Accidents will occur when ASIMO leaves the lab and enters the busy, unpredictable world we live in. That much is inevitable. How ASIMO reacts to unexpected bumps will play a major role in minimizing the potential damage, both to the robot and everything around it. That’s precisely what Seung-kook Yun and Ambarish Goswami are working on at the Honda Research Institute USA.
Previously we’ve looked at some of their safe falling strategies, and while these are important, it’s preferable to prevent falling altogether. There are two ways a robot can absorb an unexpected bump, depending on the force involved. If pushed lightly, the robot can simply adjust its posture, such as bending at the waist. If the force is stronger, ASIMO can recover by taking a step. The chosen method, which was presented in simulation by Yun and Goswami at IROS 2011, also works on uneven ground.
A couple of things worth noting: ASIMO is standing completely still before the impacts occur. Second, the simulation model seen in the video is based on the first version of ASIMO, which is still used for research purposes because it would be wasteful not to. It’s unclear how the push-recovery controller works when ASIMO is walking, and it goes without saying that using an out-dated model prevents the researchers from experimenting with All-new ASIMO’s enhanced mechanical properties.
Although ASIMO has slimmed down over the years, there are benefits to having a turtle-like shell on its back. In earlier research done by the same group, ASIMO could prevent damage to its extremities by turning onto its back during a fall, allowing the backpack to absorb the impact. However, the newer versions have decreased the size of the backpack dramatically, which limits the usefulness of this tactic in future models.
However, this is still just the beginning. Ideally, the push-recovery controller would offset impacts from every conceivable direction, and to every area of ASIMO’s body, whether it is standing still, walking, or running. How ASIMO reacts to being hit in the leg, for example, would be very different from how it reacts to being pushed around its midrift. Clearly, more work needs to be done before ASIMO will be ready to leave the lab. Meanwhile, Boston Dynamic’s PETMAN is already showing robust push-recovery while walking, much in the same way BigDog did.
[source: Ambarish Goswami homepage]