Curvilinear motion is the movement of a point or body along a curved path rather than a straight line. Unlike rectilinear motion which travels along a single axis, curvilinear motion always carries two acceleration components — tangential along the path and normal toward the centre of curvature. We use it to describe and design any mechanism where a follower, tooltip, or payload sweeps an arc or compound curve, from cam-driven indexers to robotic end-effectors. Get the radius of curvature wrong and you size the centripetal load wrong — and the bearing or pivot pays the price.
The Curvilinear Motion in Action
Any time a point traces a path that isn't a straight line, you've got curvilinear motion. The position vector changes both in magnitude and direction as the body moves, which is why a single scalar speed never tells the whole story. You need to track the path itself — the radius of curvature ρ at every instant — because that radius sets the normal acceleration, and the normal acceleration sets the force the structure has to react.
The two acceleration components do different jobs. Tangential acceleration at changes how fast the body moves along the path. Normal acceleration an changes the direction of travel and points toward the centre of curvature. At constant speed at is zero but an is not — that's why a roller-coaster car at steady velocity still pushes hard into the rails through a loop. Get the radius wrong by 20% on a high-speed cam follower and the bearing load can spike 25%, because an = v2 / ρ scales inversely with the radius.
Designers get burned when they treat curvilinear motion like rectilinear motion with a curve drawn on top. A cam-follower roller riding a profile with a sudden curvature change — say from ρ = 50 mm to ρ = 15 mm in 10 degrees of cam rotation — sees the contact force jump nearly 3× across that transition. Smooth the curvature transition (this is why we use polynomial cam profiles instead of straight-arc-and-line) and the follower stops bouncing. Common failure modes are pin shear at sharp curvature inflections, follower-roller skid when tangential acceleration exceeds available friction, and fatigue cracking on pivot pins fed by repeated normal-load cycling.
Key Components
- Path (Trajectory): The geometric curve the body traces in space. We define it parametrically as r(s) where s is arc length. The path determines ρ at every point — and ρ values below 5 mm at speeds above 1 m/s start producing centripetal accelerations above 200 m/s2, which is where structural design gets serious.
- Tangent Vector (et): Unit vector pointing along the direction of motion at any instant. The tangential acceleration component lies along this vector. If the path is fairing-smooth (continuous third derivative) the tangent rotates gradually; if not, you get jerk spikes that excite resonance in the driven member.
- Normal Vector (en): Unit vector perpendicular to et, pointing toward the instantaneous centre of curvature. The normal acceleration acts along this vector. The pivot, bearing, or guide rail must react this load — sized properly, this is straightforward; sized to rectilinear assumptions, parts fail.
- Radius of Curvature (ρ): The instantaneous radius of the osculating circle that best fits the path at that point. ρ varies along most real paths. For a circular arc ρ is constant; for a cam profile or spline trajectory ρ changes continuously, and the design tolerance on ρ is typically ±2% to keep contact stress predictable.
- Speed (v): The scalar magnitude of the velocity vector along the tangent. Speed alone tells you nothing about force — you need v together with ρ to get the centripetal acceleration v2/ρ, which is what actually loads the structure.
Who Uses the Curvilinear Motion
Curvilinear motion shows up everywhere a part doesn't move in a straight line — which is most of mechanical engineering. Where it matters most is anywhere a high-speed follower, end-effector, or payload sweeps a curved path and the designer has to size pivots, bearings, and structures to react centripetal loading. You'll see it in cam mechanisms, robot wrists, automotive steering geometry, conveyor curves, and any aerial or marine system tracking a curved trajectory. The common thread is that the engineer must compute both acceleration components and not just the path-speed.
- Automotive Engineering: Front-suspension steering knuckle motion in a MacPherson strut on a Ford Focus traces a 3D curvilinear path as the wheel both steers and rises through bump travel.
- Industrial Automation: End-effector trajectory of a FANUC LR Mate 200iD/7L six-axis robot sweeping a curved weld seam at 250 mm/s along a complex bezier-spline path.
- Packaging Machinery: Cam-driven follower roller on a Bosch Pack 403 cartoner traces a non-circular profile to deliver dwell-rise-return motion to the cartoning fingers at 200 cycles per minute.
- Aerospace: Flight-path tracking of a Lockheed Martin F-35 in a coordinated 5g turn — the airframe's centre of mass follows a 3D curvilinear path with continuously varying ρ.
