Linearize the car to read its stability
Generated from
content/lms/vehicle-dynamics-ii-theory/04-transient-vehicle-response/02-stability-derivatives-and-linearization.md; edit the source file, not this page.
Source path: content/lms/vehicle-dynamics-ii-theory/04-transient-vehicle-response/02-stability-derivatives-and-linearization.md
Course: Read the forces that steer the car
Module: Model the car's reaction to sudden inputs
Estimated duration: 55 minutes
Linearizing the car is not a trick for making vehicle dynamics look cleaner on paper. It is the skill of freezing a complicated car at one operating condition, asking a small and precise question, and then reading whether the car tends to settle, diverge, or oscillate when disturbed. For an intermediate driver, this matters because the words stable, understeer, oversteer, responsive, and nervous are often used as if they describe the whole car. They do not. The bonded material is very clear on that point: vehicle behavior changes with environment and trim, and stability has to be examined separately for each environment and trim. In this lesson, trim means the operating condition you are asking about, specified in part by steering angle, forward velocity, and lateral acceleration.
The principle is simple: do not ask whether the car is stable in the abstract. Ask whether this car, at this speed, at this lateral acceleration, with this steering condition, returns toward equilibrium after a small disturbance. That is the useful mental move behind linearization. You are not pretending the car is actually linear everywhere. You are choosing a small neighborhood around one trim state so the first response can be read cleanly. If you change the speed, the lateral acceleration, the tire operating point, the road surface, or the vehicle parameters, you have changed the question. A car can be stable under one set of operating conditions and unstable in another. It can also be understeer for small inputs and oversteer for large inputs because the real vehicle is nonlinear and does not keep the same characteristics at all trims.
That is why linearization belongs in transient response. A steady-state handling curve tells you how much steering is needed for a certain lateral acceleration at a given path radius. A transient stability read asks what happens after you disturb the car away from that condition. The two are related, but they are not the same job. If you are careless, you will mix them together and decide that a car with an understeer label must always be safe, or that a car that rotates well in one corner must be unstable everywhere. The better method is narrower. First define the trim. Then make the disturbance small enough that you are reading the local answer. Then identify the mode of the response.
The smallest readable model in the chunks is the bicycle model. The pair of real wheels on each axle is compressed into one front wheel and one rear wheel on the centerline. That simplification removes track width, roll details, and many packaging details, but it keeps the important plan-view balance: the front axle develops a lateral force, the rear axle develops a lateral force, the center of mass sits between them, and the yaw moment depends on how those forces balance. The point is not that a car behaves like a bicycle. The source explicitly warns that the name does not imply bicycle-like handling. The point is that the bicycle model is the cleanest plan-view object for seeing the balance between front and rear.
Start with the steady-state idea. At a given radius, the steer angle can be thought of as the sum of the kinematic steer angle and the understeer angle. The kinematic part is the angle required by geometry when lateral acceleration is nominally zero. The understeer angle, also called the dynamic steer angle in the source, is the extra steering required for the lateral acceleration. That separation is one of the most useful driver-level ways to keep the model honest. If you need more steering as lateral acceleration rises at the same path radius, something about the dynamic balance is asking for more front slip angle relative to rear. If you need less steering as the demand rises, or if the required steering begins to fall in the region you are testing, the local balance is different. The curve is not just a drawing. It is a way to detect which axle is becoming the limiting or more compliant participant in the cornering force balance.
Now connect that steady-state curve to stability. Understeer gradient, characteristic speed, critical speed, response factor, curvature gain, yaw gain, acceleration gain, yaw stiffness, static margin, and yaw damping all appear in the bonded corpus as part of the linear handling vocabulary. You do not need to turn every one of those into a cockpit calculation. You do need to understand what they are trying to read. The understeer gradient gives a compact measure of steady-state handling behavior in the linear case. Characteristic speed belongs to the understeer side of the story, where more speed changes the response in a way that remains bounded in the linear reading. Critical speed belongs to the oversteer stability limit, where the straight-ahead or disturbed response can become divergent. Static margin locates the relationship between the vehicle center of gravity and the neutral steer point. Yaw damping describes whether yaw motion is naturally resisted or whether the car has little ability to kill yaw-rate energy after it starts.
