you would need some other means of distribtuing the agents per trial; I have noticed that the 'distribute' in the argos-config is not doing anything for random resets, even with random seed 0, if you look at the positions of robots or any cylinders. According to the code on https://github.com/ilpincy/argos3/blob/f853c1a324b96ad5539e6de7f325a381c9bc45cb/src/core/simulator/space/space.h https://github.com/ilpincy/argos3/blob/f853c1a324b96ad5539e6de7f325a381c9bc45cb/src/core/simulator/space/space.h, it seems that the distribute is only used at the Init.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/resilient-swarms/argos-sferes/issues/14?email_source=notifications&email_token=ACAO4KINWF22GXEW4HS6YODP4QR7JA5CNFSM4H3XRRX2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4G36PR6Q, or mute the thread https://github.com/notifications/unsubscribe-auth/ACAO4KOVJZBVJOJFPIFYWK3P4QR7JANCNFSM4H3XRRXQ .
bossdm
I have run code with multiple physics engines now successfully and this greatly speeds up the code.
However, the manual setting of the robots is causing some issue:
The error can be avoided by commenting line 292 in the base_loop_functions.cpp which calls the _reset_agentpositions(). However then you would need some other means of distribtuing the agents per trial; I have noticed that the 'distribute' in the argos-config is not doing anything for random resets, even with random seed 0, if you look at the positions of robots or any cylinders. According to the code on https://github.com/ilpincy/argos3/blob/f853c1a324b96ad5539e6de7f325a381c9bc45cb/src/core/simulator/space/space.h, it seems that the distribute is only used at the Init.