Quarter-Car Active Suspension — Adams–Simulink Co-Simulation

A multibody quarter-car (Adams) coupled to a MATLAB/Simulink semi-active skyhook controller. Compares skyhook against a passive suspension on step and random roads, and examines the ride-comfort vs road-holding trade-off.

The quarter-car plant (sprung mass, unsprung mass, suspension, PAC2002 tire) is built in Adams and exported with Adams/Controls as a co-simulation plant. Simulink runs the controller and exchanges signals with the Adams Solver every communication interval. The same model is switched between passive and semi-active skyhook control, and between a step and a random (ISO 8608) road, so all four cases run from one script.

Skills demonstrated: multibody dynamics (Adams), Adams–Simulink co-simulation, semi-active control (skyhook), MATLAB scripting & signal processing, and interpreting the comfort vs road-holding trade-off from the results.

RMS metrics over a 10 s run (signals are in MMKS: mm, mm/s²):

Skyhook lowered the RMS sprung-mass acceleration (comfort) by ~38% on the step road and ~48% on the random road. On the step road road-holding also improved slightly; on the random road, road-holding was worse than passive — a trade-off discussed below. Suspension stroke increased modestly, as expected for an actively worked suspension.

An ideal skyhook law commands zero force whenever the damper cannot dissipate in the required direction (the "off" quadrant, where sprung_vel · susp_vel ≤ 0). During those intervals the unsprung mass is left essentially undamped, which raises the dynamic tire load in the wheel-hop band. A passive damper, by contrast, provides damping at all times, so it holds the road better in this respect.

A sweep of the skyhook coefficient cs_s confirmed the effect is monotonic and structural — road-holding improves as cs_s increases but never reaches the passive level — which is consistent with this mechanism rather than an implementation error.

A semi-active implementation that keeps a small minimum off-state damping (cs_s_min) would mitigate this; it is listed under future work.

 road input            control force (skyhook)
     |                          ^
     v                          |
 +----------------+   y   +-------------------+
 |  Adams         | ----> |  Simulink         |
 |  plant         |       |  skyhook / passive|
 | (Adams Solver) | <---- |                   |
 +----------------+   u   +-------------------+
   Adams/Controls           adams_sub S-Function
   (co-simulation)          (Adams Solver runs in its own process, separate from Simulink)

Two base-workspace switches select the case:

• use_skyhook : 0 = passive, 1 = skyhook
• use_random : 0 = step road, 1 = random road

.
├── adams/                  # Adams quarter-car model + Adams/Controls export
├── simulink/               # co-simulation model (.slx)
├── matlab/                 # init.m (setup), postprocess.m (metrics & plots), sweep_cs_s.m
└── results/                # figures and demo animation

Built with Adams 2024.1 and MATLAB/Simulink R2024b.

1. Open simulink/quarter_car.slx. Its PreLoadFcn runs init.m, which sets the parameters, generates the road profiles, and loads the Adams/Controls plant.
2. Run matlab/postprocess.m. Runs all four cases (2 roads × passive/skyhook), prints the metric tables, and plots the comparisons.
3. (optional) Run matlab/sweep_cs_s.m to trace road holding and comfort against the skyhook coefficient.

• Controller: semi-active skyhook, F = -cs_s · sprung_vel, clipped to the dissipative quadrant (a real damper can only remove energy).
• Excitation: the road is applied at the tire–road contact (road → body), the physically correct input for a ride study.
• Units: the Adams model uses the MMKS system (mm, kg, N, s); logged signals are in mm and mm/s² (divide by 1000 for SI).

• Add a minimum off-state damping (cs_s_min) to the semi-active law to recover road holding on broadband roads.
• Smooth the semi-active switching to remove the acceleration spikes at damper on/off.
• Extend to a full-vehicle model (Adams/Car); add a braking / cornering scenario.
• Add Python automation for parameter sweeps and sensitivity analysis.

MIT — see LICENSE.

Takuma Matsuda — multibody dynamics & CAE. LinkedIn: www.linkedin.com/in/takuma-matsuda