"... In PAGE 45: ...It is clear from the graphs that in this problem RTDP learned faster (and with less vari- ance) than AdaptiveRTDP and Real-Time Q-Learning, when learning rate is measured in terms of the number of epochs (numbers of moves are given in Table2 discussed below). This is not surprising given the di erences between the versions of the problem with com- plete information (Panel A) and with incomplete information (Panels B and C).... In PAGE 45: ... Although we do not illustrate it here, as the amount of uncertainty in the problem increases (increasing p), optimal policies generate paths that are more \conservative quot; in the sense of keeping safer distances from the track boundary and maintaining lower velocities. Table2 provides additional information about the performance of the real-time algorithms on the small track. For comparative purposes, the table includes a column for Gauss-Seidel DP.... In PAGE 45: ... We also turned o the random exploration used by the latter two algorithms. The rowof Table2 labeled \Est. path length at... ..."

"... In PAGE 7: ... As shown in Sections 2 and 3, interrupt distribution routines in a hybrid system have huge influence for interrupt latency. In Table1 , we compare the main functions and code sizes of the real-time interrupt distribution routines in RTLinux-THIN and RTAI based on this platform. in Table 1.... In PAGE 7: ... in Table 1. From Table1 , we can see that in order to lock the real- time interrupt routines into the cache on Intel PXA270... ..."

"... In PAGE 10: ...In addition to the information provided by the kiosk system, the in-home and personal information systems would provide users with paratransit and real-time paratransit information, and could be accessed through a home computer/TV or from a portable hand-held device. Table2 shows that 73 percent of the disabled would make more trips by transit, paratransit or real-time paratransit if an in-home information system were available. And 75 percent would make more trips by transit, paratransit, or real-time paratransit if a personal information system were available.... ..."

"... In PAGE 2: ... The sensor mounts are flexible, so that new sensors can easily be added. Sensors include a General Dynamics Robotics Systems (GDRS)* imaging ladar (Table 1) mounted on a tilt platform, a color camera mounted on top of the ladar on the tilt platform, and a highly accurate positioning system ( Table2 and Table 3)3. Other sensors that are commonly used on the vehicle include a pair of color stereo cameras, a Sick line-scan ladar, currently mounted on the back of the vehicle, and a Riegl high-resolution scanning ladar (Table 4).... In PAGE 3: ... This unit typically provides real-time data accurate to better than one meter and a few hundredths of a degree. With post-processing, the accuracy is a few centimeters in distance and angular accuracy is a few thousandths of a degree ( Table2 and Table 3). To date, data have been collected for two main purposes.... ..."

"... In PAGE 17: ...he standby unit cannot fail while powered-o , i.e., s = 0. 5 CONCLUSIONS Table6 shows a comparison of the introduc- tion of failure dependencies in the real-time com- puter model. Entries in this table refer to the ex- amples previously discussed, the respective size (in terms of number of states and transitions) of the CTMC underlying the stochastic Petri net model, and an observation stating if the three subsystems in the model remained statistically independent (s- independent) or not.... ..."

"... In PAGE 26: ... The accuracy of these two techniques is only slightly worse than regression, but has the advantages of lower complexity and being able to dynamically increase or decrease the range size in order to better respond to any system require- ment such as more power reduction or better video quality [13]. Table8 summarizes the different inter-task DVS techniques for MPEG decoding dis- cussed in the previous paragraphs. ... ..."

"... In PAGE 4: ... First, we consider audio only, we simulate and measure the loss percentage for a call from node 0 to the seven other nodes. Table1 shows the loss percentage for a call from node 0 to nodes 1.... ..."

"... In PAGE 4: ... First, we consider audio only, we simulate and measure the loss percentage for a call from node 0 to the seven other nodes. Table1 shows the loss percentage for a call from node 0 to nodes 1.... ..."

"... In PAGE 11: ... 4.4 Example Table1 lists the tasks of a real-time task set. The unstarred tasks are the hard real-time tasks.... ..."