1. Inr. 1975. All righrrrcrcrved.
2. 6 spectively, where tg is time to go. There is no%cceleration feedback gain with the classical law. Three-axis cruciform steering is accomplished by computing cublc curve fitacceleration commands in both pitch and yaw. The vehicle is maintained upright by a separate roll rJnt1-01.
3. The nominal i?itial condition vector cqmponents were Yo= GOO it, Yo=282 ft/sec, Zo f 0, Zo= 282 ft/ sec, = 1630ft/sec 2, o0= -1~1'. 0, = 0. The velocity magnitude is 400 feet/second and the velocity vector is inclined to the position rector by an angle of 8 = 45'. Except for the angle e, initial position and velocity values are the same as the pitch plane case. The initial left and bank angle values for the optimal solution were chosen to be the same as the values determined by the classical cubic curve fit guidance law. The two sets of control lags modelled were i) TL = 0.1 sec, T = 0.1 sec
4. Using This approach was not used because the bounds of linearity were much smaller than the bounds of interest. Instead, two norms were defined ns measures of terminal error and these norms were computed for each of the initial condition perturbations. Classical and optimal comparativeterminal errors for autopilot time lag conditions, (i)and (ii), appear in Table 2. The results of Table 2 show the position and velocity norm [I II PV defined by I I * I = dyf+sec ;'+z