Training process error:Can't instantiate abstract class DoubleIntegrator with abstract methods disconnect_mask, leader_follower_assign, unsafe_test_mask
Opened this issue · 1 comments
TypeError: Can't instantiate abstract class DoubleIntegrator with abstract methods disconnect_mask, leader_follower_assign, unsafe_test_mask 。 There is no definition of these three methods in the code, how can we solve it?(The running code is:python train.py --algo gcbf+ --env DoubleIntegrator -n 8 --area-size 4 --loss-action-coef 1e-4 --n-env-train 16 --lr-actor 1e-5 --lr-cbf 1e-5 --horizon 32)
Thanks for pointing this out. As of now, the leader-follower-based coordination is only implemented for SingleIntegrator dynamics. If you are interested in just the GCBF-based safety part, you can use the code from https://github.com/MIT-REALM/gcbfplus/ which allows you to train GCBF for a variety of dynamics.
Furthermore, it is possible to implement those functions using the SingleIntegrator dynamics as the template. We are actively working on this project and will update the repository with complete training methods for the rest of the dynamics as well.