The International Journal of Robotics Research Study, Ahead of Publish.
Joint evaluation of realized things posture and external calls is main to durable and dexterous adjustment. In this paper, we present MultiSCOPE, a state-estimation formula that leverages consecutive frictional calls (e.g., jabs) to collectively approximate call places and realized things postures making use of solely proprioception and responsive comments. Our technique addresses the issue of minimizing things posture unpredictability by utilizing 2 corresponding bit filters over a collection of activities: one to approximate call place (CPFGrasp) and one more to approximate things postures (EXTENT). Our technique addresses unpredictability in both robotic proprioception and force-torque dimensions, which is very important for approximating in-hand things posture in the real life. We execute and review our technique on substitute and real-world single-arm and dual-arm robot systems. We show that by bringing 2 items right into call a number of times, the robotics can presume call place and things postures concurrently.
发布者：Andrea Sipos，转转请注明出处：https://robotalks.cn/multiscope-disambiguating-in-hand-object-poses-with-proprioception-and-sequential-interactions/
