search

hanruihua / rl_rvo_nav

The source code of the [RA-L] paper "Reinforcement Learned Distributed Multi-Robot Navigation with Reciprocal Velocity Obstacle Shaped Rewards"
MIT License
180 stars 32 forks source link

result #9

Closed Mealoore closed 1 year ago

Mealoore commented 1 year ago

I have a question about stage_1, why can't the result reach 100%?

policy_name: r4_0_50  successful rate: 12.00% average EpLen: 90.58 std length 9.24 average speed: 1.08 std speed 0.07
policy_name: r4_0_100  successful rate: 51.00% average EpLen: 91.39 std length 12.24 average speed: 0.93 std speed 0.09
policy_name: r4_0_150  successful rate: 39.00% average EpLen: 77.46 std length 7.88 average speed: 1.08 std speed 0.09
policy_name: r4_0_200  successful rate: 84.00% average EpLen: 64.77 std length 5.52 average speed: 1.27 std speed 0.1
policy_name: r4_0_250  successful rate: 72.00% average EpLen: 64.61 std length 5.3 average speed: 1.23 std speed 0.12
hanruihua commented 1 year ago

Hi,

You can try to change the parameter about the reward in line 35 of the file train_process_s1.py.

For example, to improve the success rate, you can increase the the third value of the reward parameter from 0.0 to 0.3