Computation-Aware SDV Testbed

This repository contains a modular MATLAB/Simulink testbed for longitudinal speed tracking and lateral path tracking. A mid-fidelity vehicle model is used as the plant, while a lower-order dynamic bicycle model is used for MPC prediction.

Phase 0 currently supports two controller backends:

• acados nonlinear MPC
• MATLAB Adaptive MPC

The fixed linear MATLAB MPC backend is reserved but is not yet implemented.

Requirements

The project was developed with MATLAB R2026a on Windows. The following MATLAB products are required:

• MATLAB
• Simulink
• Control System Toolbox
• Model Predictive Control Toolbox

The acados backend additionally requires:

• Git
• CMake
• MATLAB MinGW-w64 C compiler support

The automated acados installation currently supports Windows only.

Clone and open the project

Clone the repository and open MATLAB in the repository root:

openProject("computation-aware-sdv.prj");

The project paths can also be initialized manually:

startup_project;

Select the controller backend

Open:

common/configuration/controller_parameters.m

Select one backend:

controller.controller_backend = controller.BACKEND_ACADOS;

or:

controller.controller_backend = controller.BACKEND_ADAPTIVE;

The backend identifiers are:

BACKEND_ACADOS  = 1
BACKEND_MATLAB  = 2   (not implemented)
BACKEND_ADAPTIVE = 3

Run all Phase 0 scenarios

From the repository root, run:

run("phase0/main.m");

The script checks the required MATLAB products and then runs:

1. Urban profile
2. Aggressive maneuver
3. Highway cruise

If Adaptive MPC is selected, no acados installation is required.

If acados is selected and either acados or the generated S-function is missing, MATLAB asks:

Install acados and generate the S-function now? [y/N]:

Enter y to continue. The first installation may take longer than an ordinary simulation run.

Manual acados setup

The same setup can be performed explicitly:

startup_project;
setup_acados_dependency;
s_fun_generation_acados;

The project installs the pinned acados v0.5.4 dependency outside the repository:

%LOCALAPPDATA%/SDV/dependencies/acados/v0.5.4

The generated solver is placed under:

build/acados/sdv_dynamic_bicycle/

Regenerate the S-function after changing the prediction dynamics, horizon, cost structure, constraints, solver options, or generated interface ports. Changing only the scenario reference or road profile does not require regeneration.

Run one scenario

After running startup_project, use one of:

run_urban_profile;
run_aggressive_maneuver;
run_highway_cruise;

Generated outputs

Scenario results and figures are saved by controller backend:

phase0/results/<controller>/<scenario>/
phase0/figures/<controller>/<scenario>/

Each run stores simulation results, tracking-error metrics, and figures. acados runs additionally store solve-time statistics and telemetry.

Troubleshooting

Missing MATLAB product

Install or activate the product reported by check_runtime_requirements.

No compatible C compiler

Install the MATLAB MinGW-w64 support package, then select it:

mex -setup C

acados is missing

setup_acados_dependency;

Generated acados S-function is missing

s_fun_generation_acados;

MATLAB is using stale generated code

Close the Simulink models and unload generated MEX files before rebuilding:

bdclose("all");
clear mex;
s_fun_generation_acados;

Reproducibility note

The Phase 0 regeneration time assumes MATLAB and the required MathWorks products are already installed. First-time acados installation and S-function compilation are dependency-provisioning steps and are not included in the ordinary baseline simulation time.