This optional companion to Appendix A explains the other main setup controls and develops two computed settings. The runnable checkpoint exposes fuel and rear-wing angle; the extensions below require deliberate C++ changes as well as XML edits.
In the source tree, the reference model is data/cars/models/car1-stock1/car1-stock1.xml. The installed data uses cars/car1-stock1/car1-stock1.xml, alongside categories/stock1.xml. The model and category establish which properties may vary. For a running robot, the incoming carSettings handle in initTrack is the useful merged reference.
An XML number can carry min, max, unit, and val. A string may carry an in list of permitted choices. These declarations describe values and constraints; they do not grant the driver permission to exceed the incoming reference limits. An entry with no useful adjustable interval may be fixed by the car model.
To add a numeric car setting to Tutorial A, add a readCarNumber call in readLayer, using the correct section and key macros from car.h. Rebuild, then test a sparse XML override. For example, brake pressure uses SECT_BRKSYST and PRM_BRKPRESS. That helper checks the reference bounds in internal units and copies only the accepted number into the returned physical handle.
String settings need a separate path: read the proposed string, check its allowed choices, set it on the output with GfParmSetStr, and retain the final GfParmCheckHandle check. Copy or use borrowed strings while the input handle is alive. Avoid passing the whole private configuration file through as a car setup.
Fuel adds mass and determines the available driving distance. A short qualifying run and a long race need different allowances for warm-up, measurement error, and refuelling opportunities. The lightest starting car is not useful if it cannot finish its planned stint.
Wing angle is measured relative to the car. Over the modest range used in Appendix A, more angle adds aerodynamic loading and drag in the reference physics model. Front and rear wing changes also alter aerodynamic balance. A rear-wing change can therefore change both straight-line performance and the way a corner is negotiated. See aerodynamic braking for the speed-dependent force model. Keep its approximations distinct from the full simulator.
Gear entries are nested, for example Gearbox/gears/1 and Gearbox/gears/r, with the attribute ratio. Reverse uses a negative ratio. Ignoring slip and losses,
\[v=\frac{\omega_e r_w}{|i_g i_f|},\qquad |T_w|\approx \eta |i_g i_f T_e|. \]
Here \(\omega_e\) is engine angular speed in rad/s, \(r_w\) rolling radius in metres, \(i_g\) gear ratio, \(i_f\) final-drive ratio, \(T_e\) engine torque, and \(\eta\) transmission efficiency. A larger forward ratio increases ideal wheel torque but reduces road speed at the same engine speed. A shorter first gear can improve acceleration until traction or rapid shifting becomes the limit. A taller top gear is only helpful if the engine can pull it against resistance. Recheck the driver's shift behavior after changing ratios.
Rear Differential/type selects a model such as FREE, LIMITED SLIP, or SPOOL when permitted by the car. A free differential allows the driven wheels to turn at different speeds, but a low-grip wheel can limit useful traction. A limited-slip model couples the wheels more strongly under its configured conditions. A spool constrains them to turn together, affecting cornering because the two wheel paths have different lengths. None is universally best for every surface and controller.
Compare exit acceleration and off-road recovery as well as a smooth solo lap. Understeer during acceleration is a clue to investigate; it does not by itself prove that the differential is the sole cause.
Wheel sections such as Front Right Wheel contain ride height, toe, and camber. Read both left and right definitions before editing them.
Inspect road contact, stability, and edge recovery. Do not substitute a real-car alignment convention or a remembered sign rule for the definitions in the selected simulator and car. Angles requested from the API with no output unit are radians; XML commonly uses degrees.
Suspension sections contain spring, suspension course, bellcrank, packers, and the bump/rebound damping parameters. Anti-roll-bar sections couple the left and right sides of an axle and resist their differential movement.
Change one family at a time, testing both a smooth corner and a bumpy section. Keep enough travel for the actual track. A faster best lap accompanied by repeated loss of road contact is a poor basis for a reliable race setup.
Instead of writing a fixed starting amount, you can use a configured consumption estimate and the planned driving distance. For a lap-based exercise:
\[\begin{aligned} F_{\rm wanted}&=(N+N_{\rm allowance})\widehat f+F_{\rm reserve},\\ F_{\rm start}&=\operatorname{clamp} (F_{\rm wanted},F_{\min},\min(F_{\max},F_{\rm tank})). \end{aligned} \]
\(N\) is the number of laps planned before refuelling or finishing; \(N_{\rm allowance}\) accounts for an extra distance allowance; \(\widehat f\) is litres per lap; \(F_{\rm reserve}\) is litres. Obtain \(F_{\min}\), \(F_{\max}\), and tank capacity from the current car reference. Check that the allowable interval is valid before clamping.
For six planned laps, one additional lap, 1.5 l/lap, and a 0.5 l reserve, the estimate is 11 l. If the result exceeds tank capacity, clamping does not make the race feasible without a stop: it identifies the need for a stint plan. Timed races need an estimated distance or duration model rather than blindly using a lap count.
To implement this extension, compute the amount during initTrack, then set SECT_CAR/PRM_FUEL on the output handle with GfParmSetNum(..., "l", ...). Decide and document whether this computed policy overrides an explicit XML fuel value; do not leave two competing assignments with accidental precedence. Keep the existing no-side-pits policy in mind.