A useful way to hold those ideas together is to think in questions. Understeer gradient asks how the required steering changes with lateral acceleration in the region being tested. Static margin asks where a side force could act without producing steady-state yaw velocity. Yaw damping asks whether yaw motion is damped as the car responds. Critical speed asks whether an oversteer vehicle has reached a speed where a normal disturbance can become a stability problem. The characteristic equation language in the source is the mathematical version of the same question. A linear vehicle analog is divergently unstable when the characteristic equation has a positive real root. It is oscillatorily unstable when complex roots have positive real parts. In driver language, one failure runs away in one direction, and the other keeps changing direction with growing amplitude.
Divergent instability is the one most drivers imagine first. The source gives the example of operating above the critical speed of an oversteer vehicle. Any steering input places the vehicle in a turn of ever-decreasing radius unless the driver makes compensating steering motions to maintain equilibrium. That is not just oversteer as a corner-exit attitude. It is a local stability problem. The linearized read says the trim does not naturally come back. Once nudged, it tends to depart.
Oscillatory instability is different. The source describes free-control response after a pulse displacement or force at the steering wheel. Some vehicles turn first one way, then the other, and continue until the amplitude increases enough for a spinout. The important distinction is that the car is not simply tightening one turn radius. It is failing to settle. In a transient response lesson, this is a different signature and it deserves a different diagnosis. A car can feel lively and still be well damped; it can also feel busy because the yaw and sideslip response are not dying out cleanly.
This is where the local nature of the linear read becomes non-negotiable. Temperature can change shock absorber damping characteristics. A slippery road surface can change tire cornering properties. Forward velocity and lateral acceleration strongly affect the stability question. If you test a car in cool morning conditions, then judge its afternoon response on a hotter surface as though nothing changed, you are no longer reading the same trim. If you test the car at mild lateral acceleration and then assume the same character near the tire limit, you are overextending the linear answer. The lesson is not that the first answer is useless. The lesson is that the first answer is local.
To perform the skill, begin by naming the trim in ordinary language before you think about equations. For example: steady constant-radius cornering at moderate speed and low lateral acceleration. Or straight-ahead high speed with small steering disturbances. Or high-speed transient maneuver with a four-wheel-steer system active. The bonded material supports all three as legitimate stability contexts. Once the trim is named, hold the major conditions as steady as you can. You are trying to see what the car does after a small change, not after five different things changed at once.
Second, separate the kinematic steering demand from the dynamic steering demand. On a constant radius, some steering is required by geometry. That part by itself does not tell you whether the car is understeer or oversteer. The dynamic part is the extra or reduced steering associated with lateral acceleration and tire force balance. This is why the basic handling curve plots steering angle against lateral acceleration at a given path radius. If you do not hold the radius idea steady, the curve becomes much harder to read because geometry and dynamics are moving together.
Third, use small disturbances. A linear handling vehicle is a local idea. Small steering, small yaw deviation, and small sideslip changes allow the front and rear tire force balance to be read as a first response. Once the input is large enough to move the tires into a different nonlinear region, you have changed the trim and the answer may change. The bonded source explicitly notes that a vehicle may be understeer for small inputs and oversteer for large inputs, or the opposite. That single sentence should stop you from making broad labels too early.
Fourth, identify the response mode. A stable response moves toward a new steady state or back toward the original one after the disturbance, depending on the input. A divergent response tightens or departs unless the driver supplies compensating steering. An oscillatory unstable response alternates direction and grows instead of dying away. A lightly damped but stable response may oscillate once or twice and then settle, which is different from an oscillation that builds. This distinction is why yaw damping matters. You are not only asking which direction the car initially yaws. You are asking whether yaw motion is resisted, sustained, or amplified.
Fifth, repeat the read at another trim instead of generalizing. The most common error in this topic is to test one place on the handling map and name the whole car. The bonded corpus gives you permission to be more precise. Stability must be examined separately for each environment and trim. The conditions that most affect stability are the steady-state values of forward velocity and lateral acceleration. That means your read at low lateral acceleration is evidence for low lateral acceleration. Your read at high lateral acceleration must be earned separately.
The front-rear mechanism is the driver-friendly core. In the bicycle model, steering can create a front slip angle, yaw can create rear slip angle, and equilibrium requires the yaw moment to balance so the yaw speed is constant in steady state. Handling depends on the balance between front and rear. The chunks also explain that many design factors influence the cornering force developed at a wheel, and any design factor that changes that force has a direct effect on directional response. Suspensions and steering systems are called out as primary sources of these influences. You therefore should not treat tire cornering stiffness as the only real cause. It is the clean basis for the model, not the only physical lever in the car.
