Start with imagining a typical humanoid robot. At the hips there would be 2 legs descending. However, in this design, just below the hip-joint, for each leg, there is another joint, where the leg can split to become two legs, one behind the other. Thus 4 separate legs can reach the ground (think of the 4 upper edges of a pyramid), giving the robot good stability.
The robot can walk in a mostly ordinary quadruped manner. There will be some differences because the legs all connect in a single region (not hugely unlike an insect, except the insect has 6 legs instead of 4), whereas the typical quadruped has well-separated pairs of legs.
However, whenever needed, and because of the way the legs are connected, a two-legged gait is also possible (sometime after the robot's balancing software gets perfected --that is considered to be a rather tough problem in robotics). Consider the left side of the robot, where first a short leg-length descends from the hip, as originally described, after which a special joint connects two more-ordinary legs (each with whatever other joints are desired for overall flexibility).
Imagine the special joint to be pliers-like, such that the two legs could nestle like the jaws of long-nose pliers (there is no equivalent of pliers-handle here; there is only joint and jaws- equivalent). When nestled, the two legs are equivalent to a single leg. On the right side of the robot we have the same sort of leg- arrangement, and thus when the two pairs of legs are nestled, the robot can theoretically move in a two-legged gait, and more- closely resemble a human. This could be good if you needed the robot to sit in a car seat and drive --having 4 splayed legs simply wouldn't work.
That's all.-- Vernon, Dec 23 2015 Inspiration for this Idea http://www.popsci.c...-driving-robots-drcApparently robots need to be humanoid to do the wide variety of things that humans can do --but walking on two legs is tough. So, if two legs can become 4.... [Vernon, Dec 23 2015] I like the idea for upper body work stability like crane pads but I'm unsure the complexity needed would be a valid gain for the quadruped motion. The stability is in the separation at the hip even in insects.-- wjt, Dec 23 2015 //needs better proprioception cybernetics//
[Marked-for-tagline]
Could be said of most HB ideas...-- csea, Dec 23 2015 I do need to mention a "fix" for a possible problem. It turns out that if the nested pair of legs are simply adjacent to each other, the knee action may not work correctly. One solution involves these two shapes: o U Each is a cross-section of one leg; when they nestle, the "o" cross-section fits inside the "U" cross-section. Now the knee-joints can align on the same axis of rotation. (Need it for the ankle joints, too.)-- Vernon, Dec 24 2015 random, halfbakery