Objective To propose a statistical inference method for linear additive spatial autoregressive models under constraint conditions to address spatial dependency issues.Method First， linear constraints are added to the linear additive spatial autoregressive model. By approximating the unknown smoothing function using B-spline basis functions and substituting it into the regression model， a spatial autoregressive model containing B-spline basis functions is generated. The loss function of this model is minimized to obtain a sparse estimate of the model under constraint conditions. The SCAD penalty function is used to estimate the larger coefficients without bias. The asymptotic properties of linear additive space autoregressive model estimation were studied under 10 constraint conditions， and it was found that when estimating the slope coefficient at the optimal convergence speed， the asymptotic distribution characteristics of the parameter estimation were not affected by the function term.Result The experimental results show that this method is not only feasible in theory， but also easy to operate in practice. Compared to unconstrained estimation， constrained sparse estimation is superior.Conclusion This method is practical and effective in traffic data statistical analysis.