Page - 118 - in Maximum Tire-Road Friction Coefficient Estimation

6 Results and conclusion be detected in an acceptable and reliable manner. The inferior performance during this manoeuvre that includes a low dynamic excitation on a road with highµmax road is also visible in the error values in Table 6.2 when one compares its error values to those for themanoeuvrewith thehigherdynamic excitation. Nevertheless, theMAEis still below 0.35 for all methods, which is remarkable considering the total failure to detect the dry surface. Looking more closely at the estimate using variable particles re-initialised by SD (light gray) in Figure 6.10, the SD on the dry road (before second 9) appears higher than on the correctly detected low-friction surface. Thus, the SD might be a helpful measure to estimate the current estimation accuracy. 7.5 8 8.5 9 9.5 10 10.5 11 11.5 12 12.5 13 0 0.2 0.4 0.6 0.8 1 Reference µmax(fixed particles), no resampling µmax(sampled particles), SD limited µmax(sampled particles), ∆µmax limitedˆˆ ˆ ˆ Time in s Figure 6.10.: Estimatesofµmax usingfixedparticles (blackdotted), variableparticles re- initialised based on SD (light gray) or variable particles re-initialised based on ∆µˆmax (dark gray) versus time during a braking manoeuvre with a maximumdecelerationof≈−2m/s2 foraµ step manoeuvrewithµmax≈1 before andµmax≈0.3 after second 9. MAE and maximum absolute errors are shown in Table 6.2. Figure 6.12 clearly shows that the SD for this manoeuvre is generally lower after the stepofµmax occursat second9. Knowledgeabout theactualaccuracy isofgreat interest when considering the application of an estimatedµmax for adaptation of the activation strategyofADAS.Kobialka et al. showedthat theestimationof theestimationaccuracy is possible using ANN, [KL13]. Nevertheless, further investigations are necessary to support the hypothesis that the SD of the variable particles can be used for accuracy estimation. 118