OCP Solver not being built correctly when infinity in bounds

Hi :wave:

I am currently trying to build an Optimal Control Problem (OCP) in Python using acados and am running into a problem with the solver generation.

The OCP is used to build an MPC algorithm for an inverted pendulum system.

The code itself is implemented as follows:

ocp = AcadosOcp()
 # set model
 model = pend_model()
 ocp.model = model

 Tf = 1.0
 nx = model.x.size()[0]
 nu = model.u.size()[0]
 N = 20

# set dimensions
ocp.dims.N = N

# set cost
Q_mat = 2 * np.diag([1e3, 1e3, 1e-2, 1e-2])
R_mat = 2 * np.diag([1e-2])

ocp.cost.cost_type = 'EXTERNAL'
ocp.cost.cost_type_e = 'EXTERNAL'
ocp.model.cost_expr_ext_cost = model.x.T @ Q_mat @ model.x + model.u.T @ R_mat @ model.u
ocp.model.cost_expr_ext_cost_e = model.x.T @ Q_mat @ model.x

# Input constraints
umax = 23
ocp.constraints.lbu = np.array([-umax])
ocp.constraints.ubu = np.array([+umax])
ocp.constraints.idxbu = np.array([0])

# State Constraints
xmax = [inf, inf, 0.25, inf]
xmin = [-inf, -inf, -0.25, -inf]
ocp.constraints.lbx = np.array(xmin)
ocp.constraints.ubx = np.array(xmax)
ocp.constraints.idxbx = np.array([0, 1, 2, 3])

# Initial state constraint
ocp.constraints.x0 = np.array([np.pi, 0.0, 0.0, 0.0])

# set options
ocp.solver_options.qp_solver = 'PARTIAL_CONDENSING_HPIPM' 
ocp.solver_options.hessian_approx = 'EXACT' 
ocp.solver_options.integrator_type = 'IRK'
ocp.solver_options.print_level = 1
ocp.solver_options.nlp_solver_type = 'SQP' 

# set prediction horizon
ocp.solver_options.tf = Tf

self.ocp = ocp
self.ocp_solver = AcadosOcpSolver(self.ocp, json_file='acados_ocp.json')

On further investigation, I noticed that the AcadosOCPSolver class did not generate a solver when I included my state constraints.

While trying to run the code this error message pops up

Error: Custom { kind: InvalidData, error: Error("invalid number", line: 46, column: 14) }
Traceback (most recent call last):
  File "/home/s/pg22-pendulum/Real_Pendulum/OCP_Test.py", line 92, in <module>
  File "/home/s/pg22-pendulum/Real_Pendulum/OCP_Test.py", line 87, in main
    solver = ctrl.build_ocp()
  File "/home/s/pg22-pendulum/Real_Pendulum/OCP_Test.py", line 80, in build_ocp
    self.ocp_solver = AcadosOcpSolver(self.ocp, json_file='acados_ocp.json')
  File "/home/s/acados/interfaces/acados_template/acados_template/acados_ocp_solver.py", line 937, in __init__
    self.generate(acados_ocp, json_file=json_file, simulink_opts=simulink_opts, cmake_builder=cmake_builder)
  File "/home/s/acados/interfaces/acados_template/acados_template/acados_ocp_solver.py", line 856, in generate
    ocp_render_templates(acados_ocp, json_file, cmake_builder=cmake_builder, simulink_opts=simulink_opts)
  File "/home/s/acados_template/acados_template/acados_ocp_solver.py", line 743, in ocp_render_templates
    render_template(tup[0], tup[1], output_dir, json_path)
  File "/home/s/acados/interfaces/acados_template/acados_template/utils.py", line 241, in render_template
    raise Exception(f'Rendering of {in_file} failed!\n\nAttempted to execute OS command:\n{os_cmd}\n\n')
Exception: Rendering of main.in.c failed!

I tried building the tera renderer natively using the latest git release and replaced the executable in the bin folder but this doesn’t change the outcome
Any help would be greatly appreciated :pray:

Hi, have you tried to change the inf in your state constraints to a large number, e.g. 1e10? Refer to this post: Time varying constraints, reference etc with acados in matlab - #4 by FreyJo


Yes, it works after I removed the inf. I think it is infinity by default and I don’t have to specify the constraint unless it’s there…