- Theme Park Engineering: Roller-coaster trains on Bolliger & Mabillard track designs use carefully clothoid-faired curvilinear paths to keep normal acceleration ramping smoothly between 1g and 4g without jerk spikes.
- Material Handling: Curved roller conveyors on FedEx parcel-sorting hubs guide packages around 90° turns where the outboard rollers run faster than the inboard rollers to maintain package orientation along the curvilinear path.
The Formula Behind the Curvilinear Motion
The two acceleration components — tangential and normal — are what every curvilinear-motion calculation reduces to. The tangential piece is straightforward: it's how the speed changes. The normal piece is the one that catches people out, because it exists even at constant speed and grows with the square of velocity. At the low end of typical operating ranges (say 0.1 m/s along a 100 mm radius arc) the normal acceleration is a trivial 0.1 m/s2. At the high end (5 m/s along the same radius) it explodes to 250 m/s2 — about 25g. The sweet spot for most mechanism design sits where v2/ρ stays under about 50 m/s2, because that's roughly where standard rolling-element bearings and ground-steel pivots last their rated L10 life without exotic specification.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| atotal | Magnitude of total acceleration of the point on the curved path | m/s2 | ft/s2 |
| at | Tangential acceleration — rate of change of speed along the path | m/s2 | ft/s2 |
| an | Normal (centripetal) acceleration directed toward the centre of curvature | m/s2 | ft/s2 |
| v | Instantaneous speed along the tangent to the path | m/s | ft/s |
| ρ | Instantaneous radius of curvature of the path at the point of interest | m | ft |
Worked Example: Curvilinear Motion in a textile loom shuttle-race curve
You are sizing the pivot bearings for the shuttle-guide roller on a Picanol OmniPlus-i air-jet loom retrofit, where the guide arm sweeps the leno-edge thread along a curved cam path. The roller centre traces an arc with radius of curvature ρ = 80 mm at the critical fastest point. The roller carrier mass is 0.45 kg, and the design speed at that point is 2.5 m/s, with tangential acceleration of 12 m/s2 as the cam ramps the carrier up. You need the total acceleration so you can size the pivot pin and the needle bearing for the centripetal load.
Given
- ρ = 80 mm
- vnom = 2.5 m/s
- at = 12 m/s2
- m = 0.45 kg
Solution
Step 1 — at nominal 2.5 m/s, compute the normal (centripetal) acceleration:
Step 2 — combine with tangential acceleration to get total acceleration magnitude at the nominal point:
Step 3 — convert to pivot-pin reaction force at nominal:
At the low end of the typical loom operating range — say a slow-pick threading speed of 1.2 m/s — the normal acceleration drops to an = 1.44 / 0.080 = 18 m/s2, and the pivot reaction is around 9.7 N. Easy duty, the pin sees almost nothing. At the high end, where the loom is pushed to 3.5 m/s in a faster fabric programme, an jumps to 153 m/s2 and total reaction climbs past 70 N — nearly double the nominal load, because the centripetal term scales with v2. That's where standard 4 mm needle bearings start fretting unless you upgrade to a hardened DIN 5402 spec and tighten the ρ tolerance to ±1.5%.
Result
Nominal total acceleration is 79. 0 m/s2 — about 8g — and the pivot pin sees 35.6 N at design speed. That's modest force on paper, but it cycles 600+ times per minute on a running loom, so fatigue dominates. The low-end 9.7 N case is benign; the high-end 70 N case at 3.5 m/s is where the bearing manufacturer's L10 life calculation actually constrains your service interval. If you measure pin-bore wear faster than predicted, the usual culprits are: (1) the cam profile's actual ρ is tighter than drawing-spec at the critical point because the cam was ground without proper curvature inspection, (2) the carrier mass crept up due to a heavier guide-roller substitution, lifting m × an proportionally, or (3) shuttle-thread tension spikes adding an unmodelled tangential load that pushes at well above 12 m/s2.