Static margin gives another way to visualize the same balance. The neutral steer point is the point on the vehicle where a side force produces no steady-state yaw velocity. The neutral steer line extends that idea through the x-z plane. For driver use, the key idea is that a side force applied at different places relative to the vehicle can produce different yaw tendencies. Understeer and oversteer are not magic labels. They are consequences of where lateral force, compliance, tire behavior, and mass distribution leave the yaw moment balance.
Yaw damping is the next layer. A car can have an initial yaw response that looks acceptable but still be poor at damping the motion. In a transient maneuver, this matters because your steering input is not the only event. Road disturbances, camber changes, compliance effects, and driver corrections can all add small disturbances. The bonded material asks for yaw damping to be explained and graphed for a simple linear vehicle, and it asks how yaw damping typically depends on speed and lateral acceleration. That tells you the concept belongs with stability, not only with comfort or feel. You read yaw damping by watching whether yaw-rate and sideslip motion decay after a disturbance.
The four-wheel-steer chunk adds an important caution. A properly implemented four-wheel-steer system can make a vehicle more maneuverable at low speeds and more responsive and stable in high-speed transient maneuvers. In other high-speed driving, its presence may be imperceptible. That means you cannot judge the stability derivative story from parking-lot maneuverability alone. A system that changes rear steer with speed can alter the high-speed transient read while feeling almost invisible in steady high-speed driving. For this lesson, the takeaway is not to become a four-wheel-steer engineer. It is to respect trim and operating condition. The same hardware can matter a lot in one region and barely announce itself in another.
The working procedure for a driver or engineer is therefore a disciplined loop. Define the trim. Choose the simplified model that keeps the front-rear force balance visible. Separate geometric steering from dynamic steering. Apply or observe a small disturbance. Read the first response and the settling behavior. Classify the local response as stable, divergent, oscillatory, or changing with input size. Then change only one major trim variable and repeat. That is how you linearize the car to read stability without pretending the full car is simple.
Sub-skill one is choosing the trim. A vague trim such as cornering hard is not enough. A useful trim includes speed range, lateral acceleration range, steering condition, and environment. You can say low lateral acceleration on a constant radius, or high-speed straight-ahead with small steering disturbance, or moderate lateral acceleration after a step steer. The exact wording is less important than the discipline. If the condition changes enough to change tire cornering properties or damping behavior, it is a different read.
Sub-skill two is keeping the input small. Small does not mean timid in the everyday sense. It means small relative to the operating region you are trying to linearize. If you are reading the car near a mild steady-state cornering trim, a large steering yank is not a better test. It pushes the car into a different part of the tire force curve and may produce a response that belongs to large-input nonlinear handling rather than the local trim. When you want the local stability answer, you ask a local question.
Sub-skill three is seeing front and rear separately. Understeer in the linear read is not just the nose sliding wide. Oversteer is not just the rear stepping out. In the model, the front and rear axles each need slip angle to make their lateral force. If the front axle is more compliant than the rear, a lateral disturbance produces more front sideslip and the vehicle response follows the understeer pattern. If the rear requires the greater slip or loses the balance sooner, the response moves toward oversteer. The driver version is to ask which end is demanding the correction and whether the steering you add is creating balance or chasing a departure.
Sub-skill four is separating steady-state evidence from transient evidence. A basic handling curve at a given radius tells you how steering demand varies with lateral acceleration. That is powerful, but it does not by itself tell you whether the yaw motion after a disturbance is well damped. For transient stability you must watch the time behavior: does the response settle, build, or reverse and grow. The sibling lessons on planar rigid-body equations, body-axis equations, step-steer response, and steering frequency response surround this skill. This lesson sits between the static map and those time-domain or frequency-domain tools. It teaches you what question those tools are answering.
Sub-skill five is refusing global labels. The source material says the vehicle may be stable under one operating condition and unstable under another. It also says the vehicle can be understeer for small inputs and oversteer for large inputs. Once you accept those two facts, broad single-word handling descriptions become less useful. You can still say a car tends to understeer in a certain region. You should not say the car is simply understeer as if the word covers every speed, lateral acceleration, and input size.
Calibration cues tell you whether you are improving at the skill. The first cue is cleaner language. You stop saying the car is unstable and start saying the high-speed straight-ahead trim with small steering disturbances shows divergent oversteer behavior above the critical-speed region. You stop saying the setup is fine and start saying the low-lateral-acceleration constant-radius curve shows understeer, but the larger-input region needs a separate read. Better language is not academic vanity here. It prevents wrong setup and driving conclusions.
