Bridge Env

A simple API for bridge using Double Dummy Solver to evaluate the score.

Build Double Dummy Solver library

git clone --recurse-submodules https://[email protected]/elisten/bridge_env.git

For Mac users, to build dds as multi-threaded. Reinstall gcc

brew reinstall gcc --without-multilib

Then modify the Makefile and make.

cd Mac_patch
python dynamic_makefile.py
cd ../dds/src
make -f Makefiles/Makefile_Mac_clang_patched
cp libdds.so ../../.

For Linux users

cd dds/src
make -f Makefiles/Makefile_linux_shared
cp libdds.so ../../.

Test

python test_env.py

API

• reset: return the first player's initial state: holding(52d vector) and empty bidding history(35d vecotr).
• step: receive the bidding action and return the state, reward, terminal signal and the next player's seat.

Variable

• bidding_seats: bidding particiants.
• predeal_seats: particiants who can be allocated explicit cards when the environment is reset.
• nmc: number of Monte Carlo tries.
• score: the maximum tricks to win substract the contract target Refer to config.py for more details.

Future

The code about implicit communication through actions will come soon.