The supplied checkpoint deliberately uses a direct fuel override for the first exercise. Its fuel per lap private setting feeds the pit strategy's fallback; it does not automatically compute starting fuel. Once driving begins, Chapter 6's measured consumption is the better basis for updating the plan.
This develops the setup idea introduced in Wheel Control. It is an optional analytical experiment, not an automatic feature of Tutorial A. The equations below describe current simuv2's ordinary service-brake path.
Let \(u\in[0,1]\) be the brake command, \(P_{\max}\) the max pressure setting in pascals, and \(q\) the effective front pressure fraction. With no extra brake-balance command, \(q\) is the configured front-rear brake repartition. Each front wheel receives \(P_f=uP_{\max}q\); each rear wheel receives \(P_r=uP_{\max}(1-q)\).
For wheel \(i\), simuv2 constructs the coefficient and brake torque as
\[C_i=\frac{d_i}{2}A_i\mu_{b,i},\qquad T_{b,i}=C_i P_i. \]
\(d_i\) is disk diameter in metres, \(A_i\) the modeled piston area in square metres, and \(\mu_{b,i}\) brake friction. Use the simulator coefficient as written; do not add an assumed extra pair of pads or caliper multiplier. \(C_i\) has units of cubic metres, and \(C_iP_i\) has units N·m.
Ignoring wheel rotational dynamics, a longitudinal contact force \(F_{x,i}\) requires approximately \(T_i=r_iF_{x,i}\), with rolling radius \(r_i\). For a simple straight-line grip estimate \(|F_{x,i}|\leq\mu_{t,i}F_{z,i}\),
\[P_{\max}\lesssim \min_i\frac{\mu_{t,i}F_{z,i}r_i}{u C_i q_i}, \qquad q_i=\begin{cases}q&\text{front},\\1-q&\text{rear}.\end{cases} \]
Here \(\mu_{t,i}\) is an effective tyre/surface grip estimate, distinct from brake friction; \(F_{z,i}\) is normal load in newtons. Require positive radii, coefficients, loads, and \(u q_i\) before dividing. This bound estimates the first wheel likely to reach its modeled force capacity at the chosen command. It is not a complete locked-wheel criterion.
For example, take \(F_z=4000\ \mathrm N\), \(\mu_t=1\), \(r=0.33\ \mathrm m\), \(d=0.38\ \mathrm m\), \(A=0.002\ \mathrm{m^2}\), \(\mu_b=0.3\), \(q=0.66\), and \(u=1\). Then \(C=0.000114\ \mathrm{m^3}\), and this front wheel gives approximately \(P_{\max}=17.54\ \mathrm{MPa}=17544\ \mathrm{kPa}\). These are illustrative inputs, not a measured optimum for the reference car.
For a level, straight-line approximation with mass \(m\), wheelbase \(\ell\), centre-of-mass height \(h\), static front load fraction \(\lambda_f\), and positive braking deceleration \(a\):
\[\begin{aligned} \Delta F_z&=\frac{mah}{\ell},\\ F_{z,f}&=\lambda_f mg+\Delta F_z+D_f,\\ F_{z,r}&=(1-\lambda_f)mg-\Delta F_z+D_r. \end{aligned} \]
\(D_f,D_r\) are simplified front/rear aerodynamic loads. Divide axle loads between the two wheels only when the situation is sufficiently symmetric. Actual loads change with banking, bumps, body motion, aero balance, and cornering. Cornering also uses some of the tyre's available force; a friction-circle approximation would replace the full longitudinal allowance by \(\sqrt{(\mu_t F_z)^2-F_y^2}\) where the radicand is nonnegative.
The command matters as much as the pressure setting. The normal speed controller caps its brake request below 1, while other states can command stronger braking. Do not inflate pressure to compensate for one controller cap without checking recovery, pit stopping, and the full-brake case. ABS modulates the command; it does not change the configured hardware coefficient or make every pressure equally useful.
Reducing pressure can reduce excessive slip but also lengthen stopping distance. Moving the balance rearward can help one operating point and destabilize another. The estimate is a starting hypothesis to test, not a universal automatic setup.
A possible extension is defaults → track → session-specific track. Select a session name from situation->_raceType using RM_TYPE_PRACTICE, RM_TYPE_QUALIF, and RM_TYPE_RACE, then load a third layer such as drivers/tutoriala/race/g-track-1.xml. Define an explicit fallback for other session types. Add every new directory/file to the Linux and Windows installation rules. A directory name has no effect until your loader uses it.
Qualifying may favour short-run pace with a smaller fuel allowance. A race setup must also finish its stints, handle traffic and recovery, and avoid accumulating damage. Driver skill level and the selected physics module affect comparisons; record them and keep them fixed. Multiple driver indices and independent state are covered in Appendix C.
For systematic tuning, start from a baseline, try a smaller and a larger value, and repeat the promising direction. Recheck nearby values before moving to another parameter. This is a local search, not proof of a global optimum: settings interact, and changing the controller can invalidate a previously good setup. Record flying laps, road-edge behavior, recovery, fuel use, and damage. Finish with a full-race test and restore the baseline if the supposed gain does not repeat.
Return to Appendix A.
1.14.0 