When to Use a Curvilinear Motion and When Not To
Curvilinear motion isn't really an alternative you pick — it's a class of motion that exists whenever the path bends. But when you're framing a mechanism design, you do choose between treating motion as rectilinear (straight-line), curvilinear (arc or compound curve), or rotational (pure rotation about a fixed axis). Each choice carries different analysis effort, different load-prediction accuracy, and different mechanism cost. Here's how they stack up on the dimensions a designer actually decides on.
| Property | Curvilinear Motion | Rectilinear Motion | Pure Rotational Motion |
|---|---|---|---|
| Analysis complexity | High — must track ρ, at, an simultaneously | Low — single-axis kinematics | Medium — single ω, fixed radius |
| Typical operating speed range | 0.1 to 10 m/s along path | 0.01 to 5 m/s linear | 10 to 10,000 RPM |
| Path-position accuracy achievable | ±0.05 mm with cam profiles, ±0.5 mm with splines | ±0.005 mm with ballscrews, ±0.02 mm with belt drives | ±0.001° with encoders, ±0.05° with steppers |
| Mechanism cost relative to function | Medium-high — cams, splines, multi-axis linkages | Low-medium — linear actuators and slides | Low — gearmotor or direct drive |
| Reliability / failure mode | Pivot fatigue and follower skid at curvature inflections | Side-load wear on rails, screw backlash | Bearing fatigue, brush wear (DC motors) |
| Best application fit | Cam followers, robot end-effectors, vehicle paths | Linear positioning, presses, slides | Spindles, conveyor drums, rotary indexers |
Frequently Asked Questions About Curvilinear Motion
That's almost always a curvature discontinuity — the cam profile inspection probably checked dimensional tolerance on the surface, not the second derivative of the profile. A 0.02 mm step in surface position can be invisible to a CMM but produce a sharp ρ change that spikes an by 50% over a few degrees of rotation. The follower briefly loses contact, then slams back down.
Fix it by inspecting curvature continuity, not just position. Plot ρ versus cam angle and look for any step or sharp inflection. Polynomial cam profiles (modified sine, cycloidal) are designed to be C2 continuous specifically to avoid this — if your cam was generated as straight-arc-and-line segments, that's the root cause.
Use the centripetal-to-tangential ratio. If an / at stays below about 0.1 across the operating range, the rectilinear approximation costs you less than 1% on total acceleration magnitude and you can safely ignore the curve. Above 0.1, you start getting force-prediction errors that show up as bearing or pivot wear faster than predicted.
Quick check: at your design speed, compute v2/ρ. If that's less than 10% of your tangential acceleration, treat as rectilinear. If it's comparable or larger, you must do the full curvilinear analysis.
The most common cause is path-flexibility you didn't model. If the supporting structure deflects under load, the actual ρ at speed is tighter than the static drawing ρ — the part is taking a shortcut through the curve. A 0.3 mm radial deflection on a 50 mm-radius arc tightens ρ effectively by ~5% and bumps an proportionally.
Second culprit: dynamic mass beyond what you accounted for. Cable carriers, sensor mounts, and lubricant slosh all add effective mass at the follower. Weigh the entire moving assembly, not just the design BOM.
Below about 6 mm ρ on a steel-on-steel cam at industrial speeds (1-3 m/s), Hertzian contact stress on a typical 10 mm roller follower exceeds 2 GPa and you start pitting the follower within months. That's the practical threshold to switch to a sliding flat-faced follower or to drop the speed.
Real-world example: high-speed printing-press cams running ρ = 4 mm sections universally use flat-faced followers with hardened tool-steel inserts, not rollers — Heidelberg made that switch in the 1970s after roller pitting issues on high-speed Speedmaster presses.
The controller is commanding tangential velocity correctly, but it's not accounting for the centripetal load on the arm joints. As ρ tightens, the joint torques required to hold the path climb with v2/ρ — and if the servo loop saturates on torque, the actual arm trails the commanded position outward. You see overshoot or path-following error proportional to 1/ρ.
Two fixes: either reduce v on tight-ρ segments (most modern path planners do this automatically — look for a 'feedrate scaling on curvature' option), or upgrade the joint motor torque margin. As a rule of thumb, keep v2/ρ below 30% of the joint's peak acceleration capability.
No — ρ → ∞ just means the path is locally straight at that point (an inflection point or a straight segment). an goes to zero there, which is fine. What's actually worth watching is the rate of change of curvature dκ/ds where κ = 1/ρ. Sharp transitions in κ produce jerk, even if ρ itself never goes near zero.
This is why high-speed motion paths use clothoid spirals or quintic polynomials between straight and curved segments — they ramp κ smoothly so the system never sees a step in normal acceleration. Roller coasters have used clothoid lead-ins since Anton Schwarzkopf's designs in the 1970s for exactly this reason.
References & Further Reading
- Wikipedia contributors. Curvilinear motion. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
