Add terminal constraint support to IPDDP

.gitignore

@@ -16,6 +16,9 @@ neural_models/
 examples/c_generated_code/
 examples/*_acados_ocp.json
 
+# Generated demo output (gifs, json dumps)
+examples/out_ipddp_mpcc_rc/
+
 # PDF images
 pdf_images/
 

CMakeLists.txt

@@ -20,7 +20,7 @@ set(CMAKE_POLICY_VERSION_MINIMUM "3.15" CACHE STRING "Minimum CMake version for
 
 project(
   cddp
-  VERSION 0.5.1
+  VERSION 0.5.2
   DESCRIPTION "CDDP: A C++ library for Trajectory Optimization and MPC"
   HOMEPAGE_URL  "https://github.com/astomodynamics/cddp-cpp"
 )
@@ -147,7 +147,9 @@ set(cddp_core_srcs
   src/cddp_core/dynamical_system.cpp
   src/cddp_core/objective.cpp
   src/cddp_core/constraint.cpp
+  src/cddp_core/cddp_context_utils.cpp
   src/cddp_core/helper.cpp
+  src/cddp_core/interior_point_utils.cpp
   src/cddp_core/boxqp.cpp
   src/cddp_core/qp_solver.cpp
   src/cddp_core/cddp_core.cpp

README.md

@@ -20,6 +20,10 @@ $$
 \quad \mathbf{g}(\mathbf{x}_k,\mathbf{u})_k\leq 0
 $$
 
+$$
+\quad \mathbf{h}(\mathbf{x}_N)=0,\;\mathbf{g}_N(\mathbf{x}_N)\leq 0
+$$
+
 $$
 \quad {\mathbf{x}}_{0} = \mathbf{x}{(t_0)} 
 $$
@@ -28,6 +32,13 @@ $$
 \quad k = 0,1,\cdots N-1
 $$
 
+For interior-point solves, `IPDDP` supports path inequalities, terminal
+equalities, terminal inequalities, and mixed path/terminal constrained
+problems with warm-started multiplier and slack state. In the C++ API,
+terminal constraints are attached with `CDDP::addTerminalConstraint(...)`
+and should use the built-in `TerminalEqualityConstraint` or
+`TerminalInequalityConstraint` classes.
+
 ## Examples
 The maintained example surface is now split:
 
@@ -81,6 +92,10 @@ The MPCC portfolio example is a lightweight kinematic contouring-control demo.
 Its vendored track data in `examples/data/` is derived from the
 [`alexliniger/MPCC`](https://github.com/alexliniger/MPCC) project.
 
+The portfolio currently emphasizes animated path-constrained examples. The
+terminal-constraint cases added to IPDDP are covered by the regression suite
+rather than by a dedicated gallery animation.
+
 ## Installation
 ### Dependencies
 * [CMake](https://cmake.org/) (Build System)

docs/cpp.md

@@ -30,3 +30,17 @@ That lets the repository serve both as:
 
 - a source build for C++ consumers
 - the native backend for the `pycddp` wheel
+
+## Constraint coverage
+
+`IPDDP` is the interior-point single-shooting solver in the C++ API. The
+current implementation supports:
+
+- path inequality constraints
+- terminal equality constraints via `TerminalEqualityConstraint`
+- terminal inequality constraints via `TerminalInequalityConstraint`
+- mixed path and terminal constrained problems
+
+Attach terminal constraints on `CDDP` with `addTerminalConstraint(...)`. IPDDP
+rejects unsupported terminal-constraint subclasses instead of silently ignoring
+them.

docs/index.md

@@ -5,6 +5,7 @@
 - a C++17 core library for trajectory optimization and MPC
 - `pycddp` Python bindings built with `pybind11`
 - a small animation-oriented Python portfolio for demos and regression checks
+- IPDDP support for path constraints plus terminal equality and terminal inequality constraints
 
 Use the navigation to get started with installation, local development, and
 the release workflow for PyPI and GitHub Pages.

docs/python.md

@@ -7,6 +7,22 @@ The Python package is published as `pycddp`.
 - the native extension module `_pycddp_core`
 - the public package namespace `pycddp`
 - version metadata from `pycddp._version`
+- `CDDP` problem bindings with `add_constraint(...)` and `add_terminal_constraint(...)`
+
+## Solver coverage
+
+The Python package follows the same solver capability as the C++ core. For
+`IPDDP`, the regression suite now covers:
+
+- path-only constrained problems
+- terminal-inequality-only problems
+- terminal-equality-only problems
+- mixed path + terminal-equality problems
+- warm-start reuse of path, terminal-inequality, and terminal-equality state
+
+The current Python portfolio focuses on animated demos rather than terminal
+constraint examples, so terminal-constrained IPDDP coverage is documented here
+and exercised in tests.
 
 ## Local validation
 

docs/python_portfolio.md

@@ -13,6 +13,11 @@ Matplotlib.
       actually produce: constrained solves, animated trajectories, and compact
       reproducible workflows.
     </p>
+    <p>
+      The gallery is intentionally path-focused. Terminal-equality and
+      terminal-inequality support in <code>IPDDP</code> is covered in the solver
+      regression suite rather than in a separate animation card.
+    </p>
     <p>
       <a class="md-button md-button--primary" href="../python/">Python package guide</a>
       <a class="md-button" href="https://github.com/astomodynamics/cddp-cpp/tree/master/examples">Browse examples</a>