Extended emission is a mystery in short gamma-ray bursts (SGRBs). By making time resolved spectral analyses of brightest nine events observed by ${\it Swift}$ XRT, we obviously classify the early X-ray emission of SGRBs into two types. One is the extended emission with exponentially rapid decay, which shows significant spectral softening during hundreds seconds since the SGRB trigger and is also detected by ${\it Swift}$-BAT. The other is a dim afterglow only showing power-law decay over $10^4$ s. The correlations between the temporal decay and spectral indices of the extended emissions are inconsistent with the $\alpha$-$\beta$ correlation expected for the high-latitude curvature emission from a uniform jet. The observed too-rapid decay suggests the emission from a photosphere or a patchy surface, and manifests the stopping central engine via such as magnetic reconnection at the black hole.

We are now investigating and studying a small satellite mission HiZ-GUNDAM for future observation of gamma-ray bursts (GRBs). The mission concept is to probe "the end of dark ages and the dawn of formation of astronomical objects", i.e. the physical condition of early universe beyond the redshift z > 7. We will consider two kinds of mission payloads, (1) wide field X-ray imaging detectors for GRB discovery, and (2) a near infrared telescope with 30 cm in diameter to select the high-z GRB candidates effectively. In this paper, we explain some requirements to promote the GRB cosmology based on the past observations, and also introduce the mission concept of HiZ-GUNDAM and basic development of X-ray imaging detectors.

Using 72 Short Gamma Ray Bursts (SGRBs) with well determined spectral data observed by BATSE, we determine their redshift and the luminosity by applying $E_p$--$L_p$ correlation for SGRBs found by \cite{tsutsui13}. For 53 SGRBs with the observed flux brighter than $4 \times 10^{-6}~{\rm erg~cm^{-2}s^{-1}}$, the cumulative redshift distribution up to $z=1$ agrees well with that of 22 {\it Swift}~SGRBs. This suggests that the redshift determination by the $E_p$--$L_p$ correlation for SGRBs works well. The minimum event rate at $z=0$ is estimated as $\rho_{SGRB}(0) = 6.3_{-3.9}^{+3.1} \times 10^{-10}~{\rm events~Mpc^{-3}yr^{-1}}$ so that the minimum beaming angle is $0.6^\circ-7.8^\circ$ assuming the merging rate of $10^{-7}-4\times 10^{-6}~{\rm events~Mpc^{-3}yr^{-1}}$ suggested from the binary pulsar data. Interestingly, this angle is consistent with that for SGRB130603B of $\sim 4^\circ-8^\circ$\citep{fong13b}. On the other hand, if we assume the beaming angle of $\sim 6^\circ$ suggested from four SGRBs with the observed value of beaming angle, the minimum event rate including off-axis SGRBs is estimated as $\rho_{SGRB,all}^{min}(0)=1.15_{-0.71}^{+0.57}\times 10^{-7}~{\rm events~Mpc^{-3}yr^{-1}}$. If SGRBs are induced by coalescence of binary neutron stars (NSs) and/or black holes (BHs), this event rate leads to the minimum gravitational-wave detection rate of $\rm 3.9_{-2.4}^{+1.9} (152_{-94}^{+75})~events~y^{-1}$ for NS-NS (NS-BH) binary, respectively, by a worldwide network with KAGRA, advanced-LIGO, advanced-Virgo, and GEO.