From 5d1de8684cc012a654a30c0fbd36abcb42b8c8b0 Mon Sep 17 00:00:00 2001
From: Erik Nygren <erik.nygren@sbb.ch>
Date: Fri, 5 Jul 2019 15:27:35 -0400
Subject: [PATCH] fixed reward function bug where taking illegal actions at
split switch were rewarded!
---
flatland/envs/rail_env.py | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/flatland/envs/rail_env.py b/flatland/envs/rail_env.py
index b4a56a8..abc8a73 100644
--- a/flatland/envs/rail_env.py
+++ b/flatland/envs/rail_env.py
@@ -231,7 +231,7 @@ class RailEnv(Environment):
self.rewards_dict[iAgent] += stop_penalty
if not agent.moving and not (action == RailEnvActions.DO_NOTHING or action == RailEnvActions.STOP_MOVING):
- # Only allow agent to start moving by pressing forward.
+ # Allow agent to start with any forward or direction action
agent.moving = True
self.rewards_dict[iAgent] += start_penalty
@@ -270,6 +270,7 @@ class RailEnv(Environment):
else:
# TODO: an invalid action was chosen after entering the cell. The agent cannot move.
self.rewards_dict[iAgent] += invalid_action_penalty
+ self.rewards_dict[iAgent] += step_penalty * agent.speed_data['speed']
agent.moving = False
self.rewards_dict[iAgent] += stop_penalty
@@ -277,6 +278,7 @@ class RailEnv(Environment):
else:
# TODO: an invalid action was chosen after entering the cell. The agent cannot move.
self.rewards_dict[iAgent] += invalid_action_penalty
+ self.rewards_dict[iAgent] += step_penalty * agent.speed_data['speed']
agent.moving = False
self.rewards_dict[iAgent] += stop_penalty
--
GitLab