diff --git a/graphics/verlet_and_cell_lists.jl b/graphics/verlet_and_cell_lists.jl
index 8ed2f8a..a858209 100644
--- a/graphics/verlet_and_cell_lists.jl
+++ b/graphics/verlet_and_cell_lists.jl
@@ -18,7 +18,7 @@ function gen_verlet_and_cell_lists_graphics()
     R_particle = 4.2
     σ = 2 * R_particle * 2^(-1 / 6)
     R_interaction = 2^(1 / 6) * σ
-    R_skin = 2 * R_interaction
+    R_skin = ReCo.DEFAULT_SKIN_TO_INTERACTION_RADIUS_RATIO * R_interaction
 
     setcolor("blue")
     reference_particle_x = 0.25 * R_skin
diff --git a/src/RL/RL.jl b/src/RL/RL.jl
index d183029..5da861c 100644
--- a/src/RL/RL.jl
+++ b/src/RL/RL.jl
@@ -66,11 +66,12 @@ function run_rl(;
     n_particles::Int64=100,
     seed::Int64=42,
     ϵ_stable::Float64=0.0001,
-    skin_to_interaction_radius_ratio::Float64=ReCo.DEFAULT_SKIN_TO_INTERACTION_R_RATIO,
+    skin_to_interaction_radius_ratio::Float64=ReCo.DEFAULT_SKIN_TO_INTERACTION_RADIUS_RATIO,
     packing_ratio::Float64=0.15,
     show_progress::Bool=true,
     reward_discount::Float64=0.1,
     show_simulation_progress::Bool=true,
+    n_episodes_before_env_helper_saving::Int64=10,
 ) where {E<:Env}
     @assert 0.0 <= elliptical_a_b_ratio <= 1.0
     @assert n_episodes > 0
@@ -78,6 +79,8 @@ function run_rl(;
     @assert update_actions_at in 0.001:0.001:episode_duration
     @assert n_particles > 0
     @assert 0.0 < ϵ_stable < 1.0
+    @assert 0.0 <= reward_discount <= 1.0
+    @assert n_episodes_before_env_helper_saving > 0
 
     # Setup
     Random.seed!(seed)
@@ -118,6 +121,8 @@ function run_rl(;
 
     parent_dir = "RL/" * process_dir
 
+    env_helper_path = ReCo.DEFAULT_EXPORTS_DIR * "/$parent_dir/env_helper.jld2"
+
     # Pre experiment
     hook(PRE_EXPERIMENT_STAGE, agent, env)
     agent(PRE_EXPERIMENT_STAGE, env)
@@ -149,16 +154,16 @@ function run_rl(;
         hook(POST_EPISODE_STAGE, agent, env)
         agent(POST_EPISODE_STAGE, env)
 
+        if episode % n_episodes_before_env_helper_saving == 0
+            JLD2.save_object(env_helper_path, env_helper)
+        end
+
         ProgressMeter.next!(progress; showvalues=[(:rewards, hook.rewards)])
     end
 
     # Post experiment
     hook(POST_EXPERIMENT_STAGE, agent, env)
 
-    process_dir = ReCo.DEFAULT_EXPORTS_DIR * "/$parent_dir"
-
-    JLD2.save_object("$process_dir/env_helper.jld2", env_helper)
-
     return env_helper
 end
 
diff --git a/src/setup.jl b/src/setup.jl
index 82c84ce..f7f1d4a 100644
--- a/src/setup.jl
+++ b/src/setup.jl
@@ -1,6 +1,6 @@
 const DEFAULT_PACKING_RATIO = 0.5
 const DEFAULT_δt = 1e-5
-const DEFAULT_SKIN_TO_INTERACTION_R_RATIO = 2.5
+const DEFAULT_SKIN_TO_INTERACTION_RADIUS_RATIO = 2.5
 const DEFAULT_EXPORTS_DIR = "exports"
 const DEFAULT_PARENT_DIR = ""
 const DEFAULT_COMMENT = ""
@@ -51,7 +51,7 @@ function gen_sim_consts(
     v₀::Float64;
     δt::Float64=DEFAULT_δt,
     packing_ratio::Float64=DEFAULT_PACKING_RATIO,
-    skin_to_interaction_radius_ratio::Float64=DEFAULT_SKIN_TO_INTERACTION_R_RATIO,
+    skin_to_interaction_radius_ratio::Float64=DEFAULT_SKIN_TO_INTERACTION_RADIUS_RATIO,
     half_box_len::Float64=0.0,
 )
     @assert n_particles > 0
@@ -160,7 +160,7 @@ function init_sim(;
     v₀::Float64,
     δt::Float64=DEFAULT_δt,
     packing_ratio::Float64=DEFAULT_PACKING_RATIO,
-    skin_to_interaction_radius_ratio::Float64=DEFAULT_SKIN_TO_INTERACTION_R_RATIO,
+    skin_to_interaction_radius_ratio::Float64=DEFAULT_SKIN_TO_INTERACTION_RADIUS_RATIO,
     exports_dir::String=DEFAULT_EXPORTS_DIR,
     parent_dir::String=DEFAULT_PARENT_DIR,
     comment::String=DEFAULT_COMMENT,