When iteration turns 5,program fails

Dear all,

My problem is that when iteration turns 5,program fails and the following is the program and the result:

cvx_begin
    variables u(N) xq(N) yq(N)
    expression y(N)
    for i=1:N
        y(i) = -B/log(2)/((B+(u0(i)))*u0(i))*u(i);
    end
    maximize (sum(y)/N)
    subject to
        xq(1) == xqs;
        yq(1) == yqs;
        (xq(N)-xqf)^2+(yq(N)-yqf)^2 <= D^2;
        for i =1:N-1
            (xq(i)-xq(i+1))^2+(yq(i)-yq(i+1))^2 <= D^2;
        end
        for i =1:N
            dSR2(i) = (xq(i)-xs)^2+(yq(i)-ys)^2+zq^2;
            dRD2(i) = (xq(i)-xd)^2+(yq(i)-yd)^2+zq^2;
            pow_p(u(i),(2/K)) >= 1/2*(pow_pos(dSR2(i)+dRD2(i),2)-(pow_p(dSR02(i),2)+pow_p(dRD02(i),2)))+(-2*dSR02(i)*((xqt(i)-xs)*(xq(i)-xqt(i))+(yqt(i)-ys)*(yq(i)-yqt(i)))-2*dRD02(i)*((xqt(i)-xd)*(xq(i)-xqt(i))+(yqt(i)-yd)*(yq(i)-yqt(i))));
        end  
cvx_end

Calling SDPT3 4.0: 7400 variables, 3198 equality constraints
For improved efficiency, SDPT3 is solving the dual problem.

num. of constraints = 3198
dim. of sdp var = 4400, num. of sdp blk = 2200
dim. of linear var = 800


SDPT3: Infeasible path-following algorithms


version predcorr gam expon scale_data
HKM 1 0.000 1 0
it pstep dstep pinfeas dinfeas gap prim-obj dual-obj cputime

0|0.000|0.000|5.4e+10|2.2e+00|2.9e+14|-1.395985e+14 0.000000e+00| 0:0:00| spchol 1 1
1|0.000|0.000|5.4e+10|2.2e+00|2.9e+14|-1.395985e+14 0.000000e+00| 0:0:00| spchol 1 2
2|0.000|0.000|5.4e+10|2.2e+00|2.9e+14|-1.395985e+14 0.000000e+00| 0:0:00| spchol 1 2
3|0.000|0.000|5.4e+10|2.2e+00|2.9e+14|-1.395985e+14 0.000000e+00| 0:0:00| spchol 1 1
4|0.000|0.000|5.4e+10|2.2e+00|2.9e+14|-1.395985e+14 0.000000e+00| 0:0:01|
*** Too many tiny steps: restarting with the following iterate.
*** [X,y,Z] = infeaspt(blk,At,C,b,2,1e5); spchol 10 6
5|0.000|0.000|5.4e+14|1.0e+00|5.2e+13|-1.372460e+18 0.000000e+00| 0:0:01|
*** Too many tiny steps even after restarting
stop: steps too short consecutively

prim_inf,dual_inf,relgap = 5.39e+14, 1.00e+00, 3.79e-05
sqlp stop: dual problem is suspected of being infeasible

number of iterations = 5
residual of dual infeasibility
certificate X = 3.86e+00
reldist to infeas. <= 9.27e+00
Total CPU time (secs) = 0.63
CPU time per iteration = 0.13
termination code = 2
DIMACS: 5.4e+10 0.0e+00 1.0e+01 0.0e+00 -1.0e+00 2.1e+00


Status: Infeasible
Optimal value (cvx_optval): +Inf

Crude Successive Convex Approximation is unreliable. Read my posts in this forum.