Whegs (specifically Whegs II) is modeled on a cockroach, with legs, antennae, and a segmented body.
Besides the segmented body that can lift up to climb over obstacles, Whegs uses an adaptive tripod gait that lets it run along at a brisk 5.5 km/hr, but if it hits an obstacle, the front legs sync up to help it climb over. Whegs can be teleoperated, but also has autonomous capability. The AI is quite simple and relies on basic insect-like behaviors and sensors"
After analyzing how a cockroach positions its antennae when deciding to climb over or tunnel under a shelf-like obstacle, mechanical antennae were designed and fitted to Whegs™ II. The mechanical antennae swept up and down, each at a different rate, as the robot moved through its environment. When a shelf was encountered, the position of the antennae of the obstacle helped Whegs™ II autonomously decide whether to climb or tunnel. Like the cockroach, when both mechanical antennae touched the bottom of the shelf, the robot tunneled. And when both touched the top of the shelf, a climbing behavior was initiated. When one antenna touched the top while the other touched the bottom, Whegs™ II tunneled, just like the insect in a bright environment.
The only problem with the antennae is the range, so Whegs II was upgraded with some ultrasonic sensors that mimic eyes by mimicking ears. By using a single ultrasonic emitter and a pair of receivers, Whegs can determine the direction and distance to nearby objects in the same way that your brain interpolates where a sound is coming from by comparing when each ear hears it. Interestingly, these simple behaviors were used to augment the control of a teleoperator, automatically adjusting control inputs to avoid potential collisions. It’s simple hardware and simple behaviors, and it seems like it could make robot teleoperation significantly easier, if or when it makes it into the commercial (or military) sector.