user-provided constraint gradients

Author: thowell

examples/particle.jl

@@ -42,6 +42,11 @@ goal(x, u, w) = x - xT
 
 cont = Constraint()
 conT = Constraint(goal, num_state, 0)
+# conT = Constraint(
+#       (c, x, u, w) -> c .= x - xT, 
+#       (cx, x, u, w) -> cx .= 1.0 * I(length(x)), 
+#       (cu, x, u, w) -> cu .= zeros(length(x), length(u)), 
+#       length(xT), num_state, num_action)
 constraints = [[cont for t = 1:T-1]..., conT] 
 
 # ## problem

src/constraints.jl

@@ -48,6 +48,18 @@ function Constraint()
         collect(1:0))
 end
 
+function Constraint(f::Function, fx::Function, fu::Function, num_constraint::Int, num_state::Int, num_action::Int; 
+    indices_inequality::Vector{Int}=collect(1:0), 
+    num_parameter::Int=0)
+
+    return Constraint(
+        f, 
+        fx, fu,
+        num_constraint, num_state, num_action, num_parameter,  
+        zeros(num_constraint), zeros(num_constraint, num_state), zeros(num_constraint, num_action), 
+        indices_inequality)
+end
+
 function constraint!(violations, constraints::Constraints{T}, states, actions, parameters) where T
     for (t, con) in enumerate(constraints)
         con.num_constraint == 0 && continue