The second cue is cleaner data selection. On a skidpad or constant-radius exercise, you compare steering angle against lateral acceleration at the same radius instead of mixing different radii and calling it one handling curve. On a transient exercise, you look for the first yaw response and the decay or growth after the input instead of only asking whether the car felt quick. On a high-speed straight-line read, you treat small steering disturbances as stability evidence only for that speed and surface condition.
The third cue is correction timing. When you understand divergent instability, you know why the driver has to keep making compensating steering motions to hold equilibrium above the critical-speed condition of an oversteer vehicle. When you understand oscillatory instability, you know why chasing each reversal can feed a motion that is already failing to damp. This lesson does not teach hand technique or recovery technique in detail, but it gives you the diagnostic difference. One problem runs away. The other reverses and grows.
The fourth cue is humility at higher lateral acceleration. A driver who has learned this lesson stops treating the first clean low-speed read as proof of the whole map. As lateral acceleration rises, the car can move from primary understeer to final oversteer or through other combinations of steering behavior. The Dixon chunks explicitly ask for handling steer curves with primary, secondary, and final steer combinations, and they ask how characteristic or critical speed can be applied to a nonlinear vehicle in a given trim state. That is the mindset: each region gets its own read.
There are several failure modes. The first is using a large input and calling the result linear. This gives you a dramatic answer, but it may not be the local answer. The fix is to repeat with a smaller input at the same trim and see whether the classification holds. The second is confusing geometric steering with understeer. On a curved path, steering angle is partly required by geometry. The fix is to compare the extra steering demand as lateral acceleration changes at a given radius. The third is ignoring yaw damping. A car can point quickly and still fail to settle well. The fix is to watch what happens after the first response. The fourth is assuming the environment is neutral. Temperature and road surface can change damping and tire cornering properties, so the same car can deserve a different read later in the day.
The fifth failure mode is treating four-wheel steer, compliance steer, suspension effects, or steering-system effects as distractions from the pure tire model. The chunks do not support that simplification. Tire cornering stiffness is the basis for developing the understeer and oversteer equations, but multiple vehicle design factors influence cornering force and directional response. Suspension and steering systems are primary sources. The correct use of the simple model is to reveal the balance, not to deny the mechanisms that move it.
The sixth failure mode is believing that final understeer automatically means a good transient response. The Dixon material notes that suspension tuning is basically a matter of achieving the desired understeer gradient and final understeer number, and it discusses final understeer in the context of keeping commercial vehicles away from rollover limits. That is a design aim in a particular context, not a blanket proof that every transient response is well damped. You still need to ask the local stability question.
When you finish this lesson, the practical standard is this: given a cornering or straight-line operating condition, you can state the trim, choose the local linear question, explain which front-rear balance terms matter, predict the kind of stability signature you are looking for, and avoid claiming more than the local answer supports. That is what it means to linearize the car to read its stability.
Worked example: high-speed straight-ahead oversteer above critical speed
Imagine the trim is straight-ahead travel at high speed in an oversteer vehicle. The disturbance is small: a steering-wheel displacement, a force pulse at the wheel, or another normal disturbance that moves the car away from its straight-running equilibrium. In the bonded material, oversteer vehicles have a stability limit at critical speed due to normal disturbances in straight-ahead travel. Above that region, any steering input can place the vehicle in a turn of ever-decreasing radius unless the driver supplies compensating steering to hold general equilibrium. The local linear read is divergent instability. The important part is that the problem is not merely that the car can oversteer in a corner. The trim itself has lost the tendency to stay bounded after a small disturbance. Your diagnostic statement should therefore include the speed condition, the oversteer character, and the divergence signature. A vague statement such as the rear is loose is too weak for this lesson.
Worked example: constant-radius handling curve with small-input understeer and large-input oversteer
Set the trim as constant-radius cornering and build the read from the basic handling curve: steer angle against lateral acceleration at that radius. At low lateral acceleration, the steering angle is the kinematic demand plus the dynamic understeer angle. If the extra steering required rises with lateral acceleration, the local small-input region reads as understeer. But the bonded material warns that a nonlinear vehicle can be understeer for small inputs and oversteer for large inputs, or the opposite. That means you do not extend the low-lateral-acceleration line into the limit region by faith. You test the higher lateral-acceleration trim separately. In driver language, the car may ask for more steering early in the cornering range and then begin to reduce the steering margin or rotate as the tire force balance changes. The correct conclusion is not that the car is contradictory. The correct conclusion is that the local linear answer changed with trim and input size.
