While a giant corpus of information exists on locomotion, our human manifestations of this knowledge, i. e. humanoid robots,
illustrate how little we really know. Virtually all walking machines to date are slow, inefficient, un-natural and lack the locomotory skills of a 5 year old human child.
Compounding the locomotion problem are research "silos" which should interact, but do not.
Through most of the history of A.I., and continuing today, legged locomotion was generally ignored as perhaps something easy to work out. Vision was regarded as completely decoupled from locomotion. These subjects are studied by two vastly different groups of people with different training.
We argue that vision, locomotion, and intelligence interact tightly in systems (human beings) that perform at high levels.
Vision and locomotion interact in complex ways. For example, we have shown that movement can enable visual perception of obstacles in the environment. However, the physical walking movements themselves can lead to complex motion patterns on the retina that require a large amount of computation (intelligence) to sort out.
We strive to bring together vision, locomotion, and intelligence into an integrated whole.