I've changed what was called a random walker agent to be called a random action agent. It samples actions and saccades from various distributions, with biases and correlations.
I've added a more interpretable random walker. This walker goes either forwards or backwards, taking a number of steps sampled from a specified distribution. Then it starts turning left or right, turning for the number of steps sampled from a specified distribution. Turn steps can be correlated with previous numbers of turn steps to give correlated random walkers. Forwards/backwards steps are interleaved with no-action steps to reduce the effect of momentum on agent motion. This is a more strict form of a random walker, and serves as another baseline for comparisons.
I've changed what was called a random walker agent to be called a random action agent. It samples actions and saccades from various distributions, with biases and correlations.
I've added a more interpretable random walker. This walker goes either forwards or backwards, taking a number of steps sampled from a specified distribution. Then it starts turning left or right, turning for the number of steps sampled from a specified distribution. Turn steps can be correlated with previous numbers of turn steps to give correlated random walkers. Forwards/backwards steps are interleaved with no-action steps to reduce the effect of momentum on agent motion. This is a more strict form of a random walker, and serves as another baseline for comparisons.