search

rr-learning / CausalWorld

CausalWorld: A Robotic Manipulation Benchmark for Causal Structure and Transfer Learning
https://sites.google.com/view/causal-world/home
MIT License
210 stars 26 forks source link

reward wrong in reaching task #79

Closed martius-lab closed 4 years ago

martius-lab commented 4 years ago

In 'get_reward' in base_task you add the goal_distance to the reward. Given that you typically try to maximize the reward in the RL setting, shouldn't be the negative distance added to the reward?

ossamaAhmed commented 4 years ago

Sorry for the late reply and thanks for pointing this out. Its actually the goal_reward instead which means 0 there is no overlap with the goal and 1 means full overlap with the goal (i.e succeeding in the task). We will change the naming accordingly.