Krishna Chintalapudi Anand Padmanabha Iyer Venkata N. Padmanabhan ——presented by Xu Jia-xing Motivation Main idea of EZ Optimization Experiment Conclusion Motivation Main idea of EZ Optimization Experiment Conclusion Schemes that require specialized infrastructure. requires infrastructure deployment Schemes that build RF signal maps. takes too much time Model-Based Techniques. much less efforts than RF map; but still need a lot of work to fit the models Localization in Indoor Robotics. requires special sensors and maps Ad-Hoc localization. requires enough node density to enable multihopping Can we do indoor localization without such pre-deployments or limitations? Works with existing WiFi infrastructure only Does not require knowledge of Aps(placement, power,etc) Even work with measurements by a single device Does not require any explicit user participation There are enough WiFi APs to provide excellent coverage throughout the indoor environment Users carry mobile devices, such as smartphones and netbooks, equipped with WiFi Occasionally a mobile device obtains an absolute location ﬁx, say by obtaining a GPS lock at the edges of the indoor environment, such as at the entrance or near a window. Motivation Main idea of EZ Optimization Experiment Conclusion xj: the jth location ci: the ith AP’s location Pi: the power of the ith AP pij: the RSS received by mobile in the jth location form the ith AP ri: the rate of fall of RSS in the vicinity of the ith AP Motivation Main idea of EZ Optimization Experiment Conclusion Manner 10% of the solutions with the highest ﬁtness are retained. 10% of the solutions are randomly generated. 60% of the solutions are generated by crossover. The remaining 20% solutions are generated by randomly picking a solution from the previous generation and perturbing it(Only Pi and ri) Randomly pick Pi and ri with boundaries Use the LDPL equation : if there are m APs and n locations then reduce from 4m+2n to 4m R1 : If an AP can be seen from ﬁve or more ﬁxed (or determined)locations, then all four of its parameters can be uniquely solved. R2 : If an AP can be seen from four ﬁxed locations, there exist only two possible solutions for the four parameters of the AP. R3 : If an AP can be seen from three ﬁxed locations, randomly pick ri, there exist only two possible solutions for the three parameters of the AP. R4 : If an AP can be seen from two ﬁxed locations, randomly pick Pi and ri, there exist only two possible solutions for the two parameters of the AP. R5 : If an AP can be seen from one ﬁxed location, randomly pick all parameters. R6 : If the parameters for three (or more) APs have been ﬁxed, then all unknown locations that see all these APs can be exactly determined using trilateration. Calculate all equations fit R1 Randomly generate parameters of all equations fit R2 to R5 Calculate parameters of all unknown locations There are gain differences among different device. Introduce an additional unkown parameter G Calculate △Gk1k2 is possible: ◦ represent all RSS from a device with a vector If “Close” Wide coverage Low standard 2.Letdeviation in RSS 1.Normalize pij into range(0,1) Select each AP to provide information that other selected AP do not 3.Cluster APs one by one by 入 High average signal 4.Select the AP which can be seen by most known locations. strength Common Methods APSelect algorithm Motivation Main idea of EZ Optimization Experiment Conclusion Normal accuracy. More training data greater accuracy. Great performance. Different devices are better. The same as one device. Great improvement. Motivation Main idea of EZ Optimization Experiment Conclusion The idea is good. It’s different from traditional methods. The optimization is functional. The LDPL Model is not perfect. Does not mention how to refresh the RSS Model.