max |recursive belief - enumerated posterior| over 200 histories: 5.00e-16
belief after 4 steps, from stored belief : [0.481301 0.147606 0.371093]
belief after 4 steps, from full history  : [0.481301 0.147606 0.371093]
sum_o Pr(o | b, a) = 1 for every action: verified
impossible observation -> observation has probability zero under the belief