Worked example: four-wheel steer in a high-speed transient maneuver
Now choose a trim involving a properly implemented four-wheel-steer system. The bonded chunk says such a system can make a vehicle more maneuverable at low speeds and more responsive and stable in high-speed transient maneuvers, while being imperceptible in other high-speed driving. That is a useful example because it punishes lazy testing. If you judge the system only from low-speed maneuverability, you are not reading the high-speed transient stability effect. If you cruise at high speed and feel nothing obvious, you still have not proved it has no transient role. The local question must match the operating region where the system changes the response. In this case, the lesson is to define the trim as a high-speed transient maneuver and look at lateral acceleration response and sideslip behavior after the steering input, not at parking-lot feel or steady cruising impressions.
Common mistakes
The first mistake is the global label error: calling a car understeer, oversteer, stable, or unstable without naming the trim. Good work names speed, lateral acceleration, steering condition, and environment. The second mistake is the big-input error: using a dramatic steering input and treating the response as the linear answer. Good work uses a disturbance small enough to stay local to the trim. The third mistake is the geometry error: counting all steering angle as understeer. Good work separates kinematic steer angle from the dynamic understeer angle at a given path radius. The fourth mistake is the first-response error: judging only the initial yaw and ignoring whether the motion damps, diverges, or oscillates. Good work watches the settling behavior. The fifth mistake is the hardware simplification error: treating suspension and steering effects as irrelevant because the tire model is clean. Good work uses the simple model to expose front-rear balance while remembering that many design factors change cornering force and directional response.
Drill: three-trim stability read
At your next suitable test day, skidpad session, simulator session, or data-review session, perform a three-trim read. First, choose one constant-radius cornering trim at low lateral acceleration. Hold the radius idea steady and record or estimate steering angle as lateral acceleration changes modestly. Success means you can separate the kinematic steering demand from the extra dynamic steering demand and state whether the local region reads understeer or oversteer. Second, choose a similar radius at a higher lateral acceleration, still within the rules and safety margin of the event. Success means you do not assume the first read still applies; you state whether the local answer changed. Third, choose a small transient steering disturbance at a defined speed and watch the response after the input. Success means you classify the motion as settling, divergent, or oscillatory, and you explicitly avoid using one trim to name the whole car. Do three clean repetitions per trim. Stop the drill if you cannot hold the conditions consistently enough to compare the repetitions.
When this principle breaks down
The principle breaks down when you ask the local linear model to answer a nonlocal question. It breaks down when the input is large enough to move the tires into a different force region, when the speed or lateral acceleration changes substantially during the read, when temperature or road surface changes the vehicle parameters, or when you use one operating condition to describe another. It also breaks down when you ignore nonlinear combinations such as small-input understeer with large-input oversteer. The recovery is not to abandon linearization. The recovery is to narrow the question again. Define the new trim, make the disturbance small relative to that trim, and read the response there.
Author Review
No quiz questions are attached to this lesson.
Sources
| # | Document | Chunk | Pages | Score | Collection |
|---|---|---|---|---|---|
| 1 | Fundamentals of vehicle dynamics Gillespie T. D. Thomas D. | 1f b50cf5-cd2b-1a40-fc47-8a12e487b515 | 257 | 1 | uio_books_raw_v1 |
| 2 | Tires Suspension and Handling Second Edition Dixon John C | 86d534d1-ae84-2f7d-df95-48c6a61170b5 | 350 | 1 | uio_books_raw_v1 |
| 3 | Fundamentals of vehicle dynamics Gillespie T. D. Thomas D. | 69a0a860-e886-7570-4743-ce77fc1e1157 | 139 | 1 | uio_books_raw_v1 |
| 4 | Fundamentals of vehicle dynamics Gillespie T. D. Thomas D. | 88cbdbfe-237b-b2ed-0f8b-09605b9839e3 | 139 | 1 | uio_books_raw_v1 |
| 5 | Tires Suspension and Handling Second Edition Dixon John C | 82d4669e-f578-32f3-3888-66bbe87b7d10 | 438 | 1 | uio_books_raw_v1 |
| 6 | Tires Suspension and Handling Second Edition Dixon John C | 10e2b0ca-761f-b540-5283-9fb891922240 | 437 | 1 | uio_books_raw_v1 |
| 7 | Fundamentals of vehicle dynamics Gillespie T. D. Thomas D. | 597498b8-5555-c72c-304f-ab90b75e7582 | 186 | 1 | uio_books_raw_v1 |
| 8 | Fundamentals of vehicle dynamics Gillespie T. D. Thomas D. | d6ed3070-411f-ea1b-763f-f8e83c2046fb | 2 | 1 | uio_books_raw_